<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"
	xmlns:content="http://purl.org/rss/1.0/modules/content/"
	xmlns:wfw="http://wellformedweb.org/CommentAPI/"
	xmlns:dc="http://purl.org/dc/elements/1.1/"
	xmlns:atom="http://www.w3.org/2005/Atom"
	xmlns:sy="http://purl.org/rss/1.0/modules/syndication/"
	xmlns:slash="http://purl.org/rss/1.0/modules/slash/"
	>

<channel>
	<title>AI Archives - OVHcloud Blog</title>
	<atom:link href="https://blog.ovhcloud.com/tag/ai/feed/" rel="self" type="application/rss+xml" />
	<link>https://blog.ovhcloud.com/tag/ai/</link>
	<description>Innovation for Freedom</description>
	<lastBuildDate>Wed, 01 Apr 2026 12:56:38 +0000</lastBuildDate>
	<language>en-GB</language>
	<sy:updatePeriod>
	hourly	</sy:updatePeriod>
	<sy:updateFrequency>
	1	</sy:updateFrequency>
	<generator>https://wordpress.org/?v=6.9.4</generator>

<image>
	<url>https://blog.ovhcloud.com/wp-content/uploads/2019/07/cropped-cropped-nouveau-logo-ovh-rebranding-32x32.gif</url>
	<title>AI Archives - OVHcloud Blog</title>
	<link>https://blog.ovhcloud.com/tag/ai/</link>
	<width>32</width>
	<height>32</height>
</image> 
	<item>
		<title>Extract Text from Images with OCR using Python and OVHcloud AI Endpoints</title>
		<link>https://blog.ovhcloud.com/extract-text-from-images-with-ocr-using-python-and-ovhcloud-ai-endpoints/</link>
		
		<dc:creator><![CDATA[Stéphane Philippart]]></dc:creator>
		<pubDate>Wed, 01 Apr 2026 12:55:19 +0000</pubDate>
				<category><![CDATA[OVHcloud Engineering]]></category>
		<category><![CDATA[AI]]></category>
		<category><![CDATA[AI Endpoints]]></category>
		<guid isPermaLink="false">https://blog.ovhcloud.com/?p=30992</guid>

					<description><![CDATA[If you want to have more information on&#160;AI Endpoints, please read the&#160;following blog post.&#160;You can, also, have a look at our&#160;previous blog posts&#160;on how use AI Endpoints. You can find the full code example in the GitHub repository. In this article,&#160;we will explore how to perform OCR&#160;(Optical Character Recognition)&#160;on images using a vision-capable LLM,&#160;the&#160;OpenAI Python library,&#160;and [&#8230;]<img src="//blog.ovhcloud.com/wp-content/plugins/matomo/app/matomo.php?idsite=1&amp;rec=1&amp;url=https%3A%2F%2Fblog.ovhcloud.com%2Fextract-text-from-images-with-ocr-using-python-and-ovhcloud-ai-endpoints%2F&amp;action_name=Extract%20Text%20from%20Images%20with%20OCR%20using%20Python%20and%20OVHcloud%20AI%20Endpoints&amp;urlref=https%3A%2F%2Fblog.ovhcloud.com%2Ffeed%2F" style="border:0;width:0;height:0" width="0" height="0" alt="" />]]></description>
										<content:encoded><![CDATA[
<p><em>If you want to have more information on&nbsp;<a href="https://endpoints.ai.cloud.ovh.net/" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">AI Endpoints</a>, please read the&nbsp;<a href="https://blog.ovhcloud.com/enhance-your-applications-with-ai-endpoints/" data-wpel-link="internal">following blog post</a>.</em>&nbsp;<em>You can, also, have a look at our&nbsp;<a href="https://blog.ovhcloud.com/tag/ai-endpoints/" data-wpel-link="internal">previous blog posts</a>&nbsp;on how use AI Endpoints.</em></p>



<p><em>You can find the full code example in the <a href="https://github.com/ovh/public-cloud-examples/tree/main/ai/ai-endpoints/python-ocr" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">GitHub repository</a>.</em></p>



<p>In this article,&nbsp;we will explore how to perform OCR&nbsp;(Optical Character Recognition)&nbsp;on images using a vision-capable LLM,&nbsp;the&nbsp;<a href="https://github.com/openai/openai-python" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">OpenAI Python library</a>,&nbsp;and OVHcloud&nbsp;<a href="https://endpoints.ai.cloud.ovh.net/" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">AI Endpoints</a>.</p>



<h3 class="wp-block-heading">Introduction to OCR with Vision Models</h3>



<p>Optical Character Recognition has been around for decades,&nbsp;but traditional OCR engines often struggle with complex layouts,&nbsp;handwritten text,&nbsp;or noisy images.&nbsp;Vision-capable Large Language Models bring a new approach:&nbsp;instead of relying on specialized OCR pipelines,&nbsp;you can simply send an image to a model that understands both visual and textual content.</p>



<p>In this example,&nbsp;we use the&nbsp;<a href="https://github.com/openai/openai-python" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">OpenAI Python library</a>&nbsp;to create a simple OCR script powered by a vision model hosted on OVHcloud&nbsp;<a href="https://endpoints.ai.cloud.ovh.net/" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">AI Endpoints</a>.</p>



<p>The whole application is a single Python file:  no complex setup, just <code><strong>pip install openai</strong></code> and you&#8217;re ready to go.</p>



<h3 class="wp-block-heading">Setting up the Environment Variables</h3>



<p>Before running the script, you need to set the following environment variables:</p>



<pre title="Environment variablesexport OVH_AI_ENDPOINTS_ACCESS_TOKEN=&quot;your-access-token&quot; export OVH_AI_ENDPOINTS_MODEL_URL=&quot;https://your-model-url&quot; export OVH_AI_ENDPOINTS_VLLM_MODEL=&quot;your-vision-model-name&quot;" class="wp-block-code"><code lang="" class=" line-numbers">export OVH_AI_ENDPOINTS_ACCESS_TOKEN="your-access-token"<br>export OVH_AI_ENDPOINTS_MODEL_URL="https://your-model-url"<br>export OVH_AI_ENDPOINTS_VLLM_MODEL="your-vision-model-name"</code></pre>



<p>You can find how to create your access token, model URL, and model name in the <a href="https://endpoints.ai.cloud.ovh.net/catalog" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">AI Endpoints catalog</a>. Make sure to choose a <strong>vision-capable model</strong> from the <a href="https://endpoints.ai.cloud.ovh.net/catalog" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">AI Endpoints catalog</a>.</p>



<h3 class="wp-block-heading">Installing Dependencies</h3>



<p>The only dependency is the OpenAI Python library:</p>



<pre title="OpenAI dependency" class="wp-block-code"><code lang="bash" class="language-bash">pip install openai</code></pre>



<h3 class="wp-block-heading">Define the System Prompt</h3>



<p>The first step is to define a system prompt that describes what our OCR service does.&nbsp;This prompt tells the model how to behave:</p>



<pre title="System prompt" class="wp-block-code"><code lang="" class=" line-numbers">SYSTEM_PROMPT = """You are an expert OCR engine.<br>Extract every piece of text visible in the provided image.<br>Preserve the original layout as faithfully as possible (line breaks, columns, tables).<br>Do NOT interpret, summarise, or translate the content.<br>Use markdown formatting to represent the layout (e.g. tables, lists).<br>If the image contains no text, reply with: "No text found."<br>"""</code></pre>



<p>We tell it to behave as an expert OCR engine, to preserve the original layout, and to use markdown formatting for structured content like tables or lists.<br></p>



<h3 class="wp-block-heading">Load the Image</h3>



<p>Before sending the image to the model,&nbsp;we need to encode it as a base64 string.&nbsp;Here is a simple helper function that reads a local PNG file and returns a base64-encoded string:</p>



<pre title="Image loading" class="wp-block-code"><code lang="" class=" line-numbers">import base64<br>from pathlib import Path<br><br>def load_image_as_base64(path: Path) -&gt; str:<br>    """Load a local image and encode it as base64."""<br>    with open(path, "rb") as f:<br>        return base64.b64encode(f.read()).decode("utf-8")</code></pre>



<p>The base64-encoded data is what gets sent to the vision model as part of the prompt.</p>



<p></p>



<h3 class="wp-block-heading">Extract Text from the Image</h3>



<p>The <code><strong>extract_text</strong></code> function sends the image to the vision model and returns the extracted text:</p>



<pre title="Extract text from image" class="wp-block-code"><code lang="" class=" line-numbers">def extract_text(client: OpenAI, image_base64: str, model: str) -&gt; str:<br>    """Extract text from an image using the vision model."""<br>    response = client.chat.completions.create(<br>        model=model,<br>        temperature=0.0,<br>        messages=[<br>            {"role": "system", "content": SYSTEM_PROMPT},<br>            {<br>                "role": "user",<br>                "content": [<br>                    {<br>                        "type": "image_url",<br>                        "image_url": {<br>                            "url": f"data:image/png;base64,{image_base64}"<br>                        }<br>                    }<br>                ]<br>            }<br>        ]<br>    )<br>    return response.choices[0].message.content</code></pre>



<p>The image is passed as a data URL inside the <code><strong>image_url</strong></code> field, following the OpenAI Vision API format. The temperature is set to <code>0.0</code> because we want deterministic, faithful text extraction and not creative output.</p>



<h3 class="wp-block-heading">Configure the Client</h3>



<p>This example uses a vision-capable model hosted on OVHcloud <a href="https://endpoints.ai.cloud.ovh.net/" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">AI Endpoints</a>. Since AI Endpoints exposes an OpenAI-compatible API, we use the <code>OpenAI</code> client and just point it to the OVHcloud endpoint:</p>



<pre title="Open AI client configuration" class="wp-block-code"><code lang="" class=" line-numbers">import os<br>from openai import OpenAI<br><br>client = OpenAI(<br>    api_key=os.getenv("OVH_AI_ENDPOINTS_ACCESS_TOKEN"),<br>    base_url=os.getenv("OVH_AI_ENDPOINTS_MODEL_URL"),<br>)<br><br>model_name = os.getenv("OVH_AI_ENDPOINTS_VLLM_MODEL")</code></pre>



<p>A few things to note:</p>



<ul class="wp-block-list">
<li>The <strong>API key</strong>, <strong>base URL</strong>, and <strong>model name</strong> are read from environment variables. </li>



<li>The OpenAI library is compatible with any OpenAI compatible API, making it perfect for use with AI Endpoints.</li>
</ul>



<h3 class="wp-block-heading">Assemble and Run</h3>



<p>With the client configured, extracting text from an image is straightforward:</p>



<pre title="Run the OCR" class="wp-block-code"><code lang="" class=" line-numbers">image_base64 = load_image_as_base64(Path("./doc.png"))<br>result = extract_text(client, image_base64, model_name)<br>print(result)</code></pre>



<p>And that&#8217;s it!</p>



<p>Here is the image used for this example:</p>



<figure class="wp-block-image aligncenter size-full is-resized"><img fetchpriority="high" decoding="async" width="946" height="693" src="https://blog.ovhcloud.com/wp-content/uploads/2026/03/doc-1.png" alt="Used image for OCR example" class="wp-image-31002" style="width:600px" srcset="https://blog.ovhcloud.com/wp-content/uploads/2026/03/doc-1.png 946w, https://blog.ovhcloud.com/wp-content/uploads/2026/03/doc-1-300x220.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2026/03/doc-1-768x563.png 768w" sizes="(max-width: 946px) 100vw, 946px" /></figure>



<p>And the result:</p>



<pre title="Run the OCR" class="wp-block-code"><code lang="" class=" line-numbers">$ python ocr_demo.py<br>📄 Loading image: doc.png<br>🔍 Running OCR with Qwen2.5-VL-72B-Instruct via OVHcloud AI Endpoints...<br><br>📝 Extracted text 📝<br>Every month, the OVHcloud Developer Advocate team creates content, shares knowledge, and connects with the tech community. Here’s a look at what we did in March 2026. 🚀<br><br>🎙️ “Tranches de Tech” – Our monthly podcast<br><br>A new episode of our French-language podcast Tranches de Tech🥑 just dropped!<br><br>🎧 Episode 102: Tranches de Tech #26 – Architecte, c’est une bonne situation ça ?<br><br>This month we sat down with Alexandre Touret, Architect at Worldline to discuss the evolving role of software architects and the growing impact of AI on development practices. From Spotify’s claim that their devs no longer code, to agentic tools like OpenClaw and Claude Code reshaping workflows. We also cover ANSSI’s revised open-source policy, IBM tripling junior hires, and the critical responsibility of mentoring the next generation of developers in an AI-driven world.<br><br>📺 Live on Twitch<br><br>We streamed live on Twitch this month! Here’s what we covered:<br><br>🎥 Rémy Vandepoel discussed with Hugo Allabert and François Loiseau about our Public VCFaaS. Catch the replay on YouTube ▶️.<br><br>🎤 Conference Talks<br><br>The team hit the road (and the stage) at several conferences this month:<br><br>🇳🇱 KubeCon Amsterdam – Amsterdam, Netherlands 🇳🇱<br><br>Aurélie Vache gave a talk: The Ultimate Kubernetes Challenge: An Interactive Trivia Game</code></pre>



<h3 class="wp-block-heading">Conclusion</h3>



<p>In this article,&nbsp;we have seen how to use a vision-capable LLM to perform OCR on images using the&nbsp;<a href="https://github.com/openai/openai-python" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">OpenAI Python library</a>&nbsp;and OVHcloud&nbsp;<a href="https://endpoints.ai.cloud.ovh.net/" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">AI Endpoints</a>.&nbsp;The OpenAI library makes it very easy to send images to a vision model and extract text,&nbsp;and Python allows us to run the whole thing as a simple script.</p>



<p>You have a dedicated Discord channel&nbsp;(#<em>ai-endpoints</em>)&nbsp;on our Discord server&nbsp;(<em><a href="https://discord.gg/ovhcloud" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">https://discord.gg/ovhcloud</a></em>),&nbsp;see you there!</p>



<p></p>
<img decoding="async" src="//blog.ovhcloud.com/wp-content/plugins/matomo/app/matomo.php?idsite=1&amp;rec=1&amp;url=https%3A%2F%2Fblog.ovhcloud.com%2Fextract-text-from-images-with-ocr-using-python-and-ovhcloud-ai-endpoints%2F&amp;action_name=Extract%20Text%20from%20Images%20with%20OCR%20using%20Python%20and%20OVHcloud%20AI%20Endpoints&amp;urlref=https%3A%2F%2Fblog.ovhcloud.com%2Ffeed%2F" style="border:0;width:0;height:0" width="0" height="0" alt="" />]]></content:encoded>
					
		
		
			</item>
		<item>
		<title>Pricing evolution of Public Cloud, Bare Metal and VPS at OVHcloud</title>
		<link>https://blog.ovhcloud.com/pricing-evolution-of-public-cloud-bare-metal-and-vps-at-ovhcloud/</link>
		
		<dc:creator><![CDATA[Octave Klaba]]></dc:creator>
		<pubDate>Thu, 05 Mar 2026 12:59:25 +0000</pubDate>
				<category><![CDATA[OVHcloud Product News]]></category>
		<category><![CDATA[Uncategorized]]></category>
		<category><![CDATA[AI]]></category>
		<category><![CDATA[bare metal]]></category>
		<category><![CDATA[Infrastructure]]></category>
		<category><![CDATA[OVHcloud]]></category>
		<category><![CDATA[OVHcloud News]]></category>
		<category><![CDATA[Private Cloud]]></category>
		<category><![CDATA[Public Cloud]]></category>
		<category><![CDATA[Storage]]></category>
		<guid isPermaLink="false">https://blog.ovhcloud.com/?p=30724</guid>

					<description><![CDATA[For customers in the United States, the same article with US pricing is available here : https://us.ovhcloud.com/resources/blog/pricing-evolution-of-public-cloud-bare-metal-and-vps-at-ovhcloud/ Since autumn 2025, the global memory market has been going through a major disruption. Although barely noticeable to end users, these developments are radically changing the cost of computer hardware and, as a direct result, the cost of [&#8230;]<img src="//blog.ovhcloud.com/wp-content/plugins/matomo/app/matomo.php?idsite=1&amp;rec=1&amp;url=https%3A%2F%2Fblog.ovhcloud.com%2Fpricing-evolution-of-public-cloud-bare-metal-and-vps-at-ovhcloud%2F&amp;action_name=Pricing%20evolution%20of%20Public%20Cloud%2C%20Bare%20Metal%20and%20VPS%20at%20OVHcloud&amp;urlref=https%3A%2F%2Fblog.ovhcloud.com%2Ffeed%2F" style="border:0;width:0;height:0" width="0" height="0" alt="" />]]></description>
										<content:encoded><![CDATA[
<p><em><em>For customers in the United States, the same article with <strong>US pricing</strong> is available here</em> : <a href="https://us.ovhcloud.com/resources/blog/pricing-evolution-of-public-cloud-bare-metal-and-vps-at-ovhcloud/" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">https://us.ovhcloud.com/resources/blog/pricing-evolution-of-public-cloud-bare-metal-and-vps-at-ovhcloud/</a></em></p>



<p>Since autumn 2025, the global memory market has been going through a major disruption. Although barely noticeable to end users, these developments are radically changing the cost of computer hardware and, as a direct result, the cost of the cloud.<br><br>This article will decipher this structural crisis, its real-life impacts, and the strategic choices that OVHcloud is implementing to mitigate its effects.</p>



<h2 class="wp-block-heading"><strong>An industrial shift towards GPUs</strong></h2>



<p>Globally, the three major memory manufacturers have redirected a significant portion of their production capacity to meet the massive demand for GPUs, particularly for AI-related and high-bandwidth computing applications.</p>



<p>This reallocation took place without a corresponding reduction in the historical demand for RAM and storage, generating pressure on several market segments simultaneously.<br><br>The consequences of this were immediate and noticeable:</p>



<ul class="wp-block-list">
<li>pressure on supply, with reduced stock and extended lead times</li>



<li>continuous rise in RAM and disk prices since September 2025</li>



<li>long-term market instability, which is not expected to find a new balance until late 2026</li>
</ul>



<h2 class="wp-block-heading"><strong>A sustained inflation of memory components</strong></h2>



<p>Even after the market stabilises, prices are not expected to return to their historical levels before 2028, the amount of time needed for new production capacities to become truly operational.</p>



<p>This development profoundly disrupts the economic fundamentals of computer hardware, both for on-premises infrastructures and for the cloud. Depending on configurations, the prices related to RAM and storage could increase by 15% to 300% compared to 2025 prices, depending on the volumes of memory and disk capacity deployed.</p>



<p>This change of scale is both abrupt and unprecedented, with no recent equivalent in the global market.</p>



<h2 class="wp-block-heading"><strong>A market under pressure, even with higher prices</strong></h2>



<p>Paradoxically, the rise in prices is not enough to secure the availability of components. Currently, to guarantee the delivery of their desired volume of RAM or disks, cloud providers need to order up to 12 months in advance, without being told the final price at the time of purchase.</p>



<p>In practice, prices are only communicated one to two months after delivery, depending on the changes in supply and demand during the quarter in question. This uncertainty places unprecedented pressure on industry players and cloud providers, simultaneously affecting production and distribution.</p>



<h2 class="wp-block-heading"><strong>Towards a new global balance of demand</strong></h2>



<p>This situation will inevitably have repercussions on the volumes ordered. Some customers will find the prices too high and limit their investments, while others, lacking alternatives, will continue to place orders regardless.</p>



<p>This interplay of opposing forces should lead to a new global balance, but at a significantly higher price point. Current projections anticipate a 250% to 300% increase in the price of RAM by the end of 2026, compared to September 2025.</p>



<h2 class="wp-block-heading"><strong>Our strategy to soften the blow</strong></h2>



<p>In light of this reality, OVHcloud has chosen not to automatically pass on the entire price increase of components to its customers.</p>



<p>For the cloud deployed between 2026 and 2028 (including Public Cloud, Private Cloud, and Bare Metal), the average price increase will be limited – between 9% and 11% – despite significantly higher RAM and disk costs.</p>



<p>To offset this gap, a moderate increase of 2% to 6% is planned for solutions deployed before 2025, depending on the age of the equipment, as well as a change in IPv4 pricing. The latter should not have a significant impact on our customers’ budgets, as the cost of IP addresses is a small share compared to other resources in a cloud project.</p>



<p>Our objective is clear: to maintain pricing consistency across the entire range from 2021 to 2028, and to prepare for a gradual return to normal in 2029.</p>



<h2 class="wp-block-heading"><strong>Continuous investments and developing solutions</strong></h2>



<p>Beyond pricing adjustments, this period will be characterised by sustained investments in our solutions and in the customer experience.<br><br>Despite the strong pressure from rising component costs, we are continuing to develop our services to provide more value to our customers.<br><br>In practical terms, this will result in:</p>



<ul class="wp-block-list">
<li>a gradual strengthening of support mechanisms</li>



<li>an increase in resources included in certain ranges</li>



<li>a modernization of our computing and storage infrastructures</li>
</ul>



<p>These initiatives demonstrate our commitment to not reduce this period to merely a consequence of cost increases, but to maintain a dynamic of improving our services, even in a constrained economic context.</p>



<h2 class="wp-block-heading"><strong>Time frame and implementation procedures</strong></h2>



<p>Our clients have already received emails detailing the precise impacts on their services. The new prices will come into effect on 1 April 2026.</p>



<p>Until that date, it is possible to renew services at the current rates for a duration of up to 2 years. In all cases, the new prices will only apply at the end of the current contractual period.</p>



<h2 class="wp-block-heading"><strong>A time of uncertainty and a strategic advantage</strong></h2>



<p>We are going through an exceptionally unpredictable period, where market visibility rarely lasts longer than one to two weeks. There remains hope that prices will stabilize on a long-term basis from 2026, so that we can avoid further unfavorable announcements. <br>In this tense context, having a global supply chain and two internal production facilities is a major strategic advantage. This allows us to continue receiving components and producing servers, while the memory shortage affects a large part of the market.</p>



<h2 class="wp-block-heading"><strong>Our Prices</strong></h2>



<p>You will find below our new pricing:<br>&#8211; <strong>Public Cloud:</strong> Prices below are displayed on an hourly basis and with Linux OS. Please, find on our <a href="https://www.ovhcloud.com/en-ie/public-cloud/prices/" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">Prices web page</a> our monthly-consumed virtual machine instances (b2, c2, r2) and <a href="https://www.ovhcloud.com/en-gb/public-cloud/savings-plan/" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">Savings Plan</a> options (b3, c3, r3) as well as prices with Windows licences.<br>&#8211; All our <strong>VPS, Floating IPs, and Additional IP</strong> pricing.<br>&#8211; <strong>Bare Metal</strong>: The displayed prices correspond to a 1-month commitment; additional discounts apply for 12- or 24-month prepayments. <strong>The prices for options are for new orders only</strong>. The renewal of options, which has been communicated by email to our customers, will be limited to +10% for disk options and +15% for RAM options.<br><br>For existing subscriptions renewed <strong>before April 1st</strong>, you can secure your current pricing for the full duration of the commitment you choose, effective from your renewal date.<br></p>



<ul class="wp-block-list">
<li></li>
</ul>



<p><span class="wp-rich-text-font-awesome-icon wp-font-awesome-icon"><svg aria-hidden="true" focusable="false" data-prefix="fas" data-icon="bell" class="svg-inline--fa fa-bell " role="img" xmlns="http://www.w3.org/2000/svg" viewBox="0 0 448 512"><path fill="currentColor" d="M224 0c-17.7 0-32 14.3-32 32V51.2C119 66 64 130.6 64 208v18.8c0 47-17.3 92.4-48.5 127.6l-7.4 8.3c-8.4 9.4-10.4 22.9-5.3 34.4S19.4 416 32 416H416c12.6 0 24-7.4 29.2-18.9s3.1-25-5.3-34.4l-7.4-8.3C401.3 319.2 384 273.9 384 226.8V208c0-77.4-55-142-128-156.8V32c0-17.7-14.3-32-32-32zm45.3 493.3c12-12 18.7-28.3 18.7-45.3H224 160c0 17 6.7 33.3 18.7 45.3s28.3 18.7 45.3 18.7s33.3-6.7 45.3-18.7z"></path></svg></span> Please note that the following product categories <strong><span style="text-decoration: underline;">are not affected</span></strong> by our pricing evolution:<br>&#8211; Public Cloud &#8211; Compute : <strong>Cloud GPUs</strong> <strong>and Metal Instances</strong><br>&#8211; Public Cloud &#8211; Container : <strong>Managed Kubernetes, Managed Registries &amp; Managed Rancher</strong><br>&#8211; Public Cloud &#8211; Network : <strong>Load Balancer, Gateway</strong>. Public and Private network traffic remains included.<br>&#8211; Public Cloud &#8211; Storage : <strong>Object Storage, Block Storage</strong>. <br>&#8211; Public Cloud &#8211; Analytics : <strong>Data Platform</strong><br>&#8211; Public Cloud &#8211; AI &amp; Machine Learning : <strong>AI Solutions (AI Notebook, AI Training, AI Deploy) and AI Endpoints</strong><br>&#8211; Public Cloud &#8211; Quantum : <strong>Emulators &amp; QPUs</strong><br>&#8211; Bare Metal : <strong>Kimsufi et SoYouStart</strong> ranges<br>&#8211; Bare Metal : <strong>All storage</strong> (Veeam Enterprise plus, HYCU, Back-up Agent, NAS-HA, Cloud Disk Array)<br>&#8211; Private Cloud : <strong>All VMware</strong> offers, all <strong>storage offers</strong> (Veeam Enterprise plus, HYCU, Back-up Agent)<br></p>



<!DOCTYPE html>
<html lang="fr">
<head>
  <meta charset="UTF-8">
  <title>Tableaux des prix</title>
  <style>
    body { font-family: Arial, Helvetica, sans-serif; margin: 40px; background-color: #f4f6f8; }
    h1 { color: #000E9C; }
    h2 { margin-top: 50px; color: #1a2b3c; }
    h3 { margin-top: 30px; color: #1a2b3c; }
    .table-container { margin-top: 20px; margin-bottom: 40px; }
    table { border-collapse: collapse; width: 100%; background: white; border-radius: 8px; overflow: hidden; box-shadow: 0 2px 8px rgba(0,0,0,0.08); }
    thead { background-color: #000E9C; color: white; }
    th { padding: 12px 15px; text-align: center; font-weight: bold; text-transform: uppercase; font-size: 13px; }
    td { padding: 12px 15px; font-size: 14px; }
    td:nth-child(n+2) { text-align: center; }
    tbody tr { border-bottom: 1px solid #e6e9ed; }
    tbody tr:nth-child(even) { background-color: #f8fafc; }
    tbody tr:hover { background-color: #eef3ff; }
    .section-header td { background-color: #e8edf5; font-weight: bold; font-style: italic; color: #1a2b3c; text-align: left !important; }
  </style>
</head>
<body>

<!-- PUBLIC CLOUD -->
<div class='table-container'>
  <h2>Public Cloud – Virtual Machine Instances</h2>

  <h3>General Purpose</h3>
 <p class="mention">These are the standard hourly &#038; monthly price for Linux version of the instances, without Savings Plan or any other additional discount.</p>
  <table class="dataframe">
    <thead>
      <tr><th>Reference</th><th>Old public price (Excl. VAT / Hour)</th><th>New Public Price (Excl. VAT / Hour)</th></tr>
    </thead>
    <tbody>
      <tr><td>b3-8</td><td>0,0465 €</td><td>0,0512 €</td></tr>
      <tr><td>b3-16</td><td>0,093 €</td><td>0,1023 €</td></tr>
      <tr><td>b3-32</td><td>0,186 €</td><td>0,2046 €</td></tr>
      <tr><td>b3-64</td><td>0,372 €</td><td>0,4092 €</td></tr>
      <tr><td>b3-128</td><td>0,7439 €</td><td>0,819 €</td></tr>
      <tr><td>b3-256</td><td>1,4878 €</td><td>1,637 €</td></tr>
      <tr><td>b3-512</td><td>2,9756 €</td><td>3,274 €</td></tr>
      <tr><td>b3-640</td><td>3,7195 €</td><td>4,092 €</td></tr>
      <tr><td>b2-7</td><td>0,0681 €</td><td>0,0709 €</td></tr>
      <tr><td>b2-15</td><td>0,129 €</td><td>0,1342 €</td></tr>
      <tr><td>b2-30</td><td>0,261 €</td><td>0,2715 €</td></tr>
      <tr><td>b2-60</td><td>0,505 €</td><td>0,526 €</td></tr>
      <tr><td>b2-120</td><td>0,993 €</td><td>1,033 €</td></tr>
    </tbody>
  </table>


  <h3>Compute Optimized</h3>
 <p class="mention">These are the standard hourly &#038; monthly price for Linux version of the instances, without Savings Plan or any other additional discount.</p>
  <table class="dataframe">
  <table class="dataframe">
    <thead>
      <tr><th>Reference</th><th>Old public price (Excl. VAT / Hour)</th><th>New Public Price (Excl. VAT / Hour)</th></tr>
    </thead>
    <tbody>
      <tr><td>c3-4</td><td>0,0415 €</td><td>0,0457 €</td></tr>
      <tr><td>c3-8</td><td>0,083 €</td><td>0,0913 €</td></tr>
      <tr><td>c3-16</td><td>0,1659 €</td><td>0,1825 €</td></tr>
      <tr><td>c3-32</td><td>0,3318 €</td><td>0,365 €</td></tr>
      <tr><td>c3-64</td><td>0,6637 €</td><td>0,7301 €</td></tr>
      <tr><td>c3-128</td><td>1,3274 €</td><td>1,461 €</td></tr>
      <tr><td>c3-256</td><td>2,6547 €</td><td>2,921 €</td></tr>
      <tr><td>c3-320</td><td>3,3184 €</td><td>3,651 €</td></tr>
      <tr><td>c2-7</td><td>0,0978 €</td><td>0,1018 €</td></tr>
      <tr><td>c2-15</td><td>0,19 €</td><td>0,1976 €</td></tr>
      <tr><td>c2-30</td><td>0,383 €</td><td>0,3984 €</td></tr>
      <tr><td>c2-60</td><td>0,749 €</td><td>0,779 €</td></tr>
      <tr><td>c2-120</td><td>1,48 €</td><td>1,54 €</td></tr>
    </tbody>
  </table>

  <h3>Memory Optimized</h3>
 <p class="mention">These are the standard hourly &#038; monthly price for Linux version of the instances, without Savings Plan or any other additional discount.</p>
  <table class="dataframe">
  <table class="dataframe">
    <thead>
      <tr><th>Reference</th><th>Old public price (Excl. VAT / Hour)</th><th>New Public Price (Excl. VAT / Hour)</th></tr>
    </thead>
    <tbody>
      <tr><td>r3-16</td><td>0,0602 €</td><td>0,0663 €</td></tr>
      <tr><td>r3-32</td><td>0,1203 €</td><td>0,1324 €</td></tr>
      <tr><td>r3-64</td><td>0,2407 €</td><td>0,2648 €</td></tr>
      <tr><td>r3-128</td><td>0,4813 €</td><td>0,53 €</td></tr>
      <tr><td>r3-256</td><td>0,9627 €</td><td>1,059 €</td></tr>
      <tr><td>r3-512</td><td>1,9254 €</td><td>2,118 €</td></tr>
      <tr><td>r3-1024</td><td>3,8508 €</td><td>4,236 €</td></tr>
      <tr><td>r2-15</td><td>0,0978 €</td><td>0,1018 €</td></tr>
      <tr><td>r2-30</td><td>0,113 €</td><td>0,1176 €</td></tr>
      <tr><td>r2-60</td><td>0,22 €</td><td>0,2288 €</td></tr>
      <tr><td>r2-120</td><td>0,443 €</td><td>0,461 €</td></tr>
      <tr><td>r2-240</td><td>0,871 €</td><td>0,906 €</td></tr>
    </tbody>
  </table>

</div>

<!-- PUBLIC CLOUD DATABASES -->
<div class='table-container'>
  <h2>Public Cloud – Databases</h2>

  <h3>MySQL</h3>
  <table class="dataframe">
    <thead>
      <tr><th>Reference</th><th>Old public price (Excl. VAT / Hour / Node)</th><th>New Public Price (Excl. VAT / Hour / Node)</th><th>New Public Price (Excl. VAT / Hour)</th></tr>
    </thead>
    <tbody>
      <tr><td>Essential DB1-4</td><td>0,068 €</td><td>0,0746 €</td><td>0,0746 €</td></tr>
      <tr><td>Essential DB1-7</td><td>0,1346 €</td><td>0,1477 €</td><td>0,1477 €</td></tr>
      <tr><td>Essential DB1-15</td><td>0,2705 €</td><td>0,2968 €</td><td>0,2968 €</td></tr>
      <tr><td>Essential DB1-30</td><td>0,5436 €</td><td>0,5967 €</td><td>0,5967 €</td></tr>
      <tr><td>Production B3-8</td><td>0,2129 €</td><td>0,223 €</td><td>0,446 €</td></tr>
      <tr><td>Production B3-16</td><td>0,4258 €</td><td>0,4461 €</td><td>0,8922 €</td></tr>
      <tr><td>Production B3-32</td><td>0,8515 €</td><td>0,8922 €</td><td>1,7844 €</td></tr>
      <tr><td>Production B3-64</td><td>1,703 €</td><td>1,7844 €</td><td>3,5688 €</td></tr>
      <tr><td>Production B3-128</td><td>3,4059 €</td><td>3,5688 €</td><td>7,1376 €</td></tr>
      <tr><td>Production B3-256</td><td>6,8118 €</td><td>7,1377 €</td><td>14,2754 €</td></tr>
      <tr><td>Business DB1-4</td><td>0,0865 €</td><td>0,0949 €</td><td>0,1898 €</td></tr>
      <tr><td>Business DB1-7</td><td>0,173 €</td><td>0,1899 €</td><td>0,3798 €</td></tr>
      <tr><td>Business DB1-15</td><td>0,346 €</td><td>0,3797 €</td><td>0,7594 €</td></tr>
      <tr><td>Business DB1-30</td><td>0,6933 €</td><td>0,761 €</td><td>1,522 €</td></tr>
      <tr><td>Business DB1-60</td><td>1,3878 €</td><td>1,5234 €</td><td>3,0468 €</td></tr>
      <tr><td>Business DB1-120</td><td>2,777 €</td><td>3,0484 €</td><td>6,0968 €</td></tr>
      <tr><td>Advanced B3-8</td><td>0,2295 €</td><td>0,2404 €</td><td>0,7212 €</td></tr>
      <tr><td>Advanced B3-16</td><td>0,4589 €</td><td>0,4808 €</td><td>1,4424 €</td></tr>
      <tr><td>Advanced B3-32</td><td>0,9177 €</td><td>0,9616 €</td><td>2,8848 €</td></tr>
      <tr><td>Advanced B3-64</td><td>1,8354 €</td><td>1,9232 €</td><td>5,7696 €</td></tr>
      <tr><td>Advanced B3-128</td><td>3,6708 €</td><td>3,8464 €</td><td>11,5392 €</td></tr>
      <tr><td>Advanced B3-256</td><td>7,3416 €</td><td>7,6928 €</td><td>23,0784 €</td></tr>
      <tr><td>Enterprise DB1-4</td><td>0,0879 €</td><td>0,0964 €</td><td>0,2892 €</td></tr>
      <tr><td>Enterprise DB1-7</td><td>0,173 €</td><td>0,1899 €</td><td>0,5697 €</td></tr>
      <tr><td>Enterprise DB1-15</td><td>0,346 €</td><td>0,3797 €</td><td>1,1391 €</td></tr>
      <tr><td>Enterprise DB1-30</td><td>0,6933 €</td><td>0,761 €</td><td>2,283 €</td></tr>
      <tr><td>Enterprise DB1-60</td><td>1,3878 €</td><td>1,5234 €</td><td>4,5702 €</td></tr>
      <tr><td>Enterprise DB1-120</td><td>2,777 €</td><td>3,0484 €</td><td>9,1452 €</td></tr>
    </tbody>
  </table>

  <h3>PostgreSQL</h3>
  <table class="dataframe">
    <thead>
      <tr><th>Reference</th><th>Old public price (Excl. VAT / Hour / Node)</th><th>New Public Price (Excl. VAT / Hour / Node)</th><th>New Public Price (Excl. VAT / Hour)</th></tr>
    </thead>
    <tbody>
      <tr><td>Essential DB1-4</td><td>0,068 €</td><td>0,0746 €</td><td>0,0746 €</td></tr>
      <tr><td>Essential DB1-7</td><td>0,1346 €</td><td>0,1477 €</td><td>0,1477 €</td></tr>
      <tr><td>Essential DB1-15</td><td>0,2705 €</td><td>0,2968 €</td><td>0,2968 €</td></tr>
      <tr><td>Essential DB1-30</td><td>0,5436 €</td><td>0,5967 €</td><td>0,5967 €</td></tr>
      <tr><td>Production B3-8</td><td>0,2129 €</td><td>0,223 €</td><td>0,446 €</td></tr>
      <tr><td>Production B3-16</td><td>0,4258 €</td><td>0,4461 €</td><td>0,8922 €</td></tr>
      <tr><td>Production B3-32</td><td>0,8515 €</td><td>0,8922 €</td><td>1,7844 €</td></tr>
      <tr><td>Production B3-64</td><td>1,703 €</td><td>1,7844 €</td><td>3,5688 €</td></tr>
      <tr><td>Production B3-128</td><td>3,4059 €</td><td>3,5688 €</td><td>7,1376 €</td></tr>
      <tr><td>Production B3-256</td><td>6,8118 €</td><td>7,1377 €</td><td>14,2754 €</td></tr>
      <tr><td>Business DB1-4</td><td>0,0865 €</td><td>0,0949 €</td><td>0,1898 €</td></tr>
      <tr><td>Business DB1-7</td><td>0,173 €</td><td>0,1899 €</td><td>0,3798 €</td></tr>
      <tr><td>Business DB1-15</td><td>0,346 €</td><td>0,3797 €</td><td>0,7594 €</td></tr>
      <tr><td>Business DB1-30</td><td>0,6933 €</td><td>0,761 €</td><td>1,522 €</td></tr>
      <tr><td>Business DB1-60</td><td>1,3878 €</td><td>1,5234 €</td><td>3,0468 €</td></tr>
      <tr><td>Business DB1-120</td><td>2,777 €</td><td>3,0484 €</td><td>6,0968 €</td></tr>
      <tr><td>Advanced B3-8</td><td>0,2295 €</td><td>0,2404 €</td><td>0,7212 €</td></tr>
      <tr><td>Advanced B3-16</td><td>0,4589 €</td><td>0,4808 €</td><td>1,4424 €</td></tr>
      <tr><td>Advanced B3-32</td><td>0,9177 €</td><td>0,9616 €</td><td>2,8848 €</td></tr>
      <tr><td>Advanced B3-64</td><td>1,8354 €</td><td>1,9232 €</td><td>5,7696 €</td></tr>
      <tr><td>Advanced B3-128</td><td>3,6708 €</td><td>3,8464 €</td><td>11,5392 €</td></tr>
      <tr><td>Advanced B3-256</td><td>7,3416 €</td><td>7,6928 €</td><td>23,0784 €</td></tr>
      <tr><td>Enterprise DB1-4</td><td>0,0879 €</td><td>0,0964 €</td><td>0,2892 €</td></tr>
      <tr><td>Enterprise DB1-7</td><td>0,173 €</td><td>0,1899 €</td><td>0,5697 €</td></tr>
      <tr><td>Enterprise DB1-15</td><td>0,346 €</td><td>0,3797 €</td><td>1,1391 €</td></tr>
      <tr><td>Enterprise DB1-30</td><td>0,6933 €</td><td>0,761 €</td><td>2,283 €</td></tr>
      <tr><td>Enterprise DB1-60</td><td>1,3878 €</td><td>1,5234 €</td><td>4,5702 €</td></tr>
      <tr><td>Enterprise DB1-120</td><td>2,777 €</td><td>3,0484 €</td><td>9,1452 €</td></tr>
    </tbody>
  </table>

  <h3>Valkey</h3>
  <table class="dataframe">
    <thead>
      <tr><th>Reference</th><th>Old public price (Excl. VAT / Hour / Node)</th><th>New Public Price (Excl. VAT / Hour / Node)</th><th>New Public Price (Excl. VAT / Hour)</th></tr>
    </thead>
    <tbody>
      <tr><td>Essential DB1-4</td><td>0,0591 €</td><td>0,0648 €</td><td>0,0648 €</td></tr>
      <tr><td>Essential DB1-7</td><td>0,1195 €</td><td>0,1311 €</td><td>0,1311 €</td></tr>
      <tr><td>Production B3-8</td><td>0,1409 €</td><td>0,1476 €</td><td>0,2952 €</td></tr>
      <tr><td>Production B3-16</td><td>0,3147 €</td><td>0,3297 €</td><td>0,6594 €</td></tr>
      <tr><td>Production B3-32</td><td>0,6295 €</td><td>0,6595 €</td><td>1,319 €</td></tr>
      <tr><td>Production B3-64</td><td>1,2588 €</td><td>1,319 €</td><td>2,638 €</td></tr>
      <tr><td>Production B3-128</td><td>2,5175 €</td><td>2,6379 €</td><td>5,2758 €</td></tr>
      <tr><td>Production B3-256</td><td>5,0349 €</td><td>5,2757 €</td><td>10,5514 €</td></tr>
      <tr><td>Business DB1-4</td><td>0,068 €</td><td>0,0746 €</td><td>0,1492 €</td></tr>
      <tr><td>Business DB1-7</td><td>0,151 €</td><td>0,1658 €</td><td>0,3316 €</td></tr>
      <tr><td>Business DB1-15</td><td>0,2252 €</td><td>0,2471 €</td><td>0,4942 €</td></tr>
      <tr><td>Business DB1-30</td><td>0,4448 €</td><td>0,4882 €</td><td>0,9764 €</td></tr>
      <tr><td>Business DB1-60</td><td>0,8895 €</td><td>0,9764 €</td><td>1,9528 €</td></tr>
      <tr><td>Business DB1-120</td><td>1,7736 €</td><td>1,9468 €</td><td>3,8936 €</td></tr>
    </tbody>
  </table>

  <h3>Kafka</h3>
  <table class="dataframe">
    <thead>
      <tr><th>Reference</th><th>Old public price (Excl. VAT / Hour / Node)</th><th>New Public Price (Excl. VAT / Hour / Node)</th><th>New Public Price (Excl. VAT / Hour)</th></tr>
    </thead>
    <tbody>
      <tr><td>Production B3-8</td><td>0,2656 €</td><td>0,2782 €</td><td>0,8346 €</td></tr>
      <tr><td>Production B3-16</td><td>0,5311 €</td><td>0,5565 €</td><td>1,6695 €</td></tr>
      <tr><td>Production B3-32</td><td>1,0622 €</td><td>1,113 €</td><td>3,339 €</td></tr>
      <tr><td>Business DB1-4</td><td>0,1469 €</td><td>0,1612 €</td><td>0,4836 €</td></tr>
      <tr><td>Business DB1-7</td><td>0,2911 €</td><td>0,3195 €</td><td>0,9585 €</td></tr>
      <tr><td>Business DB1-15</td><td>0,5532 €</td><td>0,6073 €</td><td>1,8219 €</td></tr>
      <tr><td>Business DB1-30</td><td>1,0707 €</td><td>1,1753 €</td><td>3,5259 €</td></tr>
      <tr><td>Business DB1-60</td><td>2,1428 €</td><td>2,3522 €</td><td>7,0566 €</td></tr>
      <tr><td>Advanced B3-8</td><td>0,2656 €</td><td>0,2782 €</td><td>1,6692 €</td></tr>
      <tr><td>Advanced B3-16</td><td>0,5311 €</td><td>0,5565 €</td><td>3,339 €</td></tr>
      <tr><td>Advanced B3-32</td><td>1,0622 €</td><td>1,113 €</td><td>6,678 €</td></tr>
      <tr><td>Enterprise DB1-7</td><td>0,2924 €</td><td>0,321 €</td><td>1,926 €</td></tr>
      <tr><td>Enterprise DB1-15</td><td>0,5532 €</td><td>0,6073 €</td><td>3,6438 €</td></tr>
      <tr><td>Enterprise DB1-30</td><td>1,0707 €</td><td>1,1753 €</td><td>7,0518 €</td></tr>
      <tr><td>Enterprise DB1-60</td><td>2,1428 €</td><td>2,3522 €</td><td>14,1132 €</td></tr>
    </tbody>
  </table>

  <h3>Kafka Connect</h3>
  <table class="dataframe">
    <thead>
      <tr><th>Reference</th><th>Old public price (Excl. VAT / Hour / Node)</th><th>New Public Price (Excl. VAT / Hour / Node)</th><th>New Public Price (Excl. VAT / Hour)</th></tr>
    </thead>
    <tbody>
      <tr><td>Essential DB1-4</td><td>0,1044 €</td><td>0,1145 €</td><td>0,1145 €</td></tr>
      <tr><td>Essential DB1-7</td><td>0,2101 €</td><td>0,2305 €</td><td>0,2305 €</td></tr>
      <tr><td>Essential DB1-15</td><td>0,3913 €</td><td>0,4295 €</td><td>0,4295 €</td></tr>
      <tr><td>Essential DB1-30</td><td>0,7084 €</td><td>0,7775 €</td><td>0,7775 €</td></tr>
      <tr><td>Production B3-8</td><td>0,1917 €</td><td>0,2008 €</td><td>0,6024 €</td></tr>
      <tr><td>Production B3-16</td><td>0,3862 €</td><td>0,4046 €</td><td>1,2138 €</td></tr>
      <tr><td>Production B3-32</td><td>0,7027 €</td><td>0,7363 €</td><td>2,2089 €</td></tr>
      <tr><td>Business DB1-7</td><td>0,2101 €</td><td>0,2305 €</td><td>0,6915 €</td></tr>
      <tr><td>Business DB1-15</td><td>0,4022 €</td><td>0,4415 €</td><td>1,3245 €</td></tr>
      <tr><td>Business DB1-30</td><td>0,7084 €</td><td>0,7775 €</td><td>2,3325 €</td></tr>
      <tr><td>Advanced B3-8</td><td>0,1908 €</td><td>0,1999 €</td><td>1,1994 €</td></tr>
      <tr><td>Advanced B3-16</td><td>0,3862 €</td><td>0,4046 €</td><td>2,4276 €</td></tr>
      <tr><td>Advanced B3-32</td><td>0,7027 €</td><td>0,7363 €</td><td>4,4178 €</td></tr>
      <tr><td>Enterprise DB1-7</td><td>0,2101 €</td><td>0,2305 €</td><td>1,383 €</td></tr>
      <tr><td>Enterprise DB1-15</td><td>0,4022 €</td><td>0,4415 €</td><td>2,649 €</td></tr>
      <tr><td>Enterprise DB1-30</td><td>0,7084 €</td><td>0,7775 €</td><td>4,665 €</td></tr>
    </tbody>
  </table>

  <h3>Kafka Mirror Maker</h3>
  <table class="dataframe">
    <thead>
      <tr><th>Reference</th><th>Old public price (Excl. VAT / Hour / Node)</th><th>New Public Price (Excl. VAT / Hour / Node)</th><th>New Public Price (Excl. VAT / Hour)</th></tr>
    </thead>
    <tbody>
      <tr><td>Essential DB1-4</td><td>0,1044 €</td><td>0,1145 €</td><td>0,1145 €</td></tr>
      <tr><td>Essential DB1-7</td><td>0,2101 €</td><td>0,2305 €</td><td>0,2305 €</td></tr>
      <tr><td>Essential DB1-15</td><td>0,3913 €</td><td>0,4295 €</td><td>0,4295 €</td></tr>
      <tr><td>Essential DB1-30</td><td>0,7084 €</td><td>0,7775 €</td><td>0,7775 €</td></tr>
      <tr><td>Production B3-8</td><td>0,1917 €</td><td>0,2008 €</td><td>0,6024 €</td></tr>
      <tr><td>Production B3-16</td><td>0,3862 €</td><td>0,4046 €</td><td>1,2138 €</td></tr>
      <tr><td>Production B3-32</td><td>0,7027 €</td><td>0,7363 €</td><td>2,2089 €</td></tr>
      <tr><td>Business DB1-4</td><td>0,1057 €</td><td>0,116 €</td><td>0,348 €</td></tr>
      <tr><td>Business DB1-7</td><td>0,2101 €</td><td>0,2305 €</td><td>0,6915 €</td></tr>
      <tr><td>Business DB1-15</td><td>0,4022 €</td><td>0,4415 €</td><td>1,3245 €</td></tr>
      <tr><td>Business DB1-30</td><td>0,7084 €</td><td>0,7775 €</td><td>2,3325 €</td></tr>
      <tr><td>Advanced B3-8</td><td>0,1908 €</td><td>0,1999 €</td><td>1,1994 €</td></tr>
      <tr><td>Advanced B3-16</td><td>0,3862 €</td><td>0,4046 €</td><td>2,4276 €</td></tr>
      <tr><td>Advanced B3-32</td><td>0,7027 €</td><td>0,7363 €</td><td>4,4178 €</td></tr>
      <tr><td>Enterprise DB1-7</td><td>0,2101 €</td><td>0,2305 €</td><td>1,383 €</td></tr>
      <tr><td>Enterprise DB1-15</td><td>0,4022 €</td><td>0,4415 €</td><td>2,649 €</td></tr>
      <tr><td>Enterprise DB1-30</td><td>0,7084 €</td><td>0,7775 €</td><td>4,665 €</td></tr>
    </tbody>
  </table>

  <h3>Opensearch</h3>
  <table class="dataframe">
    <thead>
      <tr><th>Reference</th><th>Old public price (Excl. VAT / Hour / Node)</th><th>New Public Price (Excl. VAT / Hour / Node)</th><th>New Public Price (Excl. VAT / Hour)</th></tr>
    </thead>
    <tbody>
      <tr><td>Essential DB1-4</td><td>0,0742 €</td><td>0,0814 €</td><td>0,0814 €</td></tr>
      <tr><td>Essential DB1-7</td><td>0,1497 €</td><td>0,1642 €</td><td>0,1642 €</td></tr>
      <tr><td>Essential DB1-15</td><td>0,3007 €</td><td>0,33 €</td><td>0,33 €</td></tr>
      <tr><td>Production B3-8</td><td>0,172 €</td><td>0,1801 €</td><td>0,5403 €</td></tr>
      <tr><td>Production B3-16</td><td>0,3439 €</td><td>0,3603 €</td><td>1,0809 €</td></tr>
      <tr><td>Production B3-32</td><td>0,6877 €</td><td>0,7205 €</td><td>2,1615 €</td></tr>
      <tr><td>Production B3-64</td><td>1,3754 €</td><td>1,4411 €</td><td>4,3233 €</td></tr>
      <tr><td>Business DB1-7</td><td>0,1607 €</td><td>0,1763 €</td><td>0,5289 €</td></tr>
      <tr><td>Business DB1-15</td><td>0,3213 €</td><td>0,3526 €</td><td>1,0578 €</td></tr>
      <tr><td>Business DB1-30</td><td>0,648 €</td><td>0,7112 €</td><td>2,1336 €</td></tr>
      <tr><td>Business DB1-60</td><td>1,2972 €</td><td>1,424 €</td><td>4,272 €</td></tr>
      <tr><td>Business DB1-120</td><td>2,6013 €</td><td>2,8555 €</td><td>8,5665 €</td></tr>
      <tr><td>Advanced B3-8</td><td>0,1839 €</td><td>0,1927 €</td><td>1,1562 €</td></tr>
      <tr><td>Advanced B3-16</td><td>0,3678 €</td><td>0,3854 €</td><td>2,3124 €</td></tr>
      <tr><td>Advanced B3-32</td><td>0,7357 €</td><td>0,7708 €</td><td>4,6248 €</td></tr>
      <tr><td>Advanced B3-64</td><td>1,4713 €</td><td>1,5416 €</td><td>9,2496 €</td></tr>
      <tr><td>Enterprise DB1-7</td><td>0,162 €</td><td>0,1778 €</td><td>1,0668 €</td></tr>
      <tr><td>Enterprise DB1-15</td><td>0,3254 €</td><td>0,3571 €</td><td>2,1426 €</td></tr>
      <tr><td>Enterprise DB1-30</td><td>0,6521 €</td><td>0,7158 €</td><td>4,2948 €</td></tr>
      <tr><td>Enterprise DB1-60</td><td>1,3014 €</td><td>1,4285 €</td><td>8,571 €</td></tr>
      <tr><td>Enterprise DB1-120</td><td>2,6027 €</td><td>2,857 €</td><td>17,142 €</td></tr>
    </tbody>
  </table>

  <h3>Managed Dashboard</h3>
  <table class="dataframe">
    <thead>
      <tr><th>Reference</th><th>Old public price (Excl. VAT / Hour / Node)</th><th>New Public Price (Excl. VAT / Hour / Node)</th><th>New Public Price (Excl. VAT / Hour)</th></tr>
    </thead>
    <tbody>
      <tr><td>Essential DB1-4</td><td>0,0591 €</td><td>0,0648 €</td><td>0,0648 €</td></tr>
      <tr><td>Essential DB1-7</td><td>0,1195 €</td><td>0,1311 €</td><td>0,1311 €</td></tr>
      <tr><td>Production B3-8</td><td>0,1195 €</td><td>0,1251 €</td><td>0,1251 €</td></tr>
    </tbody>
  </table>
</div>

<div class='table-container'>
  <h2>Floating IPs</h2>
  <table class="dataframe">
    <thead>
      <tr style="text-align: right;">
        <th>Reference</th>
        <th>Old public price (Excl. VAT / Hour)</th>
        <th>New Public Price (Excl. VAT / Hour)</th>
      </tr>
    </thead>
    <tbody>
      <tr><td>Floating IPs</td><td>0.0025 €</td><td>0.0027 €</td></tr>
    </tbody>
  </table>
</div>

<!DOCTYPE html>
<html lang="fr">
<head>
  <meta charset="UTF-8">
  <title>Bare Metal Options – Price Tables</title>
  <style>
    body { font-family: Arial, Helvetica, sans-serif; margin: 40px; background-color: #f4f6f8; }
    h2 { margin-top: 50px; color: #1a2b3c; }
    .table-container { margin-top: 20px; margin-bottom: 40px; }
    table { border-collapse: collapse; width: 100%; background: white; border-radius: 8px; overflow: hidden; box-shadow: 0 2px 8px rgba(0,0,0,0.08); }
    thead { background-color: #000E9C; color: white; }
    th { padding: 12px 15px; text-align: center; font-weight: bold; text-transform: uppercase; font-size: 13px; }
    td { padding: 10px 15px; font-size: 14px; }
    td:nth-child(n+2) { text-align: center; }
    tbody tr { border-bottom: 1px solid #e6e9ed; }
    tbody tr:nth-child(even) { background-color: #f8fafc; }
    tbody tr:hover { background-color: #eef3ff; }
    tr.family-header td {
      background-color: #000E9C;
      color: white;
      font-weight: bold;
      font-size: 14px;
      text-align: left !important;
      padding: 10px 15px;
      letter-spacing: 0.05em;
      text-transform: uppercase;
    }
    tr.server-header td {
      background-color: #1a3a6b;
      color: #e8efff;
      font-weight: bold;
      font-size: 13px;
      text-align: center;
      padding: 8px 15px;
      font-style: italic;
    }
    tr.server-header td:first-child {
      text-align: left !important;
    }
    tr.option-subheader td {
      background-color: #dce6f5;
      color: #1a2b3c;
      font-weight: bold;
      font-size: 12px;
      text-align: left !important;
      padding: 6px 15px;
      text-transform: uppercase;
      letter-spacing: 0.08em;
    }
  </style>
</head>
<body>
<div class='table-container'>
  <h2>Dedicated Servers &#038; Options</h2>
<p class="mention">These are the standard monthly price of the servers, without prepayment or commitment discount. The prices for options are for new orders only. The renewal of options, which has been communicated by email to our customers, will be limited to +10% for disk options and +15% for RAM options.</p>
  <table>
    <thead>
      <tr>
        <th></th>
        <th>Old Public Price (Excl. VAT / Month)</th>
        <th>NEW Public Price (Excl. VAT / Month)</th>
      </tr>
    </thead>
    <tbody>
      <tr class="family-header"><td colspan="3">ADVANCE</td></tr>
      <tr class="server-header"><td>ADVANCE-1 &#8211; 2024 &#8211; AMD EPYC 4244P</td><td>84.99 €</td><td>89.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>32GB DDR5 On-Die ECC 5200MHz</td><td>Included</td><td>Included</td></tr>
      <tr><td>64GB DDR5 On-Die ECC 5200MHz</td><td>12 €</td><td>18 €</td></tr>
      <tr><td>128GB DDR5 On-Die ECC 3600MHz</td><td>36 €</td><td>58 €</td></tr>
      <tr><td>192GB DDR5 On-Die ECC 3600MHz</td><td>60 €</td><td>78 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Storage</td></tr>
      <tr><td>2x SSD NVMe 960GB Enterprise Class Soft RAID</td><td>Included</td><td>Included</td></tr>
      <tr><td>4x SSD NVMe 960GB Enterprise Class Soft RAID</td><td>26 €</td><td>42 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>52 €</td><td>70 €</td></tr>
      <tr><td>4x SSD NVMe 1.92TB Enterprise Class Soft RAID</td><td>78 €</td><td>98 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>104 €</td><td>118 €</td></tr>
      <tr><td>4x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>182 €</td><td>200 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 7.68TB Datacenter Class Soft RAID</td><td>208 €</td><td>229 €</td></tr>
      <tr><td>4x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>390 €</td><td>429 €</td></tr>
      <tr class="server-header"><td>ADVANCE-1 &#8211; 2026 &#8211; AMD EPYC 4245P</td><td>99.99 €</td><td>104.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>32GB DDR5 On-Die ECC 5600MHz</td><td>Included</td><td>Included</td></tr>
      <tr><td>64GB DDR5 On-Die ECC 5600MHz</td><td>22 €</td><td>26 €</td></tr>
      <tr><td>128GB DDR5 On-Die ECC 3600MHz</td><td>44 €</td><td>58 €</td></tr>
      <tr><td>256GB DDR5 On-Die ECC 3600MHz</td><td>63 €</td><td>130 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Storage</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID</td><td>Included</td><td>Included</td></tr>
      <tr><td>4x SSD NVMe 960GB Datacenter Class Soft RAID</td><td>21.60 €</td><td>42 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>38 €</td><td>70 €</td></tr>
      <tr><td>4x SSD NVMe 1.92TB Enterprise Class PCIe 5.0 Soft RAID</td><td>54.40 €</td><td>98 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>100 €</td><td>118 €</td></tr>
      <tr><td>4x SSD NVMe 3.84TB Enterprise Class PCIe 5.0 Soft RAID</td><td>178.40 €</td><td>197 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 7.68TB Datacenter Class Soft RAID</td><td>149.20 €</td><td>210 €</td></tr>
      <tr><td>4x SSD NVMe 7.68TB Enterprise Class PCIe 5.0 Soft RAID</td><td>298.40 €</td><td>378 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 15.36TB Datacenter Class Soft RAID</td><td>299.99 €</td><td>392 €</td></tr>
      <tr class="server-header"><td>ADVANCE-2 &#8211; 2024 &#8211; AMD EPYC 4344P</td><td>119.99 €</td><td>124.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>64GB DDR5 On-Die ECC 5200MHz</td><td>Included</td><td>Included</td></tr>
      <tr><td>128GB DDR5 On-Die ECC 3600MHz</td><td>24 €</td><td>40 €</td></tr>
      <tr><td>192GB DDR5 On-Die ECC 3600MHz</td><td>48 €</td><td>60 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Storage</td></tr>
      <tr><td>2x SSD NVMe 960GB Enterprise Class Soft RAID</td><td>Included</td><td>Included</td></tr>
      <tr><td>4x SSD NVMe 960GB Enterprise Class Soft RAID</td><td>26 €</td><td>42 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>52 €</td><td>70 €</td></tr>
      <tr><td>4x SSD NVMe 1.92TB Enterprise Class Soft RAID</td><td>78 €</td><td>98 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>104 €</td><td>118 €</td></tr>
      <tr><td>4x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>182 €</td><td>200 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 7.68TB Datacenter Class Soft RAID</td><td>208 €</td><td>229 €</td></tr>
      <tr><td>4x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>390 €</td><td>429 €</td></tr>
      <tr class="server-header"><td>ADVANCE-2 &#8211; 2026 &#8211; AMD EPYC 4345P</td><td>119.99 €</td><td>134.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>64GB DDR5 On-Die ECC 5600MHz</td><td>Included</td><td>Included</td></tr>
      <tr><td>128GB DDR5 On-Die ECC 3600MHz</td><td>22 €</td><td>40 €</td></tr>
      <tr><td>256GB DDR5 On-Die ECC 3600MHz</td><td>52 €</td><td>112 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Storage</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID</td><td>Included</td><td>Included</td></tr>
      <tr><td>4x SSD NVMe 960GB Datacenter Class Soft RAID</td><td>21.60 €</td><td>42 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>38 €</td><td>70 €</td></tr>
      <tr><td>4x SSD NVMe 1.92TB Enterprise Class PCIe 5.0 Soft RAID</td><td>54.40 €</td><td>98 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>100 €</td><td>118 €</td></tr>
      <tr><td>4x SSD NVMe 3.84TB Enterprise Class PCIe 5.0 Soft RAID</td><td>178.40 €</td><td>197 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 7.68TB Datacenter Class Soft RAID</td><td>149.20 €</td><td>210 €</td></tr>
      <tr><td>4x SSD NVMe 7.68TB Enterprise Class PCIe 5.0 Soft RAID</td><td>298.40 €</td><td>378 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 15.36TB Datacenter Class Soft RAID</td><td>299.99 €</td><td>392 €</td></tr>
      <tr class="server-header"><td>ADVANCE-3 &#8211; 2024 &#8211; AMD EPYC 4464P</td><td>149.99 €</td><td>169.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>64GB DDR5 On-Die ECC 5200MHz</td><td>Included</td><td>Included</td></tr>
      <tr><td>128GB DDR5 On-Die ECC 3600MHz</td><td>24 €</td><td>40 €</td></tr>
      <tr><td>192GB DDR5 On-Die ECC 3600MHz</td><td>48 €</td><td>60 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Storage</td></tr>
      <tr><td>2x SSD NVMe 960GB Enterprise Class Soft RAID</td><td>Included</td><td>Included</td></tr>
      <tr><td>4x SSD NVMe 960GB Enterprise Class Soft RAID</td><td>26 €</td><td>42 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>52 €</td><td>70 €</td></tr>
      <tr><td>4x SSD NVMe 1.92TB Enterprise Class Soft RAID</td><td>78 €</td><td>98 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>104 €</td><td>118 €</td></tr>
      <tr><td>4x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>182 €</td><td>200 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 7.68TB Datacenter Class Soft RAID</td><td>208 €</td><td>229 €</td></tr>
      <tr><td>4x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>390 €</td><td>429 €</td></tr>
      <tr class="server-header"><td>ADVANCE-3 &#8211; 2026 &#8211; AMD EPYC 4464P</td><td>159.99 €</td><td>199.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>64GB DDR5 On-Die ECC 5600MHz</td><td>Included</td><td>Included</td></tr>
      <tr><td>128GB DDR5 On-Die ECC 3600MHz</td><td>22 €</td><td>40 €</td></tr>
      <tr><td>256GB DDR5 On-Die ECC 3600MHz</td><td>52 €</td><td>112 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Storage</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID</td><td>Included</td><td>Included</td></tr>
      <tr><td>4x SSD NVMe 960GB Datacenter Class Soft RAID</td><td>21.60 €</td><td>42 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>38 €</td><td>70 €</td></tr>
      <tr><td>4x SSD NVMe 1.92TB Enterprise Class PCIe 5.0 Soft RAID</td><td>54.40 €</td><td>98 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>100 €</td><td>118 €</td></tr>
      <tr><td>4x SSD NVMe 3.84TB Enterprise Class PCIe 5.0 Soft RAID</td><td>178.40 €</td><td>197 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 7.68TB Datacenter Class Soft RAID</td><td>149.20 €</td><td>210 €</td></tr>
      <tr><td>4x SSD NVMe 7.68TB Enterprise Class PCIe 5.0 Soft RAID</td><td>298.40 €</td><td>378 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 15.36TB Datacenter Class Soft RAID</td><td>299.99 €</td><td>392 €</td></tr>
      <tr class="server-header"><td>ADVANCE-4 &#8211; 2024 &#8211; AMD EPYC 4584PX</td><td>199.99 €</td><td>219.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>64GB DDR5 On-Die ECC 5200MHz</td><td>Included</td><td>Included</td></tr>
      <tr><td>128GB DDR5 On-Die ECC 3600MHz</td><td>24 €</td><td>40 €</td></tr>
      <tr><td>192GB DDR5 On-Die ECC 3600MHz</td><td>48 €</td><td>60 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Storage</td></tr>
      <tr><td>2x SSD NVMe 960GB Enterprise Class Soft RAID</td><td>Included</td><td>Included</td></tr>
      <tr><td>4x SSD NVMe 960GB Enterprise Class Soft RAID</td><td>26 €</td><td>42 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>52 €</td><td>70 €</td></tr>
      <tr><td>4x SSD NVMe 1.92TB Enterprise Class Soft RAID</td><td>78 €</td><td>98 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>104 €</td><td>118 €</td></tr>
      <tr><td>4x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>182 €</td><td>200 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 7.68TB Datacenter Class Soft RAID</td><td>208 €</td><td>229 €</td></tr>
      <tr><td>4x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>390 €</td><td>429 €</td></tr>
      <tr class="server-header"><td>ADVANCE-4 &#8211; 2026 &#8211; AMD EPYC 4585PX</td><td>199.99 €</td><td>239.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>64GB DDR5 On-Die ECC 5600MHz</td><td>Included</td><td>Included</td></tr>
      <tr><td>128GB DDR5 On-Die ECC 3600MHz</td><td>22 €</td><td>40 €</td></tr>
      <tr><td>256GB DDR5 On-Die ECC 3600MHz</td><td>52 €</td><td>112 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Storage</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID</td><td>Included</td><td>Included</td></tr>
      <tr><td>4x SSD NVMe 960GB Datacenter Class Soft RAID</td><td>21.60 €</td><td>42 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>38 €</td><td>70 €</td></tr>
      <tr><td>4x SSD NVMe 1.92TB Enterprise Class PCIe 5.0 Soft RAID</td><td>54.40 €</td><td>98 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>100 €</td><td>118 €</td></tr>
      <tr><td>4x SSD NVMe 3.84TB Enterprise Class PCIe 5.0 Soft RAID</td><td>178.40 €</td><td>197 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 7.68TB Datacenter Class Soft RAID</td><td>149.20 €</td><td>210 €</td></tr>
      <tr><td>4x SSD NVMe 7.68TB Enterprise Class PCIe 5.0 Soft RAID</td><td>298.40 €</td><td>378 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 15.36TB Datacenter Class Soft RAID</td><td>299.99 €</td><td>392 €</td></tr>
      <tr class="server-header"><td>ADVANCE-5 &#8211; 2024 &#8211; AMD EPYC 8224P</td><td>249.99 €</td><td>289.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>96GB DDR5 ECC 4800MHz</td><td>Included</td><td>Included</td></tr>
      <tr><td>192GB DDR5 ECC 4800MHz</td><td>36 €</td><td>90 €</td></tr>
      <tr><td>384GB DDR5 ECC 4800MHz</td><td>108 €</td><td>318 €</td></tr>
      <tr><td>576GB DDR5 ECC 4800MHz</td><td>180 €</td><td>552 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Storage</td></tr>
      <tr><td>2x SSD NVMe 960GB Enterprise Class Soft RAID</td><td>Included</td><td>Included</td></tr>
      <tr><td>4x SSD NVMe 960GB Enterprise Class Soft RAID</td><td>26 €</td><td>42 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>52 €</td><td>70 €</td></tr>
      <tr><td>4x SSD NVMe 1.92TB Enterprise Class Soft RAID</td><td>78 €</td><td>98 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>104 €</td><td>118 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 4x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>104 €</td><td>140 €</td></tr>
      <tr><td>4x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>182 €</td><td>200 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 7.68TB Datacenter Class Soft RAID</td><td>208 €</td><td>229 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 4x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>208 €</td><td>236 €</td></tr>
      <tr><td>4x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>390 €</td><td>429 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 4x SSD NVMe 7.68TB Datacenter Class Soft RAID</td><td>416 €</td><td>458 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 6x SSD NVMe 7.68TB Datacenter Class Soft RAID</td><td>624 €</td><td>687 €</td></tr>
      <tr><td>8x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>806 €</td><td>887 €</td></tr>
      <tr class="server-header"><td>ADVANCE-STOR &#8211; 2024 &#8211; AMD EPYC 4344P</td><td>199.99 €</td><td>199.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>32GB DDR5 On-Die ECC 5200MHz</td><td>Included</td><td>Included</td></tr>
      <tr><td>64GB DDR5 On-Die ECC 5200MHz</td><td>12 €</td><td>14 €</td></tr>
      <tr><td>128GB DDR5 On-Die ECC 3600MHz</td><td>36 €</td><td>42 €</td></tr>
      <tr><td>192GB DDR5 On-Die ECC 3600MHz</td><td>60 €</td><td>69 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Storage</td></tr>
      <tr><td>2x HDD SAS 22TB Enterprise Class Soft RAID</td><td>Included</td><td>Included</td></tr>
      <tr><td>2x SSD NVMe 960GB Enterprise Class Soft RAID</td><td>Included</td><td>Included</td></tr>
      <tr><td>4x HDD SAS 22TB Enterprise Class Soft RAID</td><td>64 €</td><td>70 €</td></tr>
      <tr><td>2x HDD SAS 22TB Enterprise Class Hard RAID</td><td>66 €</td><td>73 €</td></tr>
      <tr><td>6x HDD SAS 22TB Enterprise Class Soft RAID</td><td>128 €</td><td>141 €</td></tr>
      <tr><td>4x HDD SAS 22TB Enterprise Class Hard RAID</td><td>130 €</td><td>143 €</td></tr>
      <tr><td>8x HDD SAS 22TB Enterprise Class Soft RAID</td><td>192 €</td><td>211 €</td></tr>
      <tr><td>6x HDD SAS 22TB Enterprise Class Hard RAID</td><td>194 €</td><td>213 €</td></tr>
      <tr><td>8x HDD SAS 22TB Enterprise Class Hard RAID</td><td>258 €</td><td>284 €</td></tr>
      <tr class="server-header"><td>ADVANCE-STOR &#8211; 2026 &#8211; AMD EPYC 4345P</td><td>199.99 €</td><td>229.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>32GB DDR5 On-Die ECC 5600MHz</td><td>Included</td><td>Included</td></tr>
      <tr><td>64GB DDR5 On-Die ECC 5600MHz</td><td>22 €</td><td>25 €</td></tr>
      <tr><td>128GB DDR5 On-Die ECC 3600MHz</td><td>44 €</td><td>58 €</td></tr>
      <tr><td>256GB DDR5 On-Die ECC 3600MHz</td><td>63 €</td><td>130 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Storage</td></tr>
      <tr><td>2x HDD SAS 24TB Enterprise Class Soft RAID</td><td>Included</td><td>Included</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class PCIe 5.0 Soft RAID</td><td>Included</td><td>Included</td></tr>
      <tr><td>2x HDD SAS 24TB Enterprise Class Hard RAID</td><td>66 €</td><td>73 €</td></tr>
      <tr><td>4x HDD SAS 24TB Enterprise Class Soft RAID</td><td>64 €</td><td>94 €</td></tr>
      <tr><td>4x HDD SAS 24TB Enterprise Class Hard RAID</td><td>130 €</td><td>143 €</td></tr>
      <tr><td>6x HDD SAS 24TB Enterprise Class Soft RAID</td><td>128 €</td><td>188 €</td></tr>
      <tr><td>6x HDD SAS 24TB Enterprise Class Hard RAID</td><td>194 €</td><td>248 €</td></tr>
      <tr><td>8x HDD SAS 24TB Enterprise Class Soft RAID</td><td>192 €</td><td>282 €</td></tr>
      <tr><td>8x HDD SAS 24TB Enterprise Class Hard RAID</td><td>258 €</td><td>362 €</td></tr>
      <tr class="family-header"><td colspan="3">RISE</td></tr>
      <tr class="server-header"><td>RISE-L &#8211; 2025 &#8211; AMD RYZEN 9 9950X</td><td>134.99 €</td><td>149.99 €</td></tr>
      <tr class="server-header"><td>RISE-M &#8211; 2025 &#8211; AMD RYZEN 9 9900X</td><td>94.99 €</td><td>99.99 €</td></tr>
      <tr class="server-header"><td>RISE-S &#8211; 2025 &#8211; AMD Ryzen 7 9700X</td><td>54.99 €</td><td>64.99 €</td></tr>
      <tr class="server-header"><td>RISE-XL &#8211; 2025 &#8211; AMD EPYC TURIN 9455</td><td>269.99 €</td><td>299.99 €</td></tr>
      <tr class="family-header"><td colspan="3">GAME</td></tr>
      <tr class="server-header"><td>GAME-1 &#8211; 2026 &#8211; AMD RYZEN 7 9800X3D</td><td>129.99 €</td><td>139.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>64GB DDR5 On-Die ECC 5600MHz</td><td>Included</td><td>Included</td></tr>
      <tr><td>128GB DDR5 On-Die ECC 3600MHz</td><td>22 €</td><td>40 €</td></tr>
      <tr><td>256GB DDR5 On-Die ECC 3600MHz</td><td>63 €</td><td>112 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Storage</td></tr>
      <tr><td>2x SSD NVMe 960GB Enterprise Class Soft RAID</td><td>Included</td><td>Included</td></tr>
      <tr class="server-header"><td>GAME-2 &#8211; 2026 &#8211; AMD RYZEN 9 9950X3D</td><td>169.99 €</td><td>179.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>64GB DDR5 On-Die ECC 5600MHz</td><td>Included</td><td>Included</td></tr>
      <tr><td>128GB DDR5 On-Die ECC 3600MHz</td><td>22 €</td><td>40 €</td></tr>
      <tr><td>256GB DDR5 On-Die ECC 3600MHz</td><td>63 €</td><td>112 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Storage</td></tr>
      <tr><td>2x SSD NVMe 960GB Enterprise Class Soft RAID</td><td>Included</td><td>Included</td></tr>
      <tr class="family-header"><td colspan="3">SCALE-a</td></tr>
      <tr class="server-header"><td>SCALE-a1 &#8211; 2024 &#8211; AMD EPYC GENOA 9124</td><td>349.99 €</td><td>369.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>128GB DDR5 ECC 4800MHz</td><td>Included</td><td>Included</td></tr>
      <tr><td>256GB DDR5 ECC 4800MHz</td><td>80 €</td><td>200 €</td></tr>
      <tr><td>512GB DDR5 ECC 4800MHz</td><td>240 €</td><td>440 €</td></tr>
      <tr><td>1TB DDR5 ECC 4800MHz</td><td>560 €</td><td>1000 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Storage</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID</td><td>Included</td><td>Included</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>52 €</td><td>70 €</td></tr>
      <tr><td>2x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>104 €</td><td>118 €</td></tr>
      <tr><td>4x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>104 €</td><td>140 €</td></tr>
      <tr><td>6x SSD NVMe 1.92TB Enterprise Class Soft RAID</td><td>156 €</td><td>210 €</td></tr>
      <tr><td>2x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>208 €</td><td>229 €</td></tr>
      <tr><td>4x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>208 €</td><td>236 €</td></tr>
      <tr><td>6x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>312 €</td><td>354 €</td></tr>
      <tr><td>4x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>416 €</td><td>458 €</td></tr>
      <tr><td>6x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>624 €</td><td>687 €</td></tr>
      <tr class="server-header"><td>SCALE-a1 &#8211; 2026 &#8211; AMD EPYC 9135</td><td>389.99 €</td><td>409.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>128GB DDR5 ECC 4800MHz</td><td>Included</td><td>Included</td></tr>
      <tr><td>192GB DDR5 ECC 4800MHz</td><td>40 €</td><td>100 €</td></tr>
      <tr><td>256GB DDR5 ECC 4800MHz</td><td>80 €</td><td>120 €</td></tr>
      <tr><td>384GB DDR5 ECC 4800MHz</td><td>160 €</td><td>280 €</td></tr>
      <tr><td>512GB DDR5 ECC 4800MHz</td><td>240 €</td><td>400 €</td></tr>
      <tr><td>768GB DDR5 ECC 4800MHz</td><td>400 €</td><td>700 €</td></tr>
      <tr><td>1TB DDR5 ECC 4800MHz</td><td>560 €</td><td>1368 €</td></tr>
      <tr><td>1.5TB DDR5 ECC 4800MHz</td><td>880 €</td><td>2152 €</td></tr>
      <tr><td>3TB DDR5 ECC 3600MHz</td><td>1840 €</td><td>4504 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Storage</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class PCIe 5.0 Soft RAID</td><td>Included</td><td>Included</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>38 €</td><td>70 €</td></tr>
      <tr><td>2x SSD NVMe 3.84TB Enterprise Class PCIe 5.0 Soft RAID</td><td>100 €</td><td>118 €</td></tr>
      <tr><td>4x SSD NVMe 1.92TB Enterprise Class PCIe 5.0 Soft RAID</td><td>76 €</td><td>140 €</td></tr>
      <tr><td>2x SSD NVMe 7.68TB Enterprise Class PCIe 5.0 Soft RAID</td><td>160 €</td><td>210 €</td></tr>
      <tr><td>6x SSD NVMe 1.92TB Enterprise Class PCIe 5.0 Soft RAID</td><td>190 €</td><td>210 €</td></tr>
      <tr><td>4x SSD NVMe 3.84TB Enterprise Class PCIe 5.0 Soft RAID</td><td>200 €</td><td>236 €</td></tr>
      <tr><td>6x SSD NVMe 3.84TB Enterprise Class PCIe 5.0 Soft RAID</td><td>300 €</td><td>354 €</td></tr>
      <tr><td>4x SSD NVMe 7.68TB Enterprise Class PCIe 5.0 Soft RAID</td><td>320 €</td><td>420 €</td></tr>
      <tr><td>6x SSD NVMe 7.68TB Enterprise Class PCIe 5.0 Soft RAID</td><td>480 €</td><td>630 €</td></tr>
      <tr><td>6x SSD NVMe 15.36TB Enterprise Class Soft RAID</td><td>900 €</td><td>1176 €</td></tr>
      <tr class="server-header"><td>SCALE-a2 &#8211; 2024 &#8211; AMD EPYC GENOA 9254</td><td>379.99 €</td><td>389.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>128GB DDR5 ECC 4800MHz</td><td>Included</td><td>Included</td></tr>
      <tr><td>256GB DDR5 ECC 4800MHz</td><td>80 €</td><td>200 €</td></tr>
      <tr><td>512GB DDR5 ECC 4800MHz</td><td>240 €</td><td>440 €</td></tr>
      <tr><td>1TB DDR5 ECC 4800MHz</td><td>560 €</td><td>1000 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Storage</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID</td><td>Included</td><td>Included</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>52 €</td><td>70 €</td></tr>
      <tr><td>2x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>104 €</td><td>118 €</td></tr>
      <tr><td>4x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>104 €</td><td>140 €</td></tr>
      <tr><td>6x SSD NVMe 1.92TB Enterprise Class Soft RAID</td><td>156 €</td><td>210 €</td></tr>
      <tr><td>2x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>208 €</td><td>229 €</td></tr>
      <tr><td>4x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>208 €</td><td>236 €</td></tr>
      <tr><td>6x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>312 €</td><td>354 €</td></tr>
      <tr><td>4x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>416 €</td><td>458 €</td></tr>
      <tr><td>6x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>624 €</td><td>687 €</td></tr>
      <tr class="server-header"><td>SCALE-a2 &#8211; 2026 &#8211; AMD EPYC 9255</td><td>429.99 €</td><td>439.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>128GB DDR5 ECC 4800MHz</td><td>Included</td><td>Included</td></tr>
      <tr><td>192GB DDR5 ECC 4800MHz</td><td>40 €</td><td>100 €</td></tr>
      <tr><td>256GB DDR5 ECC 4800MHz</td><td>80 €</td><td>120 €</td></tr>
      <tr><td>384GB DDR5 ECC 4800MHz</td><td>160 €</td><td>280 €</td></tr>
      <tr><td>512GB DDR5 ECC 4800MHz</td><td>240 €</td><td>400 €</td></tr>
      <tr><td>768GB DDR5 ECC 4800MHz</td><td>400 €</td><td>700 €</td></tr>
      <tr><td>1TB DDR5 ECC 4800MHz</td><td>560 €</td><td>1368 €</td></tr>
      <tr><td>1.5TB DDR5 ECC 4800MHz</td><td>880 €</td><td>2152 €</td></tr>
      <tr><td>3TB DDR5 ECC 3600MHz</td><td>1840 €</td><td>4504 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Storage</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class PCIe 5.0 Soft RAID</td><td>Included</td><td>Included</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>38 €</td><td>70 €</td></tr>
      <tr><td>2x SSD NVMe 3.84TB Enterprise Class PCIe 5.0 Soft RAID</td><td>100 €</td><td>118 €</td></tr>
      <tr><td>4x SSD NVMe 1.92TB Enterprise Class PCIe 5.0 Soft RAID</td><td>76 €</td><td>140 €</td></tr>
      <tr><td>2x SSD NVMe 7.68TB Enterprise Class PCIe 5.0 Soft RAID</td><td>160 €</td><td>210 €</td></tr>
      <tr><td>6x SSD NVMe 1.92TB Enterprise Class PCIe 5.0 Soft RAID</td><td>190 €</td><td>210 €</td></tr>
      <tr><td>4x SSD NVMe 3.84TB Enterprise Class PCIe 5.0 Soft RAID</td><td>200 €</td><td>236 €</td></tr>
      <tr><td>6x SSD NVMe 3.84TB Enterprise Class PCIe 5.0 Soft RAID</td><td>300 €</td><td>354 €</td></tr>
      <tr><td>4x SSD NVMe 7.68TB Enterprise Class PCIe 5.0 Soft RAID</td><td>320 €</td><td>420 €</td></tr>
      <tr><td>6x SSD NVMe 7.68TB Enterprise Class PCIe 5.0 Soft RAID</td><td>480 €</td><td>630 €</td></tr>
      <tr><td>6x SSD NVMe 15.36TB Enterprise Class Soft RAID</td><td>900 €</td><td>1176 €</td></tr>
      <tr class="server-header"><td>SCALE-a3  &#8211; 2024 &#8211; AMD EPYC GENOA 9354</td><td>419.99 €</td><td>449.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>128GB DDR5 ECC 4800MHz</td><td>Included</td><td>Included</td></tr>
      <tr><td>256GB DDR5 ECC 4800MHz</td><td>80 €</td><td>200 €</td></tr>
      <tr><td>512GB DDR5 ECC 4800MHz</td><td>240 €</td><td>440 €</td></tr>
      <tr><td>1TB DDR5 ECC 4800MHz</td><td>560 €</td><td>1000 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Storage</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID</td><td>Included</td><td>Included</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>52 €</td><td>70 €</td></tr>
      <tr><td>2x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>104 €</td><td>118 €</td></tr>
      <tr><td>4x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>104 €</td><td>140 €</td></tr>
      <tr><td>6x SSD NVMe 1.92TB Enterprise Class Soft RAID</td><td>156 €</td><td>210 €</td></tr>
      <tr><td>2x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>208 €</td><td>229 €</td></tr>
      <tr><td>4x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>208 €</td><td>236 €</td></tr>
      <tr><td>6x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>312 €</td><td>354 €</td></tr>
      <tr><td>4x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>416 €</td><td>458 €</td></tr>
      <tr><td>6x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>624 €</td><td>687 €</td></tr>
      <tr class="server-header"><td>SCALE-a3 &#8211; 2026 &#8211; AMD EPYC 9355</td><td>469.99 €</td><td>499.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>128GB DDR5 ECC 4800MHz</td><td>Included</td><td>Included</td></tr>
      <tr><td>192GB DDR5 ECC 4800MHz</td><td>40 €</td><td>100 €</td></tr>
      <tr><td>256GB DDR5 ECC 4800MHz</td><td>80 €</td><td>120 €</td></tr>
      <tr><td>384GB DDR5 ECC 4800MHz</td><td>160 €</td><td>280 €</td></tr>
      <tr><td>512GB DDR5 ECC 4800MHz</td><td>240 €</td><td>400 €</td></tr>
      <tr><td>768GB DDR5 ECC 4800MHz</td><td>400 €</td><td>700 €</td></tr>
      <tr><td>1TB DDR5 ECC 4800MHz</td><td>560 €</td><td>1368 €</td></tr>
      <tr><td>1.5TB DDR5 ECC 4800MHz</td><td>880 €</td><td>2152 €</td></tr>
      <tr><td>3TB DDR5 ECC 3600MHz</td><td>1840 €</td><td>4504 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Storage</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class PCIe 5.0 Soft RAID</td><td>Included</td><td>Included</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>38 €</td><td>70 €</td></tr>
      <tr><td>2x SSD NVMe 3.84TB Enterprise Class PCIe 5.0 Soft RAID</td><td>100 €</td><td>118 €</td></tr>
      <tr><td>4x SSD NVMe 1.92TB Enterprise Class PCIe 5.0 Soft RAID</td><td>76 €</td><td>140 €</td></tr>
      <tr><td>2x SSD NVMe 7.68TB Enterprise Class PCIe 5.0 Soft RAID</td><td>160 €</td><td>210 €</td></tr>
      <tr><td>6x SSD NVMe 1.92TB Enterprise Class PCIe 5.0 Soft RAID</td><td>190 €</td><td>210 €</td></tr>
      <tr><td>4x SSD NVMe 3.84TB Enterprise Class PCIe 5.0 Soft RAID</td><td>200 €</td><td>236 €</td></tr>
      <tr><td>6x SSD NVMe 3.84TB Enterprise Class PCIe 5.0 Soft RAID</td><td>300 €</td><td>354 €</td></tr>
      <tr><td>4x SSD NVMe 7.68TB Enterprise Class PCIe 5.0 Soft RAID</td><td>320 €</td><td>420 €</td></tr>
      <tr><td>6x SSD NVMe 7.68TB Enterprise Class PCIe 5.0 Soft RAID</td><td>480 €</td><td>630 €</td></tr>
      <tr><td>6x SSD NVMe 15.36TB Enterprise Class Soft RAID</td><td>900 €</td><td>1176 €</td></tr>
      <tr class="server-header"><td>SCALE-a4 &#8211; 2024 &#8211; AMD EPYC GENOA 9454</td><td>449.99 €</td><td>459.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>128GB DDR5 ECC 4800MHz</td><td>Included</td><td>Included</td></tr>
      <tr><td>256GB DDR5 ECC 4800MHz</td><td>80 €</td><td>200 €</td></tr>
      <tr><td>512GB DDR5 ECC 4800MHz</td><td>240 €</td><td>440 €</td></tr>
      <tr><td>1TB DDR5 ECC 4800MHz</td><td>560 €</td><td>1000 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Storage</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID</td><td>Included</td><td>Included</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>52 €</td><td>70 €</td></tr>
      <tr><td>2x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>104 €</td><td>118 €</td></tr>
      <tr><td>4x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>104 €</td><td>140 €</td></tr>
      <tr><td>6x SSD NVMe 1.92TB Enterprise Class Soft RAID</td><td>156 €</td><td>210 €</td></tr>
      <tr><td>2x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>208 €</td><td>229 €</td></tr>
      <tr><td>4x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>208 €</td><td>236 €</td></tr>
      <tr><td>6x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>312 €</td><td>354 €</td></tr>
      <tr><td>4x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>416 €</td><td>458 €</td></tr>
      <tr><td>6x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>624 €</td><td>687 €</td></tr>
      <tr class="server-header"><td>SCALE-a4 &#8211; 2026 &#8211;  AMD EPYC 9455</td><td>539.99 €</td><td>549.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>128GB DDR5 ECC 4800MHz</td><td>Included</td><td>Included</td></tr>
      <tr><td>192GB DDR5 ECC 4800MHz</td><td>40 €</td><td>100 €</td></tr>
      <tr><td>256GB DDR5 ECC 4800MHz</td><td>80 €</td><td>120 €</td></tr>
      <tr><td>384GB DDR5 ECC 4800MHz</td><td>160 €</td><td>280 €</td></tr>
      <tr><td>512GB DDR5 ECC 4800MHz</td><td>240 €</td><td>400 €</td></tr>
      <tr><td>768GB DDR5 ECC 4800MHz</td><td>400 €</td><td>700 €</td></tr>
      <tr><td>1TB DDR5 ECC 4800MHz</td><td>560 €</td><td>1368 €</td></tr>
      <tr><td>1.5TB DDR5 ECC 4800MHz</td><td>880 €</td><td>2152 €</td></tr>
      <tr><td>3TB DDR5 ECC 3600MHz</td><td>1840 €</td><td>4504 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Storage</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class PCIe 5.0 Soft RAID</td><td>Included</td><td>Included</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>38 €</td><td>70 €</td></tr>
      <tr><td>2x SSD NVMe 3.84TB Enterprise Class PCIe 5.0 Soft RAID</td><td>100 €</td><td>118 €</td></tr>
      <tr><td>4x SSD NVMe 1.92TB Enterprise Class PCIe 5.0 Soft RAID</td><td>76 €</td><td>140 €</td></tr>
      <tr><td>2x SSD NVMe 7.68TB Enterprise Class PCIe 5.0 Soft RAID</td><td>160 €</td><td>210 €</td></tr>
      <tr><td>6x SSD NVMe 1.92TB Enterprise Class PCIe 5.0 Soft RAID</td><td>190 €</td><td>210 €</td></tr>
      <tr><td>4x SSD NVMe 3.84TB Enterprise Class PCIe 5.0 Soft RAID</td><td>200 €</td><td>236 €</td></tr>
      <tr><td>6x SSD NVMe 3.84TB Enterprise Class PCIe 5.0 Soft RAID</td><td>300 €</td><td>354 €</td></tr>
      <tr><td>4x SSD NVMe 7.68TB Enterprise Class PCIe 5.0 Soft RAID</td><td>320 €</td><td>420 €</td></tr>
      <tr><td>6x SSD NVMe 7.68TB Enterprise Class PCIe 5.0 Soft RAID</td><td>480 €</td><td>630 €</td></tr>
      <tr><td>6x SSD NVMe 15.36TB Enterprise Class Soft RAID</td><td>900 €</td><td>1176 €</td></tr>
      <tr class="server-header"><td>SCALE-a5 &#8211; 2024 &#8211; AMD EPYC GENOA 9554</td><td>499.99 €</td><td>539.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>128GB DDR5 ECC 4800MHz</td><td>Included</td><td>Included</td></tr>
      <tr><td>256GB DDR5 ECC 4800MHz</td><td>80 €</td><td>200 €</td></tr>
      <tr><td>512GB DDR5 ECC 4800MHz</td><td>240 €</td><td>440 €</td></tr>
      <tr><td>1TB DDR5 ECC 4800MHz</td><td>560 €</td><td>1000 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Storage</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID</td><td>Included</td><td>Included</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>52 €</td><td>70 €</td></tr>
      <tr><td>2x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>104 €</td><td>118 €</td></tr>
      <tr><td>4x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>104 €</td><td>140 €</td></tr>
      <tr><td>6x SSD NVMe 1.92TB Enterprise Class Soft RAID</td><td>156 €</td><td>210 €</td></tr>
      <tr><td>2x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>208 €</td><td>229 €</td></tr>
      <tr><td>4x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>208 €</td><td>236 €</td></tr>
      <tr><td>6x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>312 €</td><td>354 €</td></tr>
      <tr><td>4x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>416 €</td><td>458 €</td></tr>
      <tr><td>6x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>624 €</td><td>687 €</td></tr>
      <tr class="server-header"><td>SCALE-a5 &#8211; 2026 &#8211; AMD EPYC 9555</td><td>599.99 €</td><td>639.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>128GB DDR5 ECC 4800MHz</td><td>Included</td><td>Included</td></tr>
      <tr><td>192GB DDR5 ECC 4800MHz</td><td>40 €</td><td>100 €</td></tr>
      <tr><td>256GB DDR5 ECC 4800MHz</td><td>80 €</td><td>120 €</td></tr>
      <tr><td>384GB DDR5 ECC 4800MHz</td><td>160 €</td><td>280 €</td></tr>
      <tr><td>512GB DDR5 ECC 4800MHz</td><td>240 €</td><td>400 €</td></tr>
      <tr><td>768GB DDR5 ECC 4800MHz</td><td>400 €</td><td>700 €</td></tr>
      <tr><td>1TB DDR5 ECC 4800MHz</td><td>560 €</td><td>1368 €</td></tr>
      <tr><td>1.5TB DDR5 ECC 4800MHz</td><td>880 €</td><td>2152 €</td></tr>
      <tr><td>3TB DDR5 ECC 3600MHz</td><td>1840 €</td><td>4504 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Storage</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class PCIe 5.0 Soft RAID</td><td>Included</td><td>Included</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>38 €</td><td>70 €</td></tr>
      <tr><td>2x SSD NVMe 3.84TB Enterprise Class PCIe 5.0 Soft RAID</td><td>100 €</td><td>118 €</td></tr>
      <tr><td>4x SSD NVMe 1.92TB Enterprise Class PCIe 5.0 Soft RAID</td><td>76 €</td><td>140 €</td></tr>
      <tr><td>2x SSD NVMe 7.68TB Enterprise Class PCIe 5.0 Soft RAID</td><td>160 €</td><td>210 €</td></tr>
      <tr><td>6x SSD NVMe 1.92TB Enterprise Class PCIe 5.0 Soft RAID</td><td>190 €</td><td>210 €</td></tr>
      <tr><td>4x SSD NVMe 3.84TB Enterprise Class PCIe 5.0 Soft RAID</td><td>200 €</td><td>236 €</td></tr>
      <tr><td>6x SSD NVMe 3.84TB Enterprise Class PCIe 5.0 Soft RAID</td><td>300 €</td><td>354 €</td></tr>
      <tr><td>4x SSD NVMe 7.68TB Enterprise Class PCIe 5.0 Soft RAID</td><td>320 €</td><td>420 €</td></tr>
      <tr><td>6x SSD NVMe 7.68TB Enterprise Class PCIe 5.0 Soft RAID</td><td>480 €</td><td>630 €</td></tr>
      <tr><td>6x SSD NVMe 15.36TB Enterprise Class Soft RAID</td><td>900 €</td><td>1176 €</td></tr>
      <tr class="server-header"><td>SCALE-a6 &#8211; 2024 &#8211; AMD EPYC GENOA 9654</td><td>579.99 €</td><td>629.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>128GB DDR5 ECC 4800MHz</td><td>Included</td><td>Included</td></tr>
      <tr><td>256GB DDR5 ECC 4800MHz</td><td>80 €</td><td>200 €</td></tr>
      <tr><td>512GB DDR5 ECC 4800MHz</td><td>240 €</td><td>440 €</td></tr>
      <tr><td>1TB DDR5 ECC 4800MHz</td><td>560 €</td><td>1000 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Storage</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID</td><td>Included</td><td>Included</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>52 €</td><td>70 €</td></tr>
      <tr><td>2x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>104 €</td><td>118 €</td></tr>
      <tr><td>4x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>104 €</td><td>140 €</td></tr>
      <tr><td>6x SSD NVMe 1.92TB Enterprise Class Soft RAID</td><td>156 €</td><td>210 €</td></tr>
      <tr><td>2x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>208 €</td><td>229 €</td></tr>
      <tr><td>4x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>208 €</td><td>236 €</td></tr>
      <tr><td>6x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>312 €</td><td>354 €</td></tr>
      <tr><td>4x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>416 €</td><td>458 €</td></tr>
      <tr><td>6x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>624 €</td><td>687 €</td></tr>
      <tr class="server-header"><td>SCALE-a6 &#8211; 2026 &#8211; AMD EPYC 9655</td><td>699.99 €</td><td>729.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>128GB DDR5 ECC 4800MHz</td><td>Included</td><td>Included</td></tr>
      <tr><td>192GB DDR5 ECC 4800MHz</td><td>40 €</td><td>100 €</td></tr>
      <tr><td>256GB DDR5 ECC 4800MHz</td><td>80 €</td><td>120 €</td></tr>
      <tr><td>384GB DDR5 ECC 4800MHz</td><td>160 €</td><td>280 €</td></tr>
      <tr><td>512GB DDR5 ECC 4800MHz</td><td>240 €</td><td>400 €</td></tr>
      <tr><td>768GB DDR5 ECC 4800MHz</td><td>400 €</td><td>700 €</td></tr>
      <tr><td>1TB DDR5 ECC 4800MHz</td><td>560 €</td><td>1368 €</td></tr>
      <tr><td>1.5TB DDR5 ECC 4800MHz</td><td>880 €</td><td>2152 €</td></tr>
      <tr><td>3TB DDR5 ECC 3600MHz</td><td>1840 €</td><td>4504 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Storage</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class PCIe 5.0 Soft RAID</td><td>Included</td><td>Included</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>38 €</td><td>70 €</td></tr>
      <tr><td>2x SSD NVMe 3.84TB Enterprise Class PCIe 5.0 Soft RAID</td><td>100 €</td><td>118 €</td></tr>
      <tr><td>4x SSD NVMe 1.92TB Enterprise Class PCIe 5.0 Soft RAID</td><td>76 €</td><td>140 €</td></tr>
      <tr><td>2x SSD NVMe 7.68TB Enterprise Class PCIe 5.0 Soft RAID</td><td>160 €</td><td>210 €</td></tr>
      <tr><td>6x SSD NVMe 1.92TB Enterprise Class PCIe 5.0 Soft RAID</td><td>190 €</td><td>210 €</td></tr>
      <tr><td>4x SSD NVMe 3.84TB Enterprise Class PCIe 5.0 Soft RAID</td><td>200 €</td><td>236 €</td></tr>
      <tr><td>6x SSD NVMe 3.84TB Enterprise Class PCIe 5.0 Soft RAID</td><td>300 €</td><td>354 €</td></tr>
      <tr><td>4x SSD NVMe 7.68TB Enterprise Class PCIe 5.0 Soft RAID</td><td>320 €</td><td>420 €</td></tr>
      <tr><td>6x SSD NVMe 7.68TB Enterprise Class PCIe 5.0 Soft RAID</td><td>480 €</td><td>630 €</td></tr>
      <tr><td>6x SSD NVMe 15.36TB Enterprise Class Soft RAID</td><td>900 €</td><td>1176 €</td></tr>
      <tr class="server-header"><td>SCALE-a7 &#8211; 2026 &#8211; AMD EPYC 9755</td><td>809.99 €</td><td>829.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>128GB DDR5 ECC 4800MHz</td><td>Included</td><td>Included</td></tr>
      <tr><td>192GB DDR5 ECC 4800MHz</td><td>40 €</td><td>100 €</td></tr>
      <tr><td>256GB DDR5 ECC 4800MHz</td><td>80 €</td><td>120 €</td></tr>
      <tr><td>384GB DDR5 ECC 4800MHz</td><td>160 €</td><td>280 €</td></tr>
      <tr><td>512GB DDR5 ECC 4800MHz</td><td>240 €</td><td>400 €</td></tr>
      <tr><td>768GB DDR5 ECC 4800MHz</td><td>192 €</td><td>700 €</td></tr>
      <tr><td>1TB DDR5 ECC 4800MHz</td><td>560 €</td><td>1368 €</td></tr>
      <tr><td>1.5TB DDR5 ECC 4800MHz</td><td>880 €</td><td>2152 €</td></tr>
      <tr><td>3TB DDR5 ECC 3600MHz</td><td>1840 €</td><td>4504 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Storage</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class PCIe 5.0 Soft RAID</td><td>Included</td><td>Included</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>38 €</td><td>70 €</td></tr>
      <tr><td>2x SSD NVMe 3.84TB Enterprise Class PCIe 5.0 Soft RAID</td><td>100 €</td><td>118 €</td></tr>
      <tr><td>4x SSD NVMe 1.92TB Enterprise Class PCIe 5.0 Soft RAID</td><td>76 €</td><td>140 €</td></tr>
      <tr><td>2x SSD NVMe 7.68TB Enterprise Class PCIe 5.0 Soft RAID</td><td>160 €</td><td>210 €</td></tr>
      <tr><td>6x SSD NVMe 1.92TB Enterprise Class PCIe 5.0 Soft RAID</td><td>190 €</td><td>210 €</td></tr>
      <tr><td>4x SSD NVMe 3.84TB Enterprise Class PCIe 5.0 Soft RAID</td><td>200 €</td><td>236 €</td></tr>
      <tr><td>6x SSD NVMe 3.84TB Enterprise Class PCIe 5.0 Soft RAID</td><td>300 €</td><td>354 €</td></tr>
      <tr><td>4x SSD NVMe 7.68TB Enterprise Class PCIe 5.0 Soft RAID</td><td>320 €</td><td>420 €</td></tr>
      <tr><td>6x SSD NVMe 7.68TB Enterprise Class PCIe 5.0 Soft RAID</td><td>480 €</td><td>630 €</td></tr>
      <tr><td>6x SSD NVMe 15.36TB Enterprise Class Soft RAID</td><td>900 €</td><td>1176 €</td></tr>
      <tr class="server-header"><td>SCALE-a8 &#8211; 2026 &#8211; AMD EPYC 9965</td><td>869.99 €</td><td>899.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>128GB DDR5 ECC 4800MHz</td><td>Included</td><td>Included</td></tr>
      <tr><td>192GB DDR5 ECC 4800MHz</td><td>40 €</td><td>100 €</td></tr>
      <tr><td>256GB DDR5 ECC 4800MHz</td><td>80 €</td><td>120 €</td></tr>
      <tr><td>384GB DDR5 ECC 4800MHz</td><td>160 €</td><td>280 €</td></tr>
      <tr><td>512GB DDR5 ECC 4800MHz</td><td>240 €</td><td>400 €</td></tr>
      <tr><td>768GB DDR5 ECC 4800MHz</td><td>192 €</td><td>700 €</td></tr>
      <tr><td>1TB DDR5 ECC 4800MHz</td><td>560 €</td><td>1368 €</td></tr>
      <tr><td>1.5TB DDR5 ECC 4800MHz</td><td>880 €</td><td>2152 €</td></tr>
      <tr><td>3TB DDR5 ECC 3600MHz</td><td>1840 €</td><td>4504 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Storage</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class PCIe 5.0 Soft RAID</td><td>Included</td><td>Included</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>38 €</td><td>70 €</td></tr>
      <tr><td>2x SSD NVMe 3.84TB Enterprise Class PCIe 5.0 Soft RAID</td><td>100 €</td><td>118 €</td></tr>
      <tr><td>4x SSD NVMe 1.92TB Enterprise Class PCIe 5.0 Soft RAID</td><td>76 €</td><td>140 €</td></tr>
      <tr><td>2x SSD NVMe 7.68TB Enterprise Class PCIe 5.0 Soft RAID</td><td>160 €</td><td>210 €</td></tr>
      <tr><td>6x SSD NVMe 1.92TB Enterprise Class PCIe 5.0 Soft RAID</td><td>190 €</td><td>210 €</td></tr>
      <tr><td>4x SSD NVMe 3.84TB Enterprise Class PCIe 5.0 Soft RAID</td><td>200 €</td><td>236 €</td></tr>
      <tr><td>6x SSD NVMe 3.84TB Enterprise Class PCIe 5.0 Soft RAID</td><td>300 €</td><td>354 €</td></tr>
      <tr><td>4x SSD NVMe 7.68TB Enterprise Class PCIe 5.0 Soft RAID</td><td>320 €</td><td>420 €</td></tr>
      <tr><td>6x SSD NVMe 7.68TB Enterprise Class PCIe 5.0 Soft RAID</td><td>480 €</td><td>630 €</td></tr>
      <tr><td>6x SSD NVMe 15.36TB Enterprise Class Soft RAID</td><td>900 €</td><td>1176 €</td></tr>
      <tr class="server-header"><td>SCALE-a9 &#8211; 2026 &#8211; Dual AMD EPYC 9965</td><td>1349.99 €</td><td>1349.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>128GB DDR5 ECC 5600MHz</td><td>Included</td><td>Included</td></tr>
      <tr><td>192GB DDR5 ECC 5600MHz</td><td>40 €</td><td>40 €</td></tr>
      <tr><td>256GB DDR5 ECC 5600MHz</td><td>80 €</td><td>80 €</td></tr>
      <tr><td>384GB DDR5 ECC 5600MHz</td><td>160 €</td><td>160 €</td></tr>
      <tr><td>512GB DDR5 ECC 5600MHz</td><td>240 €</td><td>240 €</td></tr>
      <tr><td>768GB DDR5 ECC 5600MHz</td><td>Included</td><td>400 €</td></tr>
      <tr><td>1024GB DDR5 ECC 5600MHz</td><td>560 €</td><td>560 €</td></tr>
      <tr><td>1.5TB DDR5 ECC 5600MHz</td><td>880 €</td><td>880 €</td></tr>
      <tr><td>3TB DDR5 ECC 5600MHz</td><td>1840 €</td><td>1840 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Storage</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class PCIe 5.0 Soft RAID</td><td>Included</td><td>Included</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>38 €</td><td>38 €</td></tr>
      <tr><td>2x SSD NVMe 3.84TB Enterprise Class PCIe 5.0 Soft RAID</td><td>100 €</td><td>100 €</td></tr>
      <tr><td>2x SSD NVMe 7.68TB Enterprise Class PCIe 5.0 Soft RAID</td><td>160 €</td><td>160 €</td></tr>
      <tr class="family-header"><td colspan="3">SCALE-i</td></tr>
      <tr class="server-header"><td>SCALE-i1 &#8211; 2024 &#8211; Intel Xeon Gold 6426Y</td><td>349.99 €</td><td>369.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>128GB DDR5 ECC 4800MHz</td><td>Included</td><td>Included</td></tr>
      <tr><td>256GB DDR5 ECC 4800MHz</td><td>80 €</td><td>200 €</td></tr>
      <tr><td>512GB DDR5 ECC 4800MHz</td><td>240 €</td><td>440 €</td></tr>
      <tr><td>1TB DDR5 ECC 4800MHz</td><td>560 €</td><td>1000 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Storage</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID</td><td>Included</td><td>Included</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>52 €</td><td>70 €</td></tr>
      <tr><td>2x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>104 €</td><td>118 €</td></tr>
      <tr><td>4x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>104 €</td><td>140 €</td></tr>
      <tr><td>6x SSD NVMe 1.92TB Enterprise Class Soft RAID</td><td>156 €</td><td>210 €</td></tr>
      <tr><td>2x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>208 €</td><td>229 €</td></tr>
      <tr><td>4x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>208 €</td><td>236 €</td></tr>
      <tr><td>6x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>312 €</td><td>354 €</td></tr>
      <tr><td>4x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>416 €</td><td>458 €</td></tr>
      <tr><td>6x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>624 €</td><td>687 €</td></tr>
      <tr class="server-header"><td>SCALE-i2 &#8211; 2024 &#8211; Intel Xeon Gold 6442Y</td><td>379.99 €</td><td>389.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>128GB DDR5 ECC 4800MHz</td><td>Included</td><td>Included</td></tr>
      <tr><td>256GB DDR5 ECC 4800MHz</td><td>80 €</td><td>200 €</td></tr>
      <tr><td>512GB DDR5 ECC 4800MHz</td><td>240 €</td><td>440 €</td></tr>
      <tr><td>1TB DDR5 ECC 4800MHz</td><td>560 €</td><td>1000 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Storage</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID</td><td>Included</td><td>Included</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>52 €</td><td>70 €</td></tr>
      <tr><td>2x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>104 €</td><td>118 €</td></tr>
      <tr><td>4x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>104 €</td><td>140 €</td></tr>
      <tr><td>6x SSD NVMe 1.92TB Enterprise Class Soft RAID</td><td>156 €</td><td>210 €</td></tr>
      <tr><td>2x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>208 €</td><td>229 €</td></tr>
      <tr><td>4x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>208 €</td><td>236 €</td></tr>
      <tr><td>6x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>312 €</td><td>354 €</td></tr>
      <tr><td>4x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>416 €</td><td>458 €</td></tr>
      <tr><td>6x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>624 €</td><td>687 €</td></tr>
      <tr class="server-header"><td>SCALE-i3 &#8211; 2024 &#8211; Intel Xeon Gold 6438M</td><td>409.99 €</td><td>449.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>128GB DDR5 ECC 4800MHz</td><td>Included</td><td>Included</td></tr>
      <tr><td>256GB DDR5 ECC 4800MHz</td><td>80 €</td><td>200 €</td></tr>
      <tr><td>512GB DDR5 ECC 4800MHz</td><td>240 €</td><td>440 €</td></tr>
      <tr><td>1TB DDR5 ECC 4800MHz</td><td>560 €</td><td>1000 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Storage</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID</td><td>Included</td><td>Included</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>52 €</td><td>70 €</td></tr>
      <tr><td>2x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>104 €</td><td>118 €</td></tr>
      <tr><td>4x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>104 €</td><td>140 €</td></tr>
      <tr><td>6x SSD NVMe 1.92TB Enterprise Class Soft RAID</td><td>156 €</td><td>210 €</td></tr>
      <tr><td>2x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>208 €</td><td>229 €</td></tr>
      <tr><td>4x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>208 €</td><td>236 €</td></tr>
      <tr><td>6x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>312 €</td><td>354 €</td></tr>
      <tr><td>4x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>416 €</td><td>458 €</td></tr>
      <tr><td>6x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>624 €</td><td>687 €</td></tr>
      <tr class="family-header"><td colspan="3">SCALE-GPU</td></tr>
      <tr class="server-header"><td>SCALE-GPU-1 &#8211; 2024 &#8211; AMD EPYC GENOA 9354</td><td>969.99 €</td><td>969.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>192GB DDR5 ECC 4800MHz</td><td>Included</td><td>Included</td></tr>
      <tr><td>384GB DDR5 ECC 4800MHz</td><td>120 €</td><td>120 €</td></tr>
      <tr><td>768GB DDR5 ECC 4800MHz</td><td>240 €</td><td>240 €</td></tr>
      <tr><td>RAM 1,1TB DDR5 ECC 4800MHz</td><td>420 €</td><td>420 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Storage</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID</td><td>Included</td><td>Included</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>52 €</td><td>52 €</td></tr>
      <tr><td>2x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>104 €</td><td>104 €</td></tr>
      <tr><td>2x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>208 €</td><td>208 €</td></tr>
      <tr class="server-header"><td>SCALE-GPU-2 &#8211; 2024 &#8211; AMD EPYC GENOA 9454</td><td>999.99 €</td><td>999.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>192GB DDR5 ECC 4800MHz</td><td>Included</td><td>Included</td></tr>
      <tr><td>384GB DDR5 ECC 4800MHz</td><td>120 €</td><td>120 €</td></tr>
      <tr><td>768GB DDR5 ECC 4800MHz</td><td>240 €</td><td>240 €</td></tr>
      <tr><td>RAM 1,1TB DDR5 ECC 4800MHz</td><td>420 €</td><td>420 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Storage</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID</td><td>Included</td><td>Included</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>52 €</td><td>52 €</td></tr>
      <tr><td>2x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>104 €</td><td>104 €</td></tr>
      <tr><td>2x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>208 €</td><td>208 €</td></tr>
      <tr class="server-header"><td>SCALE-GPU-3 &#8211; 2024 &#8211; AMD EPYC GENOA 9554</td><td>1029.99 €</td><td>1029.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>192GB DDR5 ECC 4800MHz</td><td>Included</td><td>Included</td></tr>
      <tr><td>384GB DDR5 ECC 4800MHz</td><td>120 €</td><td>120 €</td></tr>
      <tr><td>768GB DDR5 ECC 4800MHz</td><td>240 €</td><td>240 €</td></tr>
      <tr><td>RAM 1,1TB DDR5 ECC 4800MHz</td><td>420 €</td><td>420 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Storage</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID</td><td>Included</td><td>Included</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>52 €</td><td>52 €</td></tr>
      <tr><td>2x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>104 €</td><td>104 €</td></tr>
      <tr><td>2x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>208 €</td><td>208 €</td></tr>
      <tr class="family-header"><td colspan="3">HGR</td></tr>
      <tr class="server-header"><td>HGR-AI-2 &#8211; 2024 &#8211; DUAL AMD EPYC 9354</td><td>2969.99 €</td><td>2969.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>384GB DDR5 ECC 4800MHz</td><td>Included</td><td>Included</td></tr>
      <tr><td>512GB DDR5 ECC 4800MHz</td><td>64 €</td><td>74 €</td></tr>
      <tr><td>768GB DDR5 ECC 4800MHz</td><td>400 €</td><td>360 €</td></tr>
      <tr><td>RAM 2304GB DDR5 ECC 4800MHz</td><td>960 €</td><td>2208 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Storage</td></tr>
      <tr><td>2x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>Included</td><td>Included</td></tr>
      <tr><td>4x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>88 €</td><td>118 €</td></tr>
      <tr><td>2x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>150 €</td><td>165 €</td></tr>
      <tr><td>4x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>300 €</td><td>330 €</td></tr>
      <tr><td>2x SSD NVMe 15.36TB Enterprise Class Soft RAID</td><td>308 €</td><td>339 €</td></tr>
      <tr><td>4x SSD NVMe 15.36TB Enterprise Class Soft RAID</td><td>616 €</td><td>680 €</td></tr>
      <tr class="server-header"><td>HGR-HCI-a1 &#8211; 2024 &#8211; DUAL AMD EPYC 9254</td><td>999.99 €</td><td>1119.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>256GB DDR5 ECC 4800MHz</td><td>Included</td><td>Included</td></tr>
      <tr><td>512GB DDR5 ECC 4800MHz</td><td>128 €</td><td>240 €</td></tr>
      <tr><td>1TB DDR5 ECC 4800MHz</td><td>384 €</td><td>800 €</td></tr>
      <tr><td>1.5TB DDR5 ECC 4800MHz</td><td>512 €</td><td>1472 €</td></tr>
      <tr><td>RAM 2304GB DDR5 ECC 4800MHz</td><td>1024 €</td><td>2408 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Storage</td></tr>
      <tr><td>6x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>Included</td><td>Included</td></tr>
      <tr><td>12x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>264 €</td><td>354 €</td></tr>
      <tr><td>18x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>528 €</td><td>708 €</td></tr>
      <tr><td>24x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>792 €</td><td>1062 €</td></tr>
      <tr class="server-header"><td>HGR-HCI-a2 &#8211; 2024 &#8211; DUAL AMD EPYC 9354</td><td>1139.99 €</td><td>1274.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>384GB DDR5 ECC 4800MHz</td><td>Included</td><td>Included</td></tr>
      <tr><td>384GB DDR5 ECC 4800MHz</td><td>Included</td><td>Included</td></tr>
      <tr><td>512GB DDR5 ECC 4800MHz</td><td>64 €</td><td>74 €</td></tr>
      <tr><td>768GB DDR5 ECC 4800MHz</td><td>400 €</td><td>360 €</td></tr>
      <tr><td>1TB DDR5 ECC 4800MHz</td><td>320 €</td><td>600 €</td></tr>
      <tr><td>1TB DDR5 ECC 4800MHz</td><td>320 €</td><td>600 €</td></tr>
      <tr><td>1.5TB DDR5 ECC 4800MHz</td><td>384 €</td><td>1272 €</td></tr>
      <tr><td>1.5TB DDR5 ECC 4800MHz</td><td>384 €</td><td>1272 €</td></tr>
      <tr><td>RAM 2304GB DDR5 ECC 4800MHz</td><td>960 €</td><td>2208 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Storage</td></tr>
      <tr><td>6x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>Included</td><td>Included</td></tr>
      <tr><td>12x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>264 €</td><td>354 €</td></tr>
      <tr><td>18x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>528 €</td><td>708 €</td></tr>
      <tr><td>24x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>792 €</td><td>1062 €</td></tr>
      <tr class="server-header"><td>HGR-HCI-i1 &#8211; 2024 &#8211; DUAL INTEL XEON GOLD 5515+</td><td>849.99 €</td><td>949.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>256GB DDR5 ECC 4800MHz</td><td>Included</td><td>Included</td></tr>
      <tr><td>512GB DDR5 ECC 4800MHz</td><td>128 €</td><td>240 €</td></tr>
      <tr><td>1TB DDR5 ECC 4800MHz</td><td>384 €</td><td>800 €</td></tr>
      <tr><td>1.5TB DDR5 ECC 4800MHz</td><td>512 €</td><td>1472 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Storage</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID</td><td>Included</td><td>Included</td></tr>
      <tr><td>6x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>Included</td><td>Included</td></tr>
      <tr><td>12x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>264 €</td><td>354 €</td></tr>
      <tr><td>18x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>528 €</td><td>708 €</td></tr>
      <tr><td>24x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>792 €</td><td>1062 €</td></tr>
      <tr class="server-header"><td>HGR-HCI-i2 &#8211; 2024 &#8211; DUAL INTEL XEON GOLD 6526Y</td><td>929.99 €</td><td>1039.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>256GB DDR5 ECC 4800MHz</td><td>Included</td><td>Included</td></tr>
      <tr><td>512GB DDR5 ECC 4800MHz</td><td>128 €</td><td>240 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Storage</td></tr>
      <tr><td>6x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>Included</td><td>Included</td></tr>
      <tr><td>12x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>264 €</td><td>354 €</td></tr>
      <tr><td>18x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>528 €</td><td>708 €</td></tr>
      <tr><td>24x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>792 €</td><td>1062 €</td></tr>
      <tr class="server-header"><td>HGR-HCI-i3 &#8211; 2024 &#8211; DUAL INTEL XEON GOLD 6542Y</td><td>999.99 €</td><td>1119.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>256GB DDR5 ECC 4800MHz</td><td>Included</td><td>Included</td></tr>
      <tr><td>512GB DDR5 ECC 4800MHz</td><td>128 €</td><td>240 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Storage</td></tr>
      <tr><td>6x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>Included</td><td>Included</td></tr>
      <tr><td>12x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>264 €</td><td>354 €</td></tr>
      <tr><td>18x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>528 €</td><td>708 €</td></tr>
      <tr><td>24x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>792 €</td><td>1062 €</td></tr>
      <tr class="server-header"><td>HGR-HCI-i4 &#8211; 2024 &#8211; DUAL INTEL XEON GOLD 6554S</td><td>1079.99 €</td><td>1209.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>256GB DDR5 ECC 4800MHz</td><td>Included</td><td>Included</td></tr>
      <tr><td>512GB DDR5 ECC 4800MHz</td><td>128 €</td><td>240 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Storage</td></tr>
      <tr><td>6x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>Included</td><td>Included</td></tr>
      <tr><td>12x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>264 €</td><td>354 €</td></tr>
      <tr><td>18x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>528 €</td><td>708 €</td></tr>
      <tr><td>24x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>792 €</td><td>1062 €</td></tr>
      <tr class="server-header"><td>HGR-SAP-1 &#8211; 2024 &#8211; DUAL INTEL XEON GOLD 6226R</td><td>1011.99 €</td><td>1254.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>192GB DDR4 ECC 2933MHz</td><td>Included</td><td>Included</td></tr>
      <tr><td>384GB DDR4 ECC 2933MHz</td><td>96 €</td><td>216 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Storage</td></tr>
      <tr><td>6x SSD SAS 3.84TB Enterprise Class Hard RAID</td><td>Included</td><td>Included</td></tr>
      <tr><td>2x SSD SATA 480GB</td><td>Included</td><td>Included</td></tr>
      <tr><td>12x SSD SAS 3.84TB Enterprise Class Hard RAID</td><td>264 €</td><td>354 €</td></tr>
      <tr><td>24x SSD SAS 3.84TB Enterprise Class Hard RAID</td><td>792 €</td><td>1062 €</td></tr>
      <tr class="server-header"><td>HGR-SAP-2 &#8211; 2024 &#8211; DUAL INTEL XEON GOLD 6242R</td><td>1121.99 €</td><td>1391.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>384GB DDR4 ECC 2933MHz</td><td>Included</td><td>Included</td></tr>
      <tr><td>RAM 768GB DDR4 ECC 2933MHz</td><td>192 €</td><td>312 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Storage</td></tr>
      <tr><td>6x SSD SAS 3.84TB Enterprise Class Hard RAID</td><td>Included</td><td>Included</td></tr>
      <tr><td>2x SSD SATA 480GB</td><td>Included</td><td>Included</td></tr>
      <tr><td>12x SSD SAS 3.84TB Enterprise Class Hard RAID</td><td>264 €</td><td>354 €</td></tr>
      <tr><td>24x SSD SAS 3.84TB Enterprise Class Hard RAID</td><td>792 €</td><td>1062 €</td></tr>
      <tr class="server-header"><td>HGR-SAP-3 &#8211; 2024 &#8211; DUAL INTEL XEON GOLD 6248R</td><td>1231.99 €</td><td>1527.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>RAM 768GB DDR4 ECC 2933MHz</td><td>192 €</td><td>Included</td></tr>
      <tr><td>RAM 1.5TB DDR4 ECC 2933MHz</td><td>384 €</td><td>1032 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Storage</td></tr>
      <tr><td>6x SSD SAS 3.84TB Enterprise Class Hard RAID</td><td>Included</td><td>Included</td></tr>
      <tr><td>2x SSD SATA 480GB</td><td>Included</td><td>Included</td></tr>
      <tr><td>12x SSD SAS 3.84TB Enterprise Class Hard RAID</td><td>264 €</td><td>354 €</td></tr>
      <tr><td>24x SSD SAS 3.84TB Enterprise Class Hard RAID</td><td>792 €</td><td>1062 €</td></tr>
      <tr class="server-header"><td>HGR-SDS-1 &#8211; 2024 &#8211; DUAL INTEL XEON GOLD 5515+</td><td>999.99 €</td><td>1119.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Storage</td></tr>
      <tr><td>6x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>Included</td><td>Included</td></tr>
      <tr><td>6x SSD NVMe 15.36TB Enterprise Class Soft RAID</td><td>475 €</td><td>546 €</td></tr>
      <tr><td>12x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>450 €</td><td>630 €</td></tr>
      <tr><td>18x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>900 €</td><td>1260 €</td></tr>
      <tr><td>12x SSD NVMe 15.36TB Enterprise Class Soft RAID</td><td>1399 €</td><td>1722 €</td></tr>
      <tr><td>24x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>1350 €</td><td>1890 €</td></tr>
      <tr><td>18x SSD NVMe 15.36TB Enterprise Class Soft RAID</td><td>2323 €</td><td>2898 €</td></tr>
      <tr><td>24x SSD NVMe 15.36TB Enterprise Class Soft RAID</td><td>3247 €</td><td>4074 €</td></tr>
      <tr class="server-header"><td>HGR-SDS-2 &#8211; 2024 &#8211; DUAL INTEL XEON GOLD 6542Y</td><td>1149.99 €</td><td>1289.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Storage</td></tr>
      <tr><td>6x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>Included</td><td>Included</td></tr>
      <tr><td>6x SSD NVMe 15.36TB Enterprise Class Soft RAID</td><td>475 €</td><td>546 €</td></tr>
      <tr><td>12x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>450 €</td><td>630 €</td></tr>
      <tr><td>18x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>900 €</td><td>1260 €</td></tr>
      <tr><td>12x SSD NVMe 15.36TB Enterprise Class Soft RAID</td><td>1399 €</td><td>1722 €</td></tr>
      <tr><td>24x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>1350 €</td><td>1890 €</td></tr>
      <tr><td>18x SSD NVMe 15.36TB Enterprise Class Soft RAID</td><td>2323 €</td><td>2898 €</td></tr>
      <tr><td>24x SSD NVMe 15.36TB Enterprise Class Soft RAID</td><td>3247 €</td><td>4074 €</td></tr>
      <tr class="server-header"><td>HGR-STOR-1 &#8211; 2024 &#8211; INTEL XEON GOLD 6554S</td><td>1199.99 €</td><td>1399.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>128GB DDR5 ECC 4800MHz</td><td>Included</td><td>Included</td></tr>
      <tr><td>256GB DDR5 ECC 4800MHz</td><td>64 €</td><td>200 €</td></tr>
      <tr><td>512GB DDR5 ECC 4800MHz</td><td>192 €</td><td>440 €</td></tr>
      <tr><td>768GB DDR5 ECC 4800MHz</td><td>400 €</td><td>760 €</td></tr>
      <tr><td>768GB DDR5 ECC 4800MHz</td><td>320 €</td><td>760 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Storage</td></tr>
      <tr><td>24x HDD SAS 22TB Enterprise Class Soft RAID</td><td>Included</td><td>Included</td></tr>
      <tr><td>24x HDD SAS 22TB + 2x SSD NVMe 3.84TB High perf. cache Enterprise Class Soft RAID</td><td>88 €</td><td>118 €</td></tr>
      <tr><td>24x HDD SAS 22TB Enterprise Class Hard RAID</td><td>66 €</td><td>120 €</td></tr>
      <tr><td>24x HDD SAS 22TB + 2x SSD NVMe 7.68TB High perf. cache Enterprise Class Soft RAID</td><td>150 €</td><td>210 €</td></tr>
      <tr><td>24x HDD SAS 22TB + 2x SSD NVMe 3.84TB High perf. cache Enterprise Class Hard RAID</td><td>154 €</td><td>238 €</td></tr>
      <tr><td>24x HDD SAS 22TB + 2x SSD NVMe 7.68TB High perf. cache Enterprise Class Hard RAID</td><td>216 €</td><td>330 €</td></tr>
      <tr><td>24x HDD SAS 22TB + 2x SSD NVMe 15.36TB High perf. cache Enterprise Class Soft RAID</td><td>308 €</td><td>392 €</td></tr>
      <tr><td>24x HDD SAS 22TB + 2x SSD NVMe 15.36TB High perf. cache Enterprise Class Hard RAID</td><td>374 €</td><td>512 €</td></tr>
      <tr><td>36x HDD SAS 22TB Enterprise Class Soft RAID</td><td>384 €</td><td>516 €</td></tr>
      <tr><td>36x HDD SAS 22TB + 2x SSD NVMe 3.84TB High perf. cache Enterprise Class Soft RAID</td><td>472 €</td><td>634 €</td></tr>
      <tr><td>36x HDD SAS 22TB Enterprise Class Hard RAID</td><td>450 €</td><td>696 €</td></tr>
      <tr><td>36x HDD SAS 22TB + 2x SSD NVMe 7.68TB High perf. cache Enterprise Class Soft RAID</td><td>534 €</td><td>726 €</td></tr>
      <tr><td>36x HDD SAS 22TB + 2x SSD NVMe 3.84TB High perf. cache Enterprise Class Hard RAID</td><td>538 €</td><td>814 €</td></tr>
      <tr><td>36x HDD SAS 22TB + 2x SSD NVMe 7.68TB High perf. cache Enterprise Class Hard RAID</td><td>600 €</td><td>906 €</td></tr>
      <tr><td>36x HDD SAS 22TB + 2x SSD NVMe 15.36TB High perf. cache Enterprise Class Soft RAID</td><td>692 €</td><td>908 €</td></tr>
      <tr><td>36x HDD SAS 22TB + 2x SSD NVMe 15.36TB High perf. cache Enterprise Class Hard RAID</td><td>758 €</td><td>1088 €</td></tr>
    </tbody>
  </table>
</div>
</body>
</html>
<div class='table-container'>
  <h2>IPs</h2>
  <table class="dataframe">
    <thead>
      <tr style="text-align: right;">
        <th>Reference</th>
        <th>Old public price (Excl. VAT / Month)</th>
        <th>New Public Price (Excl. VAT / Month)</th>
      </tr>
    </thead>
    <tbody>
      <tr><td>Additional IPv4</td><td>1.50 €</td><td>2.00 €</td></tr>
    </tbody>
  </table>
</div>



<div class='table-container'>
  <h2>VPS</h2>
  <table class="dataframe">
    <thead>
      <tr style="text-align: right;">
        <th>Family</th>
        <th>Reference</th>
        <th>Commit</th>
        <th>Old public price (Excl. VAT / Month)</th>
        <th>New Public Price (Excl. VAT / Month)</th>
      </tr>
    </thead>
    <tbody>
      <tr><td>VPS 2026</td><td>VPS-1</td><td>Monthly</td><td>4.49 €</td><td>6.49 €</td></tr>
      <tr><td>VPS 2026</td><td>VPS-2</td><td>Monthly</td><td>6.99 €</td><td>9.99 €</td></tr>
      <tr><td>VPS 2026</td><td>VPS-3</td><td>Monthly</td><td>13.99 €</td><td>19.99 €</td></tr>
      <tr><td>VPS 2026</td><td>VPS-4</td><td>Monthly</td><td>24.99 €</td><td>36.99 €</td></tr>
      <tr><td>VPS 2026</td><td>VPS-5</td><td>Monthly</td><td>36.99 €</td><td>54.99 €</td></tr>
      <tr><td>VPS 2026</td><td>VPS-6</td><td>Monthly</td><td>48.99 €</td><td>72.99 €</td></tr>
      <tr><td>VPS 2026</td><td>VPSLZ-1</td><td>Monthly</td><td>5.49 €</td><td>7.49 €</td></tr>
      <tr><td>VPS 2026</td><td>VPS-1</td><td>prepayment 6 months</td><td>25.56 €</td><td>36.99 €</td></tr>
      <tr><td>VPS 2026</td><td>VPS-2</td><td>prepayment 6 months</td><td>39.84 €</td><td>56.94 €</td></tr>
      <tr><td>VPS 2026</td><td>VPS-3</td><td>prepayment 6 months</td><td>79.74 €</td><td>113.94 €</td></tr>
      <tr><td>VPS 2026</td><td>VPS-4</td><td>prepayment 6 months</td><td>142.44 €</td><td>210.84 €</td></tr>
      <tr><td>VPS 2026</td><td>VPS-5</td><td>prepayment 6 months</td><td>210.84 €</td><td>313.44 €</td></tr>
      <tr><td>VPS 2026</td><td>VPS-6</td><td>prepayment 6 months</td><td>279.24 €</td><td>416.04 €</td></tr>
      <tr><td>VPS 2026</td><td>VPSLZ-1</td><td>prepayment 6 months</td><td>31.26 €</td><td>42.69 €</td></tr>
      <tr><td>VPS 2026</td><td>VPS-1</td><td>prepayment 12 months</td><td>45.72 €</td><td>66.19 €</td></tr>
      <tr><td>VPS 2026</td><td>VPS-2</td><td>prepayment 12 months</td><td>71.28 €</td><td>101.89 €</td></tr>
      <tr><td>VPS 2026</td><td>VPS-3</td><td>prepayment 12 months</td><td>142.68 €</td><td>203.89 €</td></tr>
      <tr><td>VPS 2026</td><td>VPS-4</td><td>prepayment 12 months</td><td>254.88 €</td><td>377.29 €</td></tr>
      <tr><td>VPS 2026</td><td>VPS-5</td><td>prepayment 12 months</td><td>377.28 €</td><td>560.89 €</td></tr>
      <tr><td>VPS 2026</td><td>VPS-6</td><td>prepayment 12 months</td><td>499.68 €</td><td>744.49 €</td></tr>
      <tr><td>VPS 2026</td><td>VPSLZ-1</td><td>prepayment 12 months</td><td>55.92 €</td><td>76.39 €</td></tr>
    </tbody>
  </table>
</div>

</body>
</html>



<p></p>



<p></p>



<p></p>
<img decoding="async" src="//blog.ovhcloud.com/wp-content/plugins/matomo/app/matomo.php?idsite=1&amp;rec=1&amp;url=https%3A%2F%2Fblog.ovhcloud.com%2Fpricing-evolution-of-public-cloud-bare-metal-and-vps-at-ovhcloud%2F&amp;action_name=Pricing%20evolution%20of%20Public%20Cloud%2C%20Bare%20Metal%20and%20VPS%20at%20OVHcloud&amp;urlref=https%3A%2F%2Fblog.ovhcloud.com%2Ffeed%2F" style="border:0;width:0;height:0" width="0" height="0" alt="" />]]></content:encoded>
					
		
		
			</item>
		<item>
		<title>Évolutions tarifaires de Public Cloud, Bare Metal et VPS chez OVHcloud</title>
		<link>https://blog.ovhcloud.com/evolutions-tarifaires-de-public-cloud-bare-metal-et-vps-chez-ovhcloud/</link>
		
		<dc:creator><![CDATA[Octave Klaba]]></dc:creator>
		<pubDate>Thu, 05 Mar 2026 12:59:14 +0000</pubDate>
				<category><![CDATA[OVHcloud Product News]]></category>
		<category><![CDATA[Uncategorized]]></category>
		<category><![CDATA[AI]]></category>
		<category><![CDATA[bare metal]]></category>
		<category><![CDATA[Infrastructure]]></category>
		<category><![CDATA[OVHcloud]]></category>
		<category><![CDATA[OVHcloud News]]></category>
		<category><![CDATA[Private Cloud]]></category>
		<category><![CDATA[Public Cloud]]></category>
		<category><![CDATA[Storage]]></category>
		<guid isPermaLink="false">https://blog.ovhcloud.com/?p=30712</guid>

					<description><![CDATA[Depuis l’automne 2025, le marché mondial de la mémoire subit une rupture majeure. Encore peu perceptible pour les utilisateurs finaux, cette évolution transforme en profondeur le coût du matériel informatique et, par effet direct, celui du cloud. Cet article propose un décryptage de cette crise structurelle, de ses impacts concrets et des choix stratégiques qu’OVHcloud [&#8230;]<img src="//blog.ovhcloud.com/wp-content/plugins/matomo/app/matomo.php?idsite=1&amp;rec=1&amp;url=https%3A%2F%2Fblog.ovhcloud.com%2Fevolutions-tarifaires-de-public-cloud-bare-metal-et-vps-chez-ovhcloud%2F&amp;action_name=%C3%89volutions%20tarifaires%20de%20Public%20Cloud%2C%20Bare%20Metal%20et%20VPS%20chez%20OVHcloud&amp;urlref=https%3A%2F%2Fblog.ovhcloud.com%2Ffeed%2F" style="border:0;width:0;height:0" width="0" height="0" alt="" />]]></description>
										<content:encoded><![CDATA[
<p>Depuis l’automne 2025, le marché mondial de la mémoire subit une rupture majeure. Encore peu perceptible pour les utilisateurs finaux, cette évolution transforme en profondeur le coût du matériel informatique et, par effet direct, celui du cloud.<br><br>Cet article propose un décryptage de cette crise structurelle, de ses impacts concrets et des choix stratégiques qu’OVHcloud met en œuvre pour en limiter les effets.</p>



<h2 class="wp-block-heading"><strong>Un basculement industriel vers les GPU</strong></h2>



<p>À l’échelle mondiale, les trois grands fabricants de mémoire ont réorienté une part importante de leurs capacités de production pour répondre à la demande massive en GPU, en particulier pour les usages liés à l’IA et au calcul haute performance.</p>



<p>Cette réallocation s’est effectuée sans réduction équivalente de la demande historique en mémoire vive et en stockage, générant une pression simultanée sur plusieurs segments du marché.<br><br>Les conséquences sont immédiates et visibles :</p>



<ul class="wp-block-list">
<li>tension sur l’offre, avec des stocks réduits et des délais d’approvisionnement allongés&nbsp;;</li>



<li>hausse continue des prix de la RAM et des disques depuis septembre 2025&nbsp;;</li>



<li>instabilité durable du marché, qui ne devrait retrouver un nouvel équilibre qu’à l’horizon fin 2026.</li>
</ul>



<h2 class="wp-block-heading"><strong>Une inflation durable des composants mémoire</strong></h2>



<p>Même après la stabilisation du marché, les prix ne devraient pas retrouver leurs niveaux historiques avant 2028, le temps nécessaire pour que de nouvelles capacités de production soient réellement opérationnelles.</p>



<p>Cette évolution bouleverse profondément les fondamentaux économiques du matériel informatique, tant pour les infrastructures <em>on-premise</em> que pour le cloud. Selon les configurations, l’impact tarifaire lié à la RAM et au stockage pourrait atteindre +15 % à +300&nbsp;% par rapport aux prix de 2025, en fonction des volumes de mémoire et de capacité disque déployés.</p>



<p>Ce changement d’échelle est à la fois brutal et inédit, sans équivalent récent sur le marché mondial.</p>



<h2 class="wp-block-heading"><strong>Un marché sous tension même à prix élevé</strong></h2>



<p>Paradoxalement, la hausse des prix ne suffit pas à sécuriser la disponibilité des composants. Aujourd’hui, pour garantir la livraison de volumes de RAM ou de disques, il est nécessaire pour les fournisseurs de cloud de passer commande jusqu’à 12 mois à l’avance, sans connaître le prix final au moment de l’achat.</p>



<p>En pratique, les tarifs ne sont communiqués qu’un à deux mois après la livraison, selon l’évolution de l’offre et de la demande sur le trimestre concerné. Cette incertitude exerce une pression inédite sur les acteurs industriels et les fournisseurs de cloud, affectant simultanément la production et la distribution.</p>



<h2 class="wp-block-heading"><strong>Vers un nouvel équilibre mondial de la demande</strong></h2>



<p>Cette situation aura inévitablement des répercussions sur les volumes commandés. Certains clients jugeront les prix trop élevés et limiteront leurs investissements, tandis que&nbsp;d’autres, faute d’alternative, continueront à passer commande malgré tout.</p>



<p>Ce jeu de forces opposées devrait conduire à un nouvel équilibre mondial, mais à un niveau de prix nettement supérieur. Les projections actuelles anticipent une augmentation de la RAM de +250 % à +300 % à la fin 2026, par rapport à septembre 2025.</p>



<h2 class="wp-block-heading"><strong>Notre stratégie pour amortir le choc</strong></h2>



<p>Face à cette réalité, OVHcloud a choisi de ne pas répercuter automatiquement l’intégralité de la hausse des composants sur ses clients.</p>



<p>Pour le cloud déployé entre 2026 et 2028 — incluant le Public Cloud, le Private Cloud et le Bare Metal — l’augmentation moyenne des prix sera limitée, entre +9 % et +11 %, malgré des coûts de RAM et de disques nettement plus élevés.</p>



<p>Pour compenser cet écart, un ajustement modéré est prévu sur les offres déployées avant 2025, de +2 % à +6 %, en fonction de l’ancienneté du matériel, ainsi qu&#8217;une évolution des tarifs des IPv4. Cette dernière ne devrait pas avoir d&#8217;impact significatif sur le budget de nos clients, le coût des adresses IP représentant une part limitée par rapport aux autres ressources d&#8217;un projet cloud.</p>



<p>Notre objectif est clair : préserver une cohérence tarifaire sur l’ensemble des gammes 2021-2028 et préparer un retour progressif à la normale en 2029.</p>



<h2 class="wp-block-heading"><strong>Investissements continus et évolution des offres</strong></h2>



<p>Au-delà des ajustements tarifaires, cette période se caractérise par des investissements soutenus dans nos offres et dans l’expérience client.<br><br>Malgré la forte pression liée à l’augmentation des coûts des composants, nous continuons à faire évoluer nos services afin d’apporter davantage de valeur à nos clients.<br><br>Concrètement, cela se traduit par&nbsp;:</p>



<ul class="wp-block-list">
<li>un renforcement progressif des dispositifs de support&nbsp;;</li>



<li>une augmentation des ressources incluses dans certaines gammes&nbsp;;</li>



<li>une modernisation de nos infrastructures de calcul et de stockage.</li>
</ul>



<p>Ces initiatives témoignent de notre volonté de ne pas réduire cette phase à une simple répercussion des hausses de coûts, mais de maintenir une dynamique d’amélioration de nos services, même dans un contexte économique contraint.</p>



<h2 class="wp-block-heading"><strong>Calendrier et modalités d’application</strong></h2>



<p>Nos clients ont déjà reçu des emails détaillant les impacts précis sur leurs services. Les nouveaux tarifs seront appliqués à compter du 1<sup>er</sup> avril 2026.</p>



<p>Jusqu’à cette date, il est possible de renouveler les services aux tarifs actuels pour une durée pouvant aller jusqu’à 2 ans. Dans tous les cas, les nouveaux prix ne s’appliqueront qu’à l’issue de la période d’engagement en cours.</p>



<h2 class="wp-block-heading"><strong>Une période d’incertitude et un avantage stratégique</strong></h2>



<p>Nous traversons une phase exceptionnellement imprévisible, où la visibilité sur les marchés dépasse rarement une à deux semaines. L’espoir demeure que les prix se stabilisent durablement dès 2026, afin d’éviter de nouvelles annonces défavorables.</p>



<p>Dans ce contexte tendu, disposer d’une chaîne d’approvisionnement mondiale et de deux usines de production internes constitue un avantage stratégique majeur. Cela nous permet de continuer à recevoir des composants et à produire des serveurs, là où la pénurie de mémoire touche une grande partie du marché.</p>



<h2 class="wp-block-heading"><strong>Nos tarifs</strong></h2>



<p>Vous trouverez ci-dessous nos nouveaux tarifs :<br>&#8211; <strong>Public Cloud</strong> : les prix ci-dessous sont affichés à l&#8217;heure et avec OS Linux. Vous trouverez sur <a href="https://www.ovhcloud.com/fr/public-cloud/prices/" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">notre page Tarifs</a> les prix des instances de machines virtuelle consommées au mois (b2, c2, r2) ou en <a href="https://www.ovhcloud.com/fr/public-cloud/savings-plan/" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">Savings Plan</a> (b3, c3, r3) ainsi que les tarifs avec licences Windows. <br>&#8211; <strong>Tous nos tarifs VPS</strong>, Floating IPs &amp; IP additionnelles.<br>&#8211; <strong>Bare Metal</strong> : les prix affichés correspondent à un engagement d’un mois ; des remises supplémentaires sont appliquées en cas de prépaiement sur 12 ou 24 mois. <strong>Les prix des options sont ceux des nouvelles commandes uniquement.</strong> Le renouvellement d&#8217;options qui a été communiqué par email à nos clients sera quant à lui limité à +10% sur les options de disques, +15% sur les options de RAM. <br></p>



<p>Pour toutes les souscriptions existantes renouvelées <strong>avant le 1er avril</strong>, vous pouvez conserver votre tarif actuel pendant toute la durée d’engagement choisie, à compter de votre date de renouvellement.</p>



<p><span class="wp-rich-text-font-awesome-icon wp-font-awesome-icon"><svg aria-hidden="true" focusable="false" data-prefix="fas" data-icon="bell" class="svg-inline--fa fa-bell " role="img" xmlns="http://www.w3.org/2000/svg" viewBox="0 0 448 512"><path fill="currentColor" d="M224 0c-17.7 0-32 14.3-32 32V51.2C119 66 64 130.6 64 208v18.8c0 47-17.3 92.4-48.5 127.6l-7.4 8.3c-8.4 9.4-10.4 22.9-5.3 34.4S19.4 416 32 416H416c12.6 0 24-7.4 29.2-18.9s3.1-25-5.3-34.4l-7.4-8.3C401.3 319.2 384 273.9 384 226.8V208c0-77.4-55-142-128-156.8V32c0-17.7-14.3-32-32-32zm45.3 493.3c12-12 18.7-28.3 18.7-45.3H224 160c0 17 6.7 33.3 18.7 45.3s28.3 18.7 45.3 18.7s33.3-6.7 45.3-18.7z"></path></svg></span> Veuillez noter que les catégories de produits suivantes <strong><span style="text-decoration: underline;">ne sont pas concernées</span></strong> par l&#8217;évolution de nos tarifs :<br>&#8211; Public Cloud &#8211; Compute : <strong>Cloud GPUs</strong> <strong>et Metal Instances</strong><br>&#8211; Public Cloud &#8211; Container : <strong>Managed Kubernetes, Managed Registries &amp; Managed Rancher</strong><br>&#8211; Public Cloud &#8211; Network : <strong>Load Balancer, Gateway</strong>. Le trafic réseau Public et Privé reste inclus. <br>&#8211; Public Cloud &#8211; Storage : <strong>Object Storage, Block Storage</strong>. <br>&#8211; Public Cloud &#8211; Analytics : <strong>Data Platform</strong><br>&#8211; Public Cloud &#8211; AI &amp; Machine Learning : <strong>AI Solutions (AI Notebook, AI Training, AI Deploy) et AI Endpoint</strong>s<br>&#8211; Public Cloud &#8211; Quantum : <strong>Emulators &amp; QPUs</strong><br>&#8211; Bare Metal &#8211; stockage : Veeam Enterprise Plus, HYCU, Back-up Agent, NAS-HA, Cloud Disk Array<br>&#8211; Bare Metal : Gammes <strong>Kimsufi et SoYouStart</strong><br>&#8211; Private Cloud : offres <strong>VMware</strong> et offres de stockage (Veeam Enterprise plus, HYCU, Back-up Agent)<br></p>



<!DOCTYPE html>
<html lang="fr">
<head>
  <meta charset="UTF-8">
  <title>Tableaux des prix</title>
  <style>
    body { font-family: Arial, Helvetica, sans-serif; margin: 40px; background-color: #f4f6f8; }
    h1 { color: #000E9C; }
    h2 { margin-top: 50px; color: #1a2b3c; }
    h3 { margin-top: 30px; color: #1a2b3c; }
    .table-container { margin-top: 20px; margin-bottom: 40px; }
    table { border-collapse: collapse; width: 100%; background: white; border-radius: 8px; overflow: hidden; box-shadow: 0 2px 8px rgba(0,0,0,0.08); }
    thead { background-color: #000E9C; color: white; }
    th { padding: 12px 15px; text-align: center; font-weight: bold; text-transform: uppercase; font-size: 13px; }
    td { padding: 12px 15px; font-size: 14px; }
    td:nth-child(n+2) { text-align: center; }
    tbody tr { border-bottom: 1px solid #e6e9ed; }
    tbody tr:nth-child(even) { background-color: #f8fafc; }
    tbody tr:hover { background-color: #eef3ff; }
    .section-header td { background-color: #e8edf5; font-weight: bold; font-style: italic; color: #1a2b3c; text-align: left !important; }
  </style>
</head>
<body>

<!-- PUBLIC CLOUD -->
<div class='table-container'>
  <h2>Public Cloud – Instances de Machines Virtuelles</h2>

  <h3>General Purpose</h3>
 <p class="mention">Ci-dessous, les tarifs standards horaires et mensuels pour les instances avec OS Linux, sans Savings Plan ni autre remise supplémentaire.</p>
  <table class="dataframe">
    <thead>
      <tr><th>Reference</th><th>Ancien prix public (HT / heure)</th><th>Nouveau prix public (HT / heure)</th></tr>
    </thead>
    <tbody>
      <tr><td>b3-8</td><td>0,0465 €</td><td>0,0512 €</td></tr>
      <tr><td>b3-16</td><td>0,093 €</td><td>0,1023 €</td></tr>
      <tr><td>b3-32</td><td>0,186 €</td><td>0,2046 €</td></tr>
      <tr><td>b3-64</td><td>0,372 €</td><td>0,4092 €</td></tr>
      <tr><td>b3-128</td><td>0,7439 €</td><td>0,819 €</td></tr>
      <tr><td>b3-256</td><td>1,4878 €</td><td>1,637 €</td></tr>
      <tr><td>b3-512</td><td>2,9756 €</td><td>3,274 €</td></tr>
      <tr><td>b3-640</td><td>3,7195 €</td><td>4,092 €</td></tr>
      <tr><td>b2-7</td><td>0,0681 €</td><td>0,0709 €</td></tr>
      <tr><td>b2-15</td><td>0,129 €</td><td>0,1342 €</td></tr>
      <tr><td>b2-30</td><td>0,261 €</td><td>0,2715 €</td></tr>
      <tr><td>b2-60</td><td>0,505 €</td><td>0,526 €</td></tr>
      <tr><td>b2-120</td><td>0,993 €</td><td>1,033 €</td></tr>
    </tbody>
  </table>

  <h3>Compute Optimized</h3>
 <p class="mention">Ci-dessous, les tarifs standards horaires et mensuels pour les instances avec OS Linux, sans Savings Plan ni autre remise supplémentaire.</p>
  <table class="dataframe">
    <thead>
      <tr><th>Reference</th><th>Ancien prix public (HT / heure)</th><th>Nouveau prix public (HT / heure)</th></tr>
    </thead>
    <tbody>
      <tr><td>c3-4</td><td>0,0415 €</td><td>0,0457 €</td></tr>
      <tr><td>c3-8</td><td>0,083 €</td><td>0,0913 €</td></tr>
      <tr><td>c3-16</td><td>0,1659 €</td><td>0,1825 €</td></tr>
      <tr><td>c3-32</td><td>0,3318 €</td><td>0,365 €</td></tr>
      <tr><td>c3-64</td><td>0,6637 €</td><td>0,7301 €</td></tr>
      <tr><td>c3-128</td><td>1,3274 €</td><td>1,461 €</td></tr>
      <tr><td>c3-256</td><td>2,6547 €</td><td>2,921 €</td></tr>
      <tr><td>c3-320</td><td>3,3184 €</td><td>3,651 €</td></tr>
      <tr><td>c2-7</td><td>0,0978 €</td><td>0,1018 €</td></tr>
      <tr><td>c2-15</td><td>0,19 €</td><td>0,1976 €</td></tr>
      <tr><td>c2-30</td><td>0,383 €</td><td>0,3984 €</td></tr>
      <tr><td>c2-60</td><td>0,749 €</td><td>0,779 €</td></tr>
      <tr><td>c2-120</td><td>1,48 €</td><td>1,54 €</td></tr>
    </tbody>
  </table>

  <h3>Memory Optimized</h3>
 <p class="mention">Ci-dessous, les tarifs standards horaires et mensuels pour les instances avec OS Linux, sans Savings Plan ni autre remise supplémentaire.</p>
  <table class="dataframe">
    <thead>
      <tr><th>Reference</th><th>Ancien prix public (HT / heure)</th><th>Nouveau prix public (HT / heure)</th></tr>
    </thead>
    <tbody>
      <tr><td>r3-16</td><td>0,0602 €</td><td>0,0663 €</td></tr>
      <tr><td>r3-32</td><td>0,1203 €</td><td>0,1324 €</td></tr>
      <tr><td>r3-64</td><td>0,2407 €</td><td>0,2648 €</td></tr>
      <tr><td>r3-128</td><td>0,4813 €</td><td>0,53 €</td></tr>
      <tr><td>r3-256</td><td>0,9627 €</td><td>1,059 €</td></tr>
      <tr><td>r3-512</td><td>1,9254 €</td><td>2,118 €</td></tr>
      <tr><td>r3-1024</td><td>3,8508 €</td><td>4,236 €</td></tr>
      <tr><td>r2-15</td><td>0,0978 €</td><td>0,1018 €</td></tr>
      <tr><td>r2-30</td><td>0,113 €</td><td>0,1176 €</td></tr>
      <tr><td>r2-60</td><td>0,22 €</td><td>0,2288 €</td></tr>
      <tr><td>r2-120</td><td>0,443 €</td><td>0,461 €</td></tr>
      <tr><td>r2-240</td><td>0,871 €</td><td>0,906 €</td></tr>
    </tbody>
  </table>
</div>

<!-- PUBLIC CLOUD DATABASES -->
<div class='table-container'>
  <h2>Public Cloud – Databases</h2>

  <h3>MySQL</h3>
  <table class="dataframe">
    <thead>
      <tr><th>Reference</th><th>Ancien prix public (HT / heure / node)</th><th>Nouveau prix public (HT / heure / node)</th><th>Nouveau prix public (HT / heure)</th></tr>
    </thead>
    <tbody>
      <tr><td>Essential DB1-4</td><td>0,068 €</td><td>0,0746 €</td><td>0,0746 €</td></tr>
      <tr><td>Essential DB1-7</td><td>0,1346 €</td><td>0,1477 €</td><td>0,1477 €</td></tr>
      <tr><td>Essential DB1-15</td><td>0,2705 €</td><td>0,2968 €</td><td>0,2968 €</td></tr>
      <tr><td>Essential DB1-30</td><td>0,5436 €</td><td>0,5967 €</td><td>0,5967 €</td></tr>
      <tr><td>Production B3-8</td><td>0,2129 €</td><td>0,223 €</td><td>0,446 €</td></tr>
      <tr><td>Production B3-16</td><td>0,4258 €</td><td>0,4461 €</td><td>0,8922 €</td></tr>
      <tr><td>Production B3-32</td><td>0,8515 €</td><td>0,8922 €</td><td>1,7844 €</td></tr>
      <tr><td>Production B3-64</td><td>1,703 €</td><td>1,7844 €</td><td>3,5688 €</td></tr>
      <tr><td>Production B3-128</td><td>3,4059 €</td><td>3,5688 €</td><td>7,1376 €</td></tr>
      <tr><td>Production B3-256</td><td>6,8118 €</td><td>7,1377 €</td><td>14,2754 €</td></tr>
      <tr><td>Business DB1-4</td><td>0,0865 €</td><td>0,0949 €</td><td>0,1898 €</td></tr>
      <tr><td>Business DB1-7</td><td>0,173 €</td><td>0,1899 €</td><td>0,3798 €</td></tr>
      <tr><td>Business DB1-15</td><td>0,346 €</td><td>0,3797 €</td><td>0,7594 €</td></tr>
      <tr><td>Business DB1-30</td><td>0,6933 €</td><td>0,761 €</td><td>1,522 €</td></tr>
      <tr><td>Business DB1-60</td><td>1,3878 €</td><td>1,5234 €</td><td>3,0468 €</td></tr>
      <tr><td>Business DB1-120</td><td>2,777 €</td><td>3,0484 €</td><td>6,0968 €</td></tr>
      <tr><td>Advanced B3-8</td><td>0,2295 €</td><td>0,2404 €</td><td>0,7212 €</td></tr>
      <tr><td>Advanced B3-16</td><td>0,4589 €</td><td>0,4808 €</td><td>1,4424 €</td></tr>
      <tr><td>Advanced B3-32</td><td>0,9177 €</td><td>0,9616 €</td><td>2,8848 €</td></tr>
      <tr><td>Advanced B3-64</td><td>1,8354 €</td><td>1,9232 €</td><td>5,7696 €</td></tr>
      <tr><td>Advanced B3-128</td><td>3,6708 €</td><td>3,8464 €</td><td>11,5392 €</td></tr>
      <tr><td>Advanced B3-256</td><td>7,3416 €</td><td>7,6928 €</td><td>23,0784 €</td></tr>
      <tr><td>Enterprise DB1-4</td><td>0,0879 €</td><td>0,0964 €</td><td>0,2892 €</td></tr>
      <tr><td>Enterprise DB1-7</td><td>0,173 €</td><td>0,1899 €</td><td>0,5697 €</td></tr>
      <tr><td>Enterprise DB1-15</td><td>0,346 €</td><td>0,3797 €</td><td>1,1391 €</td></tr>
      <tr><td>Enterprise DB1-30</td><td>0,6933 €</td><td>0,761 €</td><td>2,283 €</td></tr>
      <tr><td>Enterprise DB1-60</td><td>1,3878 €</td><td>1,5234 €</td><td>4,5702 €</td></tr>
      <tr><td>Enterprise DB1-120</td><td>2,777 €</td><td>3,0484 €</td><td>9,1452 €</td></tr>
    </tbody>
  </table>

  <h3>PostgreSQL</h3>
  <table class="dataframe">
    <thead>
      <tr><th>Reference</th><th>Ancien prix public (HT / heure / node)</th><th>Nouveau prix public (HT / heure / node)</th><th>Nouveau prix public (HT / heure)</th></tr>
    </thead>
    <tbody>
      <tr><td>Essential DB1-4</td><td>0,068 €</td><td>0,0746 €</td><td>0,0746 €</td></tr>
      <tr><td>Essential DB1-7</td><td>0,1346 €</td><td>0,1477 €</td><td>0,1477 €</td></tr>
      <tr><td>Essential DB1-15</td><td>0,2705 €</td><td>0,2968 €</td><td>0,2968 €</td></tr>
      <tr><td>Essential DB1-30</td><td>0,5436 €</td><td>0,5967 €</td><td>0,5967 €</td></tr>
      <tr><td>Production B3-8</td><td>0,2129 €</td><td>0,223 €</td><td>0,446 €</td></tr>
      <tr><td>Production B3-16</td><td>0,4258 €</td><td>0,4461 €</td><td>0,8922 €</td></tr>
      <tr><td>Production B3-32</td><td>0,8515 €</td><td>0,8922 €</td><td>1,7844 €</td></tr>
      <tr><td>Production B3-64</td><td>1,703 €</td><td>1,7844 €</td><td>3,5688 €</td></tr>
      <tr><td>Production B3-128</td><td>3,4059 €</td><td>3,5688 €</td><td>7,1376 €</td></tr>
      <tr><td>Production B3-256</td><td>6,8118 €</td><td>7,1377 €</td><td>14,2754 €</td></tr>
      <tr><td>Business DB1-4</td><td>0,0865 €</td><td>0,0949 €</td><td>0,1898 €</td></tr>
      <tr><td>Business DB1-7</td><td>0,173 €</td><td>0,1899 €</td><td>0,3798 €</td></tr>
      <tr><td>Business DB1-15</td><td>0,346 €</td><td>0,3797 €</td><td>0,7594 €</td></tr>
      <tr><td>Business DB1-30</td><td>0,6933 €</td><td>0,761 €</td><td>1,522 €</td></tr>
      <tr><td>Business DB1-60</td><td>1,3878 €</td><td>1,5234 €</td><td>3,0468 €</td></tr>
      <tr><td>Business DB1-120</td><td>2,777 €</td><td>3,0484 €</td><td>6,0968 €</td></tr>
      <tr><td>Advanced B3-8</td><td>0,2295 €</td><td>0,2404 €</td><td>0,7212 €</td></tr>
      <tr><td>Advanced B3-16</td><td>0,4589 €</td><td>0,4808 €</td><td>1,4424 €</td></tr>
      <tr><td>Advanced B3-32</td><td>0,9177 €</td><td>0,9616 €</td><td>2,8848 €</td></tr>
      <tr><td>Advanced B3-64</td><td>1,8354 €</td><td>1,9232 €</td><td>5,7696 €</td></tr>
      <tr><td>Advanced B3-128</td><td>3,6708 €</td><td>3,8464 €</td><td>11,5392 €</td></tr>
      <tr><td>Advanced B3-256</td><td>7,3416 €</td><td>7,6928 €</td><td>23,0784 €</td></tr>
      <tr><td>Enterprise DB1-4</td><td>0,0879 €</td><td>0,0964 €</td><td>0,2892 €</td></tr>
      <tr><td>Enterprise DB1-7</td><td>0,173 €</td><td>0,1899 €</td><td>0,5697 €</td></tr>
      <tr><td>Enterprise DB1-15</td><td>0,346 €</td><td>0,3797 €</td><td>1,1391 €</td></tr>
      <tr><td>Enterprise DB1-30</td><td>0,6933 €</td><td>0,761 €</td><td>2,283 €</td></tr>
      <tr><td>Enterprise DB1-60</td><td>1,3878 €</td><td>1,5234 €</td><td>4,5702 €</td></tr>
      <tr><td>Enterprise DB1-120</td><td>2,777 €</td><td>3,0484 €</td><td>9,1452 €</td></tr>
    </tbody>
  </table>

  <h3>Valkey</h3>
  <table class="dataframe">
    <thead>
      <tr><th>Reference</th><th>Ancien prix public (HT / heure / node)</th><th>Nouveau prix public (HT / heure / node)</th><th>Nouveau prix public (HT / heure)</th></tr>
    </thead>
    <tbody>
      <tr><td>Essential DB1-4</td><td>0,0591 €</td><td>0,0648 €</td><td>0,0648 €</td></tr>
      <tr><td>Essential DB1-7</td><td>0,1195 €</td><td>0,1311 €</td><td>0,1311 €</td></tr>
      <tr><td>Production B3-8</td><td>0,1409 €</td><td>0,1476 €</td><td>0,2952 €</td></tr>
      <tr><td>Production B3-16</td><td>0,3147 €</td><td>0,3297 €</td><td>0,6594 €</td></tr>
      <tr><td>Production B3-32</td><td>0,6295 €</td><td>0,6595 €</td><td>1,319 €</td></tr>
      <tr><td>Production B3-64</td><td>1,2588 €</td><td>1,319 €</td><td>2,638 €</td></tr>
      <tr><td>Production B3-128</td><td>2,5175 €</td><td>2,6379 €</td><td>5,2758 €</td></tr>
      <tr><td>Production B3-256</td><td>5,0349 €</td><td>5,2757 €</td><td>10,5514 €</td></tr>
      <tr><td>Business DB1-4</td><td>0,068 €</td><td>0,0746 €</td><td>0,1492 €</td></tr>
      <tr><td>Business DB1-7</td><td>0,151 €</td><td>0,1658 €</td><td>0,3316 €</td></tr>
      <tr><td>Business DB1-15</td><td>0,2252 €</td><td>0,2471 €</td><td>0,4942 €</td></tr>
      <tr><td>Business DB1-30</td><td>0,4448 €</td><td>0,4882 €</td><td>0,9764 €</td></tr>
      <tr><td>Business DB1-60</td><td>0,8895 €</td><td>0,9764 €</td><td>1,9528 €</td></tr>
      <tr><td>Business DB1-120</td><td>1,7736 €</td><td>1,9468 €</td><td>3,8936 €</td></tr>
    </tbody>
  </table>

  <h3>Kafka</h3>
  <table class="dataframe">
    <thead>
      <tr><th>Reference</th><th>Ancien prix public (HT / heure / node)</th><th>Nouveau prix public (HT / heure / node)</th><th>Nouveau prix public (HT / heure)</th></tr>
    </thead>
    <tbody>
      <tr><td>Production B3-8</td><td>0,2656 €</td><td>0,2782 €</td><td>0,8346 €</td></tr>
      <tr><td>Production B3-16</td><td>0,5311 €</td><td>0,5565 €</td><td>1,6695 €</td></tr>
      <tr><td>Production B3-32</td><td>1,0622 €</td><td>1,113 €</td><td>3,339 €</td></tr>
      <tr><td>Business DB1-4</td><td>0,1469 €</td><td>0,1612 €</td><td>0,4836 €</td></tr>
      <tr><td>Business DB1-7</td><td>0,2911 €</td><td>0,3195 €</td><td>0,9585 €</td></tr>
      <tr><td>Business DB1-15</td><td>0,5532 €</td><td>0,6073 €</td><td>1,8219 €</td></tr>
      <tr><td>Business DB1-30</td><td>1,0707 €</td><td>1,1753 €</td><td>3,5259 €</td></tr>
      <tr><td>Business DB1-60</td><td>2,1428 €</td><td>2,3522 €</td><td>7,0566 €</td></tr>
      <tr><td>Advanced B3-8</td><td>0,2656 €</td><td>0,2782 €</td><td>1,6692 €</td></tr>
      <tr><td>Advanced B3-16</td><td>0,5311 €</td><td>0,5565 €</td><td>3,339 €</td></tr>
      <tr><td>Advanced B3-32</td><td>1,0622 €</td><td>1,113 €</td><td>6,678 €</td></tr>
      <tr><td>Enterprise DB1-7</td><td>0,2924 €</td><td>0,321 €</td><td>1,926 €</td></tr>
      <tr><td>Enterprise DB1-15</td><td>0,5532 €</td><td>0,6073 €</td><td>3,6438 €</td></tr>
      <tr><td>Enterprise DB1-30</td><td>1,0707 €</td><td>1,1753 €</td><td>7,0518 €</td></tr>
      <tr><td>Enterprise DB1-60</td><td>2,1428 €</td><td>2,3522 €</td><td>14,1132 €</td></tr>
    </tbody>
  </table>

  <h3>Kafka Connect</h3>
  <table class="dataframe">
    <thead>
      <tr><th>Reference</th><th>Ancien prix public (HT / heure / node)</th><th>Nouveau prix public (HT / heure / node)</th><th>Nouveau prix public (HT / heure)</th></tr>
    </thead>
    <tbody>
      <tr><td>Essential DB1-4</td><td>0,1044 €</td><td>0,1145 €</td><td>0,1145 €</td></tr>
      <tr><td>Essential DB1-7</td><td>0,2101 €</td><td>0,2305 €</td><td>0,2305 €</td></tr>
      <tr><td>Essential DB1-15</td><td>0,3913 €</td><td>0,4295 €</td><td>0,4295 €</td></tr>
      <tr><td>Essential DB1-30</td><td>0,7084 €</td><td>0,7775 €</td><td>0,7775 €</td></tr>
      <tr><td>Production B3-8</td><td>0,1917 €</td><td>0,2008 €</td><td>0,6024 €</td></tr>
      <tr><td>Production B3-16</td><td>0,3862 €</td><td>0,4046 €</td><td>1,2138 €</td></tr>
      <tr><td>Production B3-32</td><td>0,7027 €</td><td>0,7363 €</td><td>2,2089 €</td></tr>
      <tr><td>Business DB1-7</td><td>0,2101 €</td><td>0,2305 €</td><td>0,6915 €</td></tr>
      <tr><td>Business DB1-15</td><td>0,4022 €</td><td>0,4415 €</td><td>1,3245 €</td></tr>
      <tr><td>Business DB1-30</td><td>0,7084 €</td><td>0,7775 €</td><td>2,3325 €</td></tr>
      <tr><td>Advanced B3-8</td><td>0,1908 €</td><td>0,1999 €</td><td>1,1994 €</td></tr>
      <tr><td>Advanced B3-16</td><td>0,3862 €</td><td>0,4046 €</td><td>2,4276 €</td></tr>
      <tr><td>Advanced B3-32</td><td>0,7027 €</td><td>0,7363 €</td><td>4,4178 €</td></tr>
      <tr><td>Enterprise DB1-7</td><td>0,2101 €</td><td>0,2305 €</td><td>1,383 €</td></tr>
      <tr><td>Enterprise DB1-15</td><td>0,4022 €</td><td>0,4415 €</td><td>2,649 €</td></tr>
      <tr><td>Enterprise DB1-30</td><td>0,7084 €</td><td>0,7775 €</td><td>4,665 €</td></tr>
    </tbody>
  </table>

  <h3>Kafka Mirror Maker</h3>
  <table class="dataframe">
    <thead>
      <tr><th>Reference</th><th>Ancien prix public (HT / heure / node)</th><th>Nouveau prix public (HT / heure / node)</th><th>Nouveau prix public (HT / heure)</th></tr>
    </thead>
    <tbody>
      <tr><td>Essential DB1-4</td><td>0,1044 €</td><td>0,1145 €</td><td>0,1145 €</td></tr>
      <tr><td>Essential DB1-7</td><td>0,2101 €</td><td>0,2305 €</td><td>0,2305 €</td></tr>
      <tr><td>Essential DB1-15</td><td>0,3913 €</td><td>0,4295 €</td><td>0,4295 €</td></tr>
      <tr><td>Essential DB1-30</td><td>0,7084 €</td><td>0,7775 €</td><td>0,7775 €</td></tr>
      <tr><td>Production B3-8</td><td>0,1917 €</td><td>0,2008 €</td><td>0,6024 €</td></tr>
      <tr><td>Production B3-16</td><td>0,3862 €</td><td>0,4046 €</td><td>1,2138 €</td></tr>
      <tr><td>Production B3-32</td><td>0,7027 €</td><td>0,7363 €</td><td>2,2089 €</td></tr>
      <tr><td>Business DB1-4</td><td>0,1057 €</td><td>0,116 €</td><td>0,348 €</td></tr>
      <tr><td>Business DB1-7</td><td>0,2101 €</td><td>0,2305 €</td><td>0,6915 €</td></tr>
      <tr><td>Business DB1-15</td><td>0,4022 €</td><td>0,4415 €</td><td>1,3245 €</td></tr>
      <tr><td>Business DB1-30</td><td>0,7084 €</td><td>0,7775 €</td><td>2,3325 €</td></tr>
      <tr><td>Advanced B3-8</td><td>0,1908 €</td><td>0,1999 €</td><td>1,1994 €</td></tr>
      <tr><td>Advanced B3-16</td><td>0,3862 €</td><td>0,4046 €</td><td>2,4276 €</td></tr>
      <tr><td>Advanced B3-32</td><td>0,7027 €</td><td>0,7363 €</td><td>4,4178 €</td></tr>
      <tr><td>Enterprise DB1-7</td><td>0,2101 €</td><td>0,2305 €</td><td>1,383 €</td></tr>
      <tr><td>Enterprise DB1-15</td><td>0,4022 €</td><td>0,4415 €</td><td>2,649 €</td></tr>
      <tr><td>Enterprise DB1-30</td><td>0,7084 €</td><td>0,7775 €</td><td>4,665 €</td></tr>
    </tbody>
  </table>

  <h3>Opensearch</h3>
  <table class="dataframe">
    <thead>
      <tr><th>Reference</th><th>Ancien prix public (HT / heure / node)</th><th>Nouveau prix public (HT / heure / node)</th><th>Nouveau prix public (HT / heure)</th></tr>
    </thead>
    <tbody>
      <tr><td>Essential DB1-4</td><td>0,0742 €</td><td>0,0814 €</td><td>0,0814 €</td></tr>
      <tr><td>Essential DB1-7</td><td>0,1497 €</td><td>0,1642 €</td><td>0,1642 €</td></tr>
      <tr><td>Essential DB1-15</td><td>0,3007 €</td><td>0,33 €</td><td>0,33 €</td></tr>
      <tr><td>Production B3-8</td><td>0,172 €</td><td>0,1801 €</td><td>0,5403 €</td></tr>
      <tr><td>Production B3-16</td><td>0,3439 €</td><td>0,3603 €</td><td>1,0809 €</td></tr>
      <tr><td>Production B3-32</td><td>0,6877 €</td><td>0,7205 €</td><td>2,1615 €</td></tr>
      <tr><td>Production B3-64</td><td>1,3754 €</td><td>1,4411 €</td><td>4,3233 €</td></tr>
      <tr><td>Business DB1-7</td><td>0,1607 €</td><td>0,1763 €</td><td>0,5289 €</td></tr>
      <tr><td>Business DB1-15</td><td>0,3213 €</td><td>0,3526 €</td><td>1,0578 €</td></tr>
      <tr><td>Business DB1-30</td><td>0,648 €</td><td>0,7112 €</td><td>2,1336 €</td></tr>
      <tr><td>Business DB1-60</td><td>1,2972 €</td><td>1,424 €</td><td>4,272 €</td></tr>
      <tr><td>Business DB1-120</td><td>2,6013 €</td><td>2,8555 €</td><td>8,5665 €</td></tr>
      <tr><td>Advanced B3-8</td><td>0,1839 €</td><td>0,1927 €</td><td>1,1562 €</td></tr>
      <tr><td>Advanced B3-16</td><td>0,3678 €</td><td>0,3854 €</td><td>2,3124 €</td></tr>
      <tr><td>Advanced B3-32</td><td>0,7357 €</td><td>0,7708 €</td><td>4,6248 €</td></tr>
      <tr><td>Advanced B3-64</td><td>1,4713 €</td><td>1,5416 €</td><td>9,2496 €</td></tr>
      <tr><td>Enterprise DB1-7</td><td>0,162 €</td><td>0,1778 €</td><td>1,0668 €</td></tr>
      <tr><td>Enterprise DB1-15</td><td>0,3254 €</td><td>0,3571 €</td><td>2,1426 €</td></tr>
      <tr><td>Enterprise DB1-30</td><td>0,6521 €</td><td>0,7158 €</td><td>4,2948 €</td></tr>
      <tr><td>Enterprise DB1-60</td><td>1,3014 €</td><td>1,4285 €</td><td>8,571 €</td></tr>
      <tr><td>Enterprise DB1-120</td><td>2,6027 €</td><td>2,857 €</td><td>17,142 €</td></tr>
    </tbody>
  </table>

  <h3>Managed Dashboard</h3>
  <table class="dataframe">
    <thead>
      <tr><th>Reference</th><th>Ancien prix public (HT / heure / node)</th><th>Nouveau prix public (HT / heure / node)</th><th>Nouveau prix public (HT / heure)</th></tr>
    </thead>
    <tbody>
      <tr><td>Essential DB1-4</td><td>0,0591 €</td><td>0,0648 €</td><td>0,0648 €</td></tr>
      <tr><td>Essential DB1-7</td><td>0,1195 €</td><td>0,1311 €</td><td>0,1311 €</td></tr>
      <tr><td>Production B3-8</td><td>0,1195 €</td><td>0,1251 €</td><td>0,1251 €</td></tr>
    </tbody>
  </table>
</div>

<div class='table-container'>
  <h2>Floating IPs</h2>
  <table class="dataframe">
    <thead>
      <tr style="text-align: right;">
        <th>Reference</th>
        <th>Ancien prix public (HT / heure)</th>
        <th>Nouveau prix public (HT / heure)</th>
      </tr>
    </thead>
    <tbody>
      <tr><td>Floating Ips</td><td>0.0025 €</td><td>0.0027 €</td></tr>
    </tbody>
  </table>
</div>

<!DOCTYPE html>
<html lang="fr">
<head>
  <meta charset="UTF-8">
  <title>Bare Metal – Tableaux des prix</title>
  <style>
    body { font-family: Arial, Helvetica, sans-serif; margin: 40px; background-color: #f4f6f8; }
    h2 { margin-top: 50px; color: #1a2b3c; }
    .table-container { margin-top: 20px; margin-bottom: 40px; }
    table { border-collapse: collapse; width: 100%; background: white; border-radius: 8px; overflow: hidden; box-shadow: 0 2px 8px rgba(0,0,0,0.08); }
    thead { background-color: #000E9C; color: white; }
    th { padding: 12px 15px; text-align: center; font-weight: bold; text-transform: uppercase; font-size: 13px; }
    td { padding: 10px 15px; font-size: 14px; }
    td:nth-child(n+2) { text-align: center; }
    tbody tr { border-bottom: 1px solid #e6e9ed; }
    tbody tr:nth-child(even) { background-color: #f8fafc; }
    tbody tr:hover { background-color: #eef3ff; }
    tr.family-header td {
      background-color: #000E9C;
      color: white;
      font-weight: bold;
      font-size: 14px;
      text-align: left !important;
      padding: 10px 15px;
      letter-spacing: 0.05em;
      text-transform: uppercase;
    }
    tr.server-header td {
      background-color: #1a3a6b;
      color: #e8efff;
      font-weight: bold;
      font-size: 13px;
      text-align: center;
      padding: 8px 15px;
      font-style: italic;
    }
    tr.server-header td:first-child {
      text-align: left !important;
    }
    tr.option-subheader td {
      background-color: #dce6f5;
      color: #1a2b3c;
      font-weight: bold;
      font-size: 12px;
      text-align: left !important;
      padding: 6px 15px;
      text-transform: uppercase;
      letter-spacing: 0.08em;
    }
  </style>
</head>
<body>
<div class='table-container'>
  <h2>Serveurs Dédiés &#038; Options</h2>
<p class="mention">Ci-dessous, les tarifs mensuels standards des serveurs, sans prépaiement ni remise liée à un engagement. Les prix des options s’appliquent uniquement aux nouvelles commandes. Le renouvellement des options, communiqué à nos clients par e-mail, sera limité à +10 % pour les options de disque et +15 % pour les options de RAM.</p>
  <table>
    <thead>
      <tr>
        <th></th>
        <th>Ancien prix public (HT / mois)</th>
        <th>Nouveau prix public (HT / mois)</th>
      </tr>
    </thead>
    <tbody>
      <tr class="family-header"><td colspan="3">ADVANCE</td></tr>
      <tr class="server-header"><td>ADVANCE-1 &#8211; 2024 &#8211; AMD EPYC 4244P</td><td>84.99 €</td><td>89.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>32GB DDR5 On-Die ECC 5200MHz</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>64GB DDR5 On-Die ECC 5200MHz</td><td>12 €</td><td>18 €</td></tr>
      <tr><td>128GB DDR5 On-Die ECC 3600MHz</td><td>36 €</td><td>58 €</td></tr>
      <tr><td>192GB DDR5 On-Die ECC 3600MHz</td><td>60 €</td><td>78 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Stockage</td></tr>
      <tr><td>2x SSD NVMe 960GB Enterprise Class Soft RAID</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>4x SSD NVMe 960GB Enterprise Class Soft RAID</td><td>26 €</td><td>42 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>52 €</td><td>70 €</td></tr>
      <tr><td>4x SSD NVMe 1.92TB Enterprise Class Soft RAID</td><td>78 €</td><td>98 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>104 €</td><td>118 €</td></tr>
      <tr><td>4x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>182 €</td><td>200 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 7.68TB Datacenter Class Soft RAID</td><td>208 €</td><td>229 €</td></tr>
      <tr><td>4x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>390 €</td><td>429 €</td></tr>
      <tr class="server-header"><td>ADVANCE-1 &#8211; 2026 &#8211; AMD EPYC 4245P</td><td>99.99 €</td><td>104.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>32GB DDR5 On-Die ECC 5600MHz</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>64GB DDR5 On-Die ECC 5600MHz</td><td>22 €</td><td>26 €</td></tr>
      <tr><td>128GB DDR5 On-Die ECC 3600MHz</td><td>44 €</td><td>58 €</td></tr>
      <tr><td>256GB DDR5 On-Die ECC 3600MHz</td><td>63 €</td><td>130 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Stockage</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>4x SSD NVMe 960GB Datacenter Class Soft RAID</td><td>21.60 €</td><td>42 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>38 €</td><td>70 €</td></tr>
      <tr><td>4x SSD NVMe 1.92TB Enterprise Class PCIe 5.0 Soft RAID</td><td>54.40 €</td><td>98 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>100 €</td><td>118 €</td></tr>
      <tr><td>4x SSD NVMe 3.84TB Enterprise Class PCIe 5.0 Soft RAID</td><td>178.40 €</td><td>197 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 7.68TB Datacenter Class Soft RAID</td><td>149.20 €</td><td>210 €</td></tr>
      <tr><td>4x SSD NVMe 7.68TB Enterprise Class PCIe 5.0 Soft RAID</td><td>298.40 €</td><td>378 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 15.36TB Datacenter Class Soft RAID</td><td>299.99 €</td><td>392 €</td></tr>
      <tr class="server-header"><td>ADVANCE-2 &#8211; 2024 &#8211; AMD EPYC 4344P</td><td>119.99 €</td><td>124.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>64GB DDR5 On-Die ECC 5200MHz</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>128GB DDR5 On-Die ECC 3600MHz</td><td>24 €</td><td>40 €</td></tr>
      <tr><td>192GB DDR5 On-Die ECC 3600MHz</td><td>48 €</td><td>60 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Stockage</td></tr>
      <tr><td>2x SSD NVMe 960GB Enterprise Class Soft RAID</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>4x SSD NVMe 960GB Enterprise Class Soft RAID</td><td>26 €</td><td>42 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>52 €</td><td>70 €</td></tr>
      <tr><td>4x SSD NVMe 1.92TB Enterprise Class Soft RAID</td><td>78 €</td><td>98 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>104 €</td><td>118 €</td></tr>
      <tr><td>4x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>182 €</td><td>200 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 7.68TB Datacenter Class Soft RAID</td><td>208 €</td><td>229 €</td></tr>
      <tr><td>4x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>390 €</td><td>429 €</td></tr>
      <tr class="server-header"><td>ADVANCE-2 &#8211; 2026 &#8211; AMD EPYC 4345P</td><td>119.99 €</td><td>134.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>64GB DDR5 On-Die ECC 5600MHz</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>128GB DDR5 On-Die ECC 3600MHz</td><td>22 €</td><td>40 €</td></tr>
      <tr><td>256GB DDR5 On-Die ECC 3600MHz</td><td>52 €</td><td>112 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Stockage</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>4x SSD NVMe 960GB Datacenter Class Soft RAID</td><td>21.60 €</td><td>42 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>38 €</td><td>70 €</td></tr>
      <tr><td>4x SSD NVMe 1.92TB Enterprise Class PCIe 5.0 Soft RAID</td><td>54.40 €</td><td>98 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>100 €</td><td>118 €</td></tr>
      <tr><td>4x SSD NVMe 3.84TB Enterprise Class PCIe 5.0 Soft RAID</td><td>178.40 €</td><td>197 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 7.68TB Datacenter Class Soft RAID</td><td>149.20 €</td><td>210 €</td></tr>
      <tr><td>4x SSD NVMe 7.68TB Enterprise Class PCIe 5.0 Soft RAID</td><td>298.40 €</td><td>378 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 15.36TB Datacenter Class Soft RAID</td><td>299.99 €</td><td>392 €</td></tr>
      <tr class="server-header"><td>ADVANCE-3 &#8211; 2024 &#8211; AMD EPYC 4464P</td><td>149.99 €</td><td>169.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>64GB DDR5 On-Die ECC 5200MHz</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>128GB DDR5 On-Die ECC 3600MHz</td><td>24 €</td><td>40 €</td></tr>
      <tr><td>192GB DDR5 On-Die ECC 3600MHz</td><td>48 €</td><td>60 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Stockage</td></tr>
      <tr><td>2x SSD NVMe 960GB Enterprise Class Soft RAID</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>4x SSD NVMe 960GB Enterprise Class Soft RAID</td><td>26 €</td><td>42 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>52 €</td><td>70 €</td></tr>
      <tr><td>4x SSD NVMe 1.92TB Enterprise Class Soft RAID</td><td>78 €</td><td>98 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>104 €</td><td>118 €</td></tr>
      <tr><td>4x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>182 €</td><td>200 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 7.68TB Datacenter Class Soft RAID</td><td>208 €</td><td>229 €</td></tr>
      <tr><td>4x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>390 €</td><td>429 €</td></tr>
      <tr class="server-header"><td>ADVANCE-3 &#8211; 2026 &#8211; AMD EPYC 4464P</td><td>159.99 €</td><td>199.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>64GB DDR5 On-Die ECC 5600MHz</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>128GB DDR5 On-Die ECC 3600MHz</td><td>22 €</td><td>40 €</td></tr>
      <tr><td>256GB DDR5 On-Die ECC 3600MHz</td><td>52 €</td><td>112 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Stockage</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>4x SSD NVMe 960GB Datacenter Class Soft RAID</td><td>21.60 €</td><td>42 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>38 €</td><td>70 €</td></tr>
      <tr><td>4x SSD NVMe 1.92TB Enterprise Class PCIe 5.0 Soft RAID</td><td>54.40 €</td><td>98 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>100 €</td><td>118 €</td></tr>
      <tr><td>4x SSD NVMe 3.84TB Enterprise Class PCIe 5.0 Soft RAID</td><td>178.40 €</td><td>197 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 7.68TB Datacenter Class Soft RAID</td><td>149.20 €</td><td>210 €</td></tr>
      <tr><td>4x SSD NVMe 7.68TB Enterprise Class PCIe 5.0 Soft RAID</td><td>298.40 €</td><td>378 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 15.36TB Datacenter Class Soft RAID</td><td>299.99 €</td><td>392 €</td></tr>
      <tr class="server-header"><td>ADVANCE-4 &#8211; 2024 &#8211; AMD EPYC 4584PX</td><td>199.99 €</td><td>219.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>64GB DDR5 On-Die ECC 5200MHz</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>128GB DDR5 On-Die ECC 3600MHz</td><td>24 €</td><td>40 €</td></tr>
      <tr><td>192GB DDR5 On-Die ECC 3600MHz</td><td>48 €</td><td>60 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Stockage</td></tr>
      <tr><td>2x SSD NVMe 960GB Enterprise Class Soft RAID</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>4x SSD NVMe 960GB Enterprise Class Soft RAID</td><td>26 €</td><td>42 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>52 €</td><td>70 €</td></tr>
      <tr><td>4x SSD NVMe 1.92TB Enterprise Class Soft RAID</td><td>78 €</td><td>98 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>104 €</td><td>118 €</td></tr>
      <tr><td>4x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>182 €</td><td>200 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 7.68TB Datacenter Class Soft RAID</td><td>208 €</td><td>229 €</td></tr>
      <tr><td>4x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>390 €</td><td>429 €</td></tr>
      <tr class="server-header"><td>ADVANCE-4 &#8211; 2026 &#8211; AMD EPYC 4585PX</td><td>199.99 €</td><td>239.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>64GB DDR5 On-Die ECC 5600MHz</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>128GB DDR5 On-Die ECC 3600MHz</td><td>22 €</td><td>40 €</td></tr>
      <tr><td>256GB DDR5 On-Die ECC 3600MHz</td><td>52 €</td><td>112 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Stockage</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>4x SSD NVMe 960GB Datacenter Class Soft RAID</td><td>21.60 €</td><td>42 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>38 €</td><td>70 €</td></tr>
      <tr><td>4x SSD NVMe 1.92TB Enterprise Class PCIe 5.0 Soft RAID</td><td>54.40 €</td><td>98 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>100 €</td><td>118 €</td></tr>
      <tr><td>4x SSD NVMe 3.84TB Enterprise Class PCIe 5.0 Soft RAID</td><td>178.40 €</td><td>197 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 7.68TB Datacenter Class Soft RAID</td><td>149.20 €</td><td>210 €</td></tr>
      <tr><td>4x SSD NVMe 7.68TB Enterprise Class PCIe 5.0 Soft RAID</td><td>298.40 €</td><td>378 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 15.36TB Datacenter Class Soft RAID</td><td>299.99 €</td><td>392 €</td></tr>
      <tr class="server-header"><td>ADVANCE-5 &#8211; 2024 &#8211; AMD EPYC 8224P</td><td>249.99 €</td><td>289.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>96GB DDR5 ECC 4800MHz</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>192GB DDR5 ECC 4800MHz</td><td>36 €</td><td>90 €</td></tr>
      <tr><td>384GB DDR5 ECC 4800MHz</td><td>108 €</td><td>318 €</td></tr>
      <tr><td>576GB DDR5 ECC 4800MHz</td><td>180 €</td><td>552 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Stockage</td></tr>
      <tr><td>2x SSD NVMe 960GB Enterprise Class Soft RAID</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>4x SSD NVMe 960GB Enterprise Class Soft RAID</td><td>26 €</td><td>42 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>52 €</td><td>70 €</td></tr>
      <tr><td>4x SSD NVMe 1.92TB Enterprise Class Soft RAID</td><td>78 €</td><td>98 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>104 €</td><td>118 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 4x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>104 €</td><td>140 €</td></tr>
      <tr><td>4x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>182 €</td><td>200 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 2x SSD NVMe 7.68TB Datacenter Class Soft RAID</td><td>208 €</td><td>229 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 4x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>208 €</td><td>236 €</td></tr>
      <tr><td>4x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>390 €</td><td>429 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 4x SSD NVMe 7.68TB Datacenter Class Soft RAID</td><td>416 €</td><td>458 €</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID + 6x SSD NVMe 7.68TB Datacenter Class Soft RAID</td><td>624 €</td><td>687 €</td></tr>
      <tr><td>8x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>806 €</td><td>887 €</td></tr>
      <tr class="server-header"><td>ADVANCE-STOR &#8211; 2024 &#8211; AMD EPYC 4344P</td><td>199.99 €</td><td>199.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>32GB DDR5 On-Die ECC 5200MHz</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>64GB DDR5 On-Die ECC 5200MHz</td><td>12 €</td><td>14 €</td></tr>
      <tr><td>128GB DDR5 On-Die ECC 3600MHz</td><td>36 €</td><td>42 €</td></tr>
      <tr><td>192GB DDR5 On-Die ECC 3600MHz</td><td>60 €</td><td>69 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Stockage</td></tr>
      <tr><td>2x HDD SAS 22TB Enterprise Class Soft RAID</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>2x SSD NVMe 960GB Enterprise Class Soft RAID</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>4x HDD SAS 22TB Enterprise Class Soft RAID</td><td>64 €</td><td>70 €</td></tr>
      <tr><td>2x HDD SAS 22TB Enterprise Class Hard RAID</td><td>66 €</td><td>73 €</td></tr>
      <tr><td>6x HDD SAS 22TB Enterprise Class Soft RAID</td><td>128 €</td><td>141 €</td></tr>
      <tr><td>4x HDD SAS 22TB Enterprise Class Hard RAID</td><td>130 €</td><td>143 €</td></tr>
      <tr><td>8x HDD SAS 22TB Enterprise Class Soft RAID</td><td>192 €</td><td>211 €</td></tr>
      <tr><td>6x HDD SAS 22TB Enterprise Class Hard RAID</td><td>194 €</td><td>213 €</td></tr>
      <tr><td>8x HDD SAS 22TB Enterprise Class Hard RAID</td><td>258 €</td><td>284 €</td></tr>
      <tr class="server-header"><td>ADVANCE-STOR &#8211; 2026 &#8211; AMD EPYC 4345P</td><td>199.99 €</td><td>229.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>32GB DDR5 On-Die ECC 5600MHz</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>64GB DDR5 On-Die ECC 5600MHz</td><td>22 €</td><td>25 €</td></tr>
      <tr><td>128GB DDR5 On-Die ECC 3600MHz</td><td>44 €</td><td>58 €</td></tr>
      <tr><td>256GB DDR5 On-Die ECC 3600MHz</td><td>63 €</td><td>130 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Stockage</td></tr>
      <tr><td>2x HDD SAS 24TB Enterprise Class Soft RAID</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class PCIe 5.0 Soft RAID</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>2x HDD SAS 24TB Enterprise Class Hard RAID</td><td>66 €</td><td>73 €</td></tr>
      <tr><td>4x HDD SAS 24TB Enterprise Class Soft RAID</td><td>64 €</td><td>94 €</td></tr>
      <tr><td>4x HDD SAS 24TB Enterprise Class Hard RAID</td><td>130 €</td><td>143 €</td></tr>
      <tr><td>6x HDD SAS 24TB Enterprise Class Soft RAID</td><td>128 €</td><td>188 €</td></tr>
      <tr><td>6x HDD SAS 24TB Enterprise Class Hard RAID</td><td>194 €</td><td>248 €</td></tr>
      <tr><td>8x HDD SAS 24TB Enterprise Class Soft RAID</td><td>192 €</td><td>282 €</td></tr>
      <tr><td>8x HDD SAS 24TB Enterprise Class Hard RAID</td><td>258 €</td><td>362 €</td></tr>
      <tr class="family-header"><td colspan="3">RISE</td></tr>
      <tr class="server-header"><td>RISE-L &#8211; 2025 &#8211; AMD RYZEN 9 9950X</td><td>134.99 €</td><td>149.99 €</td></tr>
      <tr class="server-header"><td>RISE-M &#8211; 2025 &#8211; AMD RYZEN 9 9900X</td><td>94.99 €</td><td>99.99 €</td></tr>
      <tr class="server-header"><td>RISE-S &#8211; 2025 &#8211; AMD Ryzen 7 9700X</td><td>54.99 €</td><td>64.99 €</td></tr>
      <tr class="server-header"><td>RISE-XL &#8211; 2025 &#8211; AMD EPYC TURIN 9455</td><td>269.99 €</td><td>299.99 €</td></tr>
      <tr class="family-header"><td colspan="3">GAME</td></tr>
      <tr class="server-header"><td>GAME-1 &#8211; 2026 &#8211; AMD RYZEN 7 9800X3D</td><td>129.99 €</td><td>139.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>64GB DDR5 On-Die ECC 5600MHz</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>128GB DDR5 On-Die ECC 3600MHz</td><td>22 €</td><td>40 €</td></tr>
      <tr><td>256GB DDR5 On-Die ECC 3600MHz</td><td>63 €</td><td>112 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Stockage</td></tr>
      <tr><td>2x SSD NVMe 960GB Enterprise Class Soft RAID</td><td>inclus</td><td>inclus</td></tr>
      <tr class="server-header"><td>GAME-2 &#8211; 2026 &#8211; AMD RYZEN 9 9950X3D</td><td>169.99 €</td><td>179.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>64GB DDR5 On-Die ECC 5600MHz</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>128GB DDR5 On-Die ECC 3600MHz</td><td>22 €</td><td>40 €</td></tr>
      <tr><td>256GB DDR5 On-Die ECC 3600MHz</td><td>63 €</td><td>112 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Stockage</td></tr>
      <tr><td>2x SSD NVMe 960GB Enterprise Class Soft RAID</td><td>inclus</td><td>inclus</td></tr>
      <tr class="family-header"><td colspan="3">SCALE-a</td></tr>
      <tr class="server-header"><td>SCALE-a1 &#8211; 2024 &#8211; AMD EPYC GENOA 9124</td><td>349.99 €</td><td>369.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>128GB DDR5 ECC 4800MHz</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>256GB DDR5 ECC 4800MHz</td><td>80 €</td><td>200 €</td></tr>
      <tr><td>512GB DDR5 ECC 4800MHz</td><td>240 €</td><td>440 €</td></tr>
      <tr><td>1TB DDR5 ECC 4800MHz</td><td>560 €</td><td>1000 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Stockage</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>52 €</td><td>70 €</td></tr>
      <tr><td>2x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>104 €</td><td>118 €</td></tr>
      <tr><td>4x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>104 €</td><td>140 €</td></tr>
      <tr><td>6x SSD NVMe 1.92TB Enterprise Class Soft RAID</td><td>156 €</td><td>210 €</td></tr>
      <tr><td>2x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>208 €</td><td>229 €</td></tr>
      <tr><td>4x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>208 €</td><td>236 €</td></tr>
      <tr><td>6x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>312 €</td><td>354 €</td></tr>
      <tr><td>4x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>416 €</td><td>458 €</td></tr>
      <tr><td>6x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>624 €</td><td>687 €</td></tr>
      <tr class="server-header"><td>SCALE-a1 &#8211; 2026 &#8211; AMD EPYC 9135</td><td>389.99 €</td><td>409.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>128GB DDR5 ECC 4800MHz</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>192GB DDR5 ECC 4800MHz</td><td>40 €</td><td>100 €</td></tr>
      <tr><td>256GB DDR5 ECC 4800MHz</td><td>80 €</td><td>120 €</td></tr>
      <tr><td>384GB DDR5 ECC 4800MHz</td><td>160 €</td><td>280 €</td></tr>
      <tr><td>512GB DDR5 ECC 4800MHz</td><td>240 €</td><td>400 €</td></tr>
      <tr><td>768GB DDR5 ECC 4800MHz</td><td>400 €</td><td>700 €</td></tr>
      <tr><td>1TB DDR5 ECC 4800MHz</td><td>560 €</td><td>1368 €</td></tr>
      <tr><td>1.5TB DDR5 ECC 4800MHz</td><td>880 €</td><td>2152 €</td></tr>
      <tr><td>3TB DDR5 ECC 3600MHz</td><td>1840 €</td><td>4504 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Stockage</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class PCIe 5.0 Soft RAID</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>38 €</td><td>70 €</td></tr>
      <tr><td>2x SSD NVMe 3.84TB Enterprise Class PCIe 5.0 Soft RAID</td><td>100 €</td><td>118 €</td></tr>
      <tr><td>4x SSD NVMe 1.92TB Enterprise Class PCIe 5.0 Soft RAID</td><td>76 €</td><td>140 €</td></tr>
      <tr><td>2x SSD NVMe 7.68TB Enterprise Class PCIe 5.0 Soft RAID</td><td>160 €</td><td>210 €</td></tr>
      <tr><td>6x SSD NVMe 1.92TB Enterprise Class PCIe 5.0 Soft RAID</td><td>190 €</td><td>210 €</td></tr>
      <tr><td>4x SSD NVMe 3.84TB Enterprise Class PCIe 5.0 Soft RAID</td><td>200 €</td><td>236 €</td></tr>
      <tr><td>6x SSD NVMe 3.84TB Enterprise Class PCIe 5.0 Soft RAID</td><td>300 €</td><td>354 €</td></tr>
      <tr><td>4x SSD NVMe 7.68TB Enterprise Class PCIe 5.0 Soft RAID</td><td>320 €</td><td>420 €</td></tr>
      <tr><td>6x SSD NVMe 7.68TB Enterprise Class PCIe 5.0 Soft RAID</td><td>480 €</td><td>630 €</td></tr>
      <tr><td>6x SSD NVMe 15.36TB Enterprise Class Soft RAID</td><td>900 €</td><td>1176 €</td></tr>
      <tr class="server-header"><td>SCALE-a2 &#8211; 2024 &#8211; AMD EPYC GENOA 9254</td><td>379.99 €</td><td>389.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>128GB DDR5 ECC 4800MHz</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>256GB DDR5 ECC 4800MHz</td><td>80 €</td><td>200 €</td></tr>
      <tr><td>512GB DDR5 ECC 4800MHz</td><td>240 €</td><td>440 €</td></tr>
      <tr><td>1TB DDR5 ECC 4800MHz</td><td>560 €</td><td>1000 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Stockage</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>52 €</td><td>70 €</td></tr>
      <tr><td>2x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>104 €</td><td>118 €</td></tr>
      <tr><td>4x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>104 €</td><td>140 €</td></tr>
      <tr><td>6x SSD NVMe 1.92TB Enterprise Class Soft RAID</td><td>156 €</td><td>210 €</td></tr>
      <tr><td>2x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>208 €</td><td>229 €</td></tr>
      <tr><td>4x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>208 €</td><td>236 €</td></tr>
      <tr><td>6x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>312 €</td><td>354 €</td></tr>
      <tr><td>4x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>416 €</td><td>458 €</td></tr>
      <tr><td>6x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>624 €</td><td>687 €</td></tr>
      <tr class="server-header"><td>SCALE-a2 &#8211; 2026 &#8211; AMD EPYC 9255</td><td>429.99 €</td><td>439.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>128GB DDR5 ECC 4800MHz</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>192GB DDR5 ECC 4800MHz</td><td>40 €</td><td>100 €</td></tr>
      <tr><td>256GB DDR5 ECC 4800MHz</td><td>80 €</td><td>120 €</td></tr>
      <tr><td>384GB DDR5 ECC 4800MHz</td><td>160 €</td><td>280 €</td></tr>
      <tr><td>512GB DDR5 ECC 4800MHz</td><td>240 €</td><td>400 €</td></tr>
      <tr><td>768GB DDR5 ECC 4800MHz</td><td>400 €</td><td>700 €</td></tr>
      <tr><td>1TB DDR5 ECC 4800MHz</td><td>560 €</td><td>1368 €</td></tr>
      <tr><td>1.5TB DDR5 ECC 4800MHz</td><td>880 €</td><td>2152 €</td></tr>
      <tr><td>3TB DDR5 ECC 3600MHz</td><td>1840 €</td><td>4504 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Stockage</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class PCIe 5.0 Soft RAID</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>38 €</td><td>70 €</td></tr>
      <tr><td>2x SSD NVMe 3.84TB Enterprise Class PCIe 5.0 Soft RAID</td><td>100 €</td><td>118 €</td></tr>
      <tr><td>4x SSD NVMe 1.92TB Enterprise Class PCIe 5.0 Soft RAID</td><td>76 €</td><td>140 €</td></tr>
      <tr><td>2x SSD NVMe 7.68TB Enterprise Class PCIe 5.0 Soft RAID</td><td>160 €</td><td>210 €</td></tr>
      <tr><td>6x SSD NVMe 1.92TB Enterprise Class PCIe 5.0 Soft RAID</td><td>190 €</td><td>210 €</td></tr>
      <tr><td>4x SSD NVMe 3.84TB Enterprise Class PCIe 5.0 Soft RAID</td><td>200 €</td><td>236 €</td></tr>
      <tr><td>6x SSD NVMe 3.84TB Enterprise Class PCIe 5.0 Soft RAID</td><td>300 €</td><td>354 €</td></tr>
      <tr><td>4x SSD NVMe 7.68TB Enterprise Class PCIe 5.0 Soft RAID</td><td>320 €</td><td>420 €</td></tr>
      <tr><td>6x SSD NVMe 7.68TB Enterprise Class PCIe 5.0 Soft RAID</td><td>480 €</td><td>630 €</td></tr>
      <tr><td>6x SSD NVMe 15.36TB Enterprise Class Soft RAID</td><td>900 €</td><td>1176 €</td></tr>
      <tr class="server-header"><td>SCALE-a3  &#8211; 2024 &#8211; AMD EPYC GENOA 9354</td><td>419.99 €</td><td>449.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>128GB DDR5 ECC 4800MHz</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>256GB DDR5 ECC 4800MHz</td><td>80 €</td><td>200 €</td></tr>
      <tr><td>512GB DDR5 ECC 4800MHz</td><td>240 €</td><td>440 €</td></tr>
      <tr><td>1TB DDR5 ECC 4800MHz</td><td>560 €</td><td>1000 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Stockage</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>52 €</td><td>70 €</td></tr>
      <tr><td>2x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>104 €</td><td>118 €</td></tr>
      <tr><td>4x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>104 €</td><td>140 €</td></tr>
      <tr><td>6x SSD NVMe 1.92TB Enterprise Class Soft RAID</td><td>156 €</td><td>210 €</td></tr>
      <tr><td>2x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>208 €</td><td>229 €</td></tr>
      <tr><td>4x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>208 €</td><td>236 €</td></tr>
      <tr><td>6x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>312 €</td><td>354 €</td></tr>
      <tr><td>4x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>416 €</td><td>458 €</td></tr>
      <tr><td>6x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>624 €</td><td>687 €</td></tr>
      <tr class="server-header"><td>SCALE-a3 &#8211; 2026 &#8211; AMD EPYC 9355</td><td>469.99 €</td><td>499.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>128GB DDR5 ECC 4800MHz</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>192GB DDR5 ECC 4800MHz</td><td>40 €</td><td>100 €</td></tr>
      <tr><td>256GB DDR5 ECC 4800MHz</td><td>80 €</td><td>120 €</td></tr>
      <tr><td>384GB DDR5 ECC 4800MHz</td><td>160 €</td><td>280 €</td></tr>
      <tr><td>512GB DDR5 ECC 4800MHz</td><td>240 €</td><td>400 €</td></tr>
      <tr><td>768GB DDR5 ECC 4800MHz</td><td>400 €</td><td>700 €</td></tr>
      <tr><td>1TB DDR5 ECC 4800MHz</td><td>560 €</td><td>1368 €</td></tr>
      <tr><td>1.5TB DDR5 ECC 4800MHz</td><td>880 €</td><td>2152 €</td></tr>
      <tr><td>3TB DDR5 ECC 3600MHz</td><td>1840 €</td><td>4504 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Stockage</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class PCIe 5.0 Soft RAID</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>38 €</td><td>70 €</td></tr>
      <tr><td>2x SSD NVMe 3.84TB Enterprise Class PCIe 5.0 Soft RAID</td><td>100 €</td><td>118 €</td></tr>
      <tr><td>4x SSD NVMe 1.92TB Enterprise Class PCIe 5.0 Soft RAID</td><td>76 €</td><td>140 €</td></tr>
      <tr><td>2x SSD NVMe 7.68TB Enterprise Class PCIe 5.0 Soft RAID</td><td>160 €</td><td>210 €</td></tr>
      <tr><td>6x SSD NVMe 1.92TB Enterprise Class PCIe 5.0 Soft RAID</td><td>190 €</td><td>210 €</td></tr>
      <tr><td>4x SSD NVMe 3.84TB Enterprise Class PCIe 5.0 Soft RAID</td><td>200 €</td><td>236 €</td></tr>
      <tr><td>6x SSD NVMe 3.84TB Enterprise Class PCIe 5.0 Soft RAID</td><td>300 €</td><td>354 €</td></tr>
      <tr><td>4x SSD NVMe 7.68TB Enterprise Class PCIe 5.0 Soft RAID</td><td>320 €</td><td>420 €</td></tr>
      <tr><td>6x SSD NVMe 7.68TB Enterprise Class PCIe 5.0 Soft RAID</td><td>480 €</td><td>630 €</td></tr>
      <tr><td>6x SSD NVMe 15.36TB Enterprise Class Soft RAID</td><td>900 €</td><td>1176 €</td></tr>
      <tr class="server-header"><td>SCALE-a4 &#8211; 2024 &#8211; AMD EPYC GENOA 9454</td><td>449.99 €</td><td>459.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>128GB DDR5 ECC 4800MHz</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>256GB DDR5 ECC 4800MHz</td><td>80 €</td><td>200 €</td></tr>
      <tr><td>512GB DDR5 ECC 4800MHz</td><td>240 €</td><td>440 €</td></tr>
      <tr><td>1TB DDR5 ECC 4800MHz</td><td>560 €</td><td>1000 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Stockage</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>52 €</td><td>70 €</td></tr>
      <tr><td>2x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>104 €</td><td>118 €</td></tr>
      <tr><td>4x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>104 €</td><td>140 €</td></tr>
      <tr><td>6x SSD NVMe 1.92TB Enterprise Class Soft RAID</td><td>156 €</td><td>210 €</td></tr>
      <tr><td>2x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>208 €</td><td>229 €</td></tr>
      <tr><td>4x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>208 €</td><td>236 €</td></tr>
      <tr><td>6x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>312 €</td><td>354 €</td></tr>
      <tr><td>4x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>416 €</td><td>458 €</td></tr>
      <tr><td>6x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>624 €</td><td>687 €</td></tr>
      <tr class="server-header"><td>SCALE-a4 &#8211; 2026 &#8211;  AMD EPYC 9455</td><td>539.99 €</td><td>549.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>128GB DDR5 ECC 4800MHz</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>192GB DDR5 ECC 4800MHz</td><td>40 €</td><td>100 €</td></tr>
      <tr><td>256GB DDR5 ECC 4800MHz</td><td>80 €</td><td>120 €</td></tr>
      <tr><td>384GB DDR5 ECC 4800MHz</td><td>160 €</td><td>280 €</td></tr>
      <tr><td>512GB DDR5 ECC 4800MHz</td><td>240 €</td><td>400 €</td></tr>
      <tr><td>768GB DDR5 ECC 4800MHz</td><td>400 €</td><td>700 €</td></tr>
      <tr><td>1TB DDR5 ECC 4800MHz</td><td>560 €</td><td>1368 €</td></tr>
      <tr><td>1.5TB DDR5 ECC 4800MHz</td><td>880 €</td><td>2152 €</td></tr>
      <tr><td>3TB DDR5 ECC 3600MHz</td><td>1840 €</td><td>4504 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Stockage</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class PCIe 5.0 Soft RAID</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>38 €</td><td>70 €</td></tr>
      <tr><td>2x SSD NVMe 3.84TB Enterprise Class PCIe 5.0 Soft RAID</td><td>100 €</td><td>118 €</td></tr>
      <tr><td>4x SSD NVMe 1.92TB Enterprise Class PCIe 5.0 Soft RAID</td><td>76 €</td><td>140 €</td></tr>
      <tr><td>2x SSD NVMe 7.68TB Enterprise Class PCIe 5.0 Soft RAID</td><td>160 €</td><td>210 €</td></tr>
      <tr><td>6x SSD NVMe 1.92TB Enterprise Class PCIe 5.0 Soft RAID</td><td>190 €</td><td>210 €</td></tr>
      <tr><td>4x SSD NVMe 3.84TB Enterprise Class PCIe 5.0 Soft RAID</td><td>200 €</td><td>236 €</td></tr>
      <tr><td>6x SSD NVMe 3.84TB Enterprise Class PCIe 5.0 Soft RAID</td><td>300 €</td><td>354 €</td></tr>
      <tr><td>4x SSD NVMe 7.68TB Enterprise Class PCIe 5.0 Soft RAID</td><td>320 €</td><td>420 €</td></tr>
      <tr><td>6x SSD NVMe 7.68TB Enterprise Class PCIe 5.0 Soft RAID</td><td>480 €</td><td>630 €</td></tr>
      <tr><td>6x SSD NVMe 15.36TB Enterprise Class Soft RAID</td><td>900 €</td><td>1176 €</td></tr>
      <tr class="server-header"><td>SCALE-a5 &#8211; 2024 &#8211; AMD EPYC GENOA 9554</td><td>499.99 €</td><td>539.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>128GB DDR5 ECC 4800MHz</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>256GB DDR5 ECC 4800MHz</td><td>80 €</td><td>200 €</td></tr>
      <tr><td>512GB DDR5 ECC 4800MHz</td><td>240 €</td><td>440 €</td></tr>
      <tr><td>1TB DDR5 ECC 4800MHz</td><td>560 €</td><td>1000 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Stockage</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>52 €</td><td>70 €</td></tr>
      <tr><td>2x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>104 €</td><td>118 €</td></tr>
      <tr><td>4x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>104 €</td><td>140 €</td></tr>
      <tr><td>6x SSD NVMe 1.92TB Enterprise Class Soft RAID</td><td>156 €</td><td>210 €</td></tr>
      <tr><td>2x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>208 €</td><td>229 €</td></tr>
      <tr><td>4x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>208 €</td><td>236 €</td></tr>
      <tr><td>6x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>312 €</td><td>354 €</td></tr>
      <tr><td>4x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>416 €</td><td>458 €</td></tr>
      <tr><td>6x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>624 €</td><td>687 €</td></tr>
      <tr class="server-header"><td>SCALE-a5 &#8211; 2026 &#8211; AMD EPYC 9555</td><td>599.99 €</td><td>639.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>128GB DDR5 ECC 4800MHz</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>192GB DDR5 ECC 4800MHz</td><td>40 €</td><td>100 €</td></tr>
      <tr><td>256GB DDR5 ECC 4800MHz</td><td>80 €</td><td>120 €</td></tr>
      <tr><td>384GB DDR5 ECC 4800MHz</td><td>160 €</td><td>280 €</td></tr>
      <tr><td>512GB DDR5 ECC 4800MHz</td><td>240 €</td><td>400 €</td></tr>
      <tr><td>768GB DDR5 ECC 4800MHz</td><td>400 €</td><td>700 €</td></tr>
      <tr><td>1TB DDR5 ECC 4800MHz</td><td>560 €</td><td>1368 €</td></tr>
      <tr><td>1.5TB DDR5 ECC 4800MHz</td><td>880 €</td><td>2152 €</td></tr>
      <tr><td>3TB DDR5 ECC 3600MHz</td><td>1840 €</td><td>4504 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Stockage</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class PCIe 5.0 Soft RAID</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>38 €</td><td>70 €</td></tr>
      <tr><td>2x SSD NVMe 3.84TB Enterprise Class PCIe 5.0 Soft RAID</td><td>100 €</td><td>118 €</td></tr>
      <tr><td>4x SSD NVMe 1.92TB Enterprise Class PCIe 5.0 Soft RAID</td><td>76 €</td><td>140 €</td></tr>
      <tr><td>2x SSD NVMe 7.68TB Enterprise Class PCIe 5.0 Soft RAID</td><td>160 €</td><td>210 €</td></tr>
      <tr><td>6x SSD NVMe 1.92TB Enterprise Class PCIe 5.0 Soft RAID</td><td>190 €</td><td>210 €</td></tr>
      <tr><td>4x SSD NVMe 3.84TB Enterprise Class PCIe 5.0 Soft RAID</td><td>200 €</td><td>236 €</td></tr>
      <tr><td>6x SSD NVMe 3.84TB Enterprise Class PCIe 5.0 Soft RAID</td><td>300 €</td><td>354 €</td></tr>
      <tr><td>4x SSD NVMe 7.68TB Enterprise Class PCIe 5.0 Soft RAID</td><td>320 €</td><td>420 €</td></tr>
      <tr><td>6x SSD NVMe 7.68TB Enterprise Class PCIe 5.0 Soft RAID</td><td>480 €</td><td>630 €</td></tr>
      <tr><td>6x SSD NVMe 15.36TB Enterprise Class Soft RAID</td><td>900 €</td><td>1176 €</td></tr>
      <tr class="server-header"><td>SCALE-a6 &#8211; 2024 &#8211; AMD EPYC GENOA 9654</td><td>579.99 €</td><td>629.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>128GB DDR5 ECC 4800MHz</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>256GB DDR5 ECC 4800MHz</td><td>80 €</td><td>200 €</td></tr>
      <tr><td>512GB DDR5 ECC 4800MHz</td><td>240 €</td><td>440 €</td></tr>
      <tr><td>1TB DDR5 ECC 4800MHz</td><td>560 €</td><td>1000 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Stockage</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>52 €</td><td>70 €</td></tr>
      <tr><td>2x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>104 €</td><td>118 €</td></tr>
      <tr><td>4x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>104 €</td><td>140 €</td></tr>
      <tr><td>6x SSD NVMe 1.92TB Enterprise Class Soft RAID</td><td>156 €</td><td>210 €</td></tr>
      <tr><td>2x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>208 €</td><td>229 €</td></tr>
      <tr><td>4x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>208 €</td><td>236 €</td></tr>
      <tr><td>6x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>312 €</td><td>354 €</td></tr>
      <tr><td>4x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>416 €</td><td>458 €</td></tr>
      <tr><td>6x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>624 €</td><td>687 €</td></tr>
      <tr class="server-header"><td>SCALE-a6 &#8211; 2026 &#8211; AMD EPYC 9655</td><td>699.99 €</td><td>729.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>128GB DDR5 ECC 4800MHz</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>192GB DDR5 ECC 4800MHz</td><td>40 €</td><td>100 €</td></tr>
      <tr><td>256GB DDR5 ECC 4800MHz</td><td>80 €</td><td>120 €</td></tr>
      <tr><td>384GB DDR5 ECC 4800MHz</td><td>160 €</td><td>280 €</td></tr>
      <tr><td>512GB DDR5 ECC 4800MHz</td><td>240 €</td><td>400 €</td></tr>
      <tr><td>768GB DDR5 ECC 4800MHz</td><td>400 €</td><td>700 €</td></tr>
      <tr><td>1TB DDR5 ECC 4800MHz</td><td>560 €</td><td>1368 €</td></tr>
      <tr><td>1.5TB DDR5 ECC 4800MHz</td><td>880 €</td><td>2152 €</td></tr>
      <tr><td>3TB DDR5 ECC 3600MHz</td><td>1840 €</td><td>4504 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Stockage</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class PCIe 5.0 Soft RAID</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>38 €</td><td>70 €</td></tr>
      <tr><td>2x SSD NVMe 3.84TB Enterprise Class PCIe 5.0 Soft RAID</td><td>100 €</td><td>118 €</td></tr>
      <tr><td>4x SSD NVMe 1.92TB Enterprise Class PCIe 5.0 Soft RAID</td><td>76 €</td><td>140 €</td></tr>
      <tr><td>2x SSD NVMe 7.68TB Enterprise Class PCIe 5.0 Soft RAID</td><td>160 €</td><td>210 €</td></tr>
      <tr><td>6x SSD NVMe 1.92TB Enterprise Class PCIe 5.0 Soft RAID</td><td>190 €</td><td>210 €</td></tr>
      <tr><td>4x SSD NVMe 3.84TB Enterprise Class PCIe 5.0 Soft RAID</td><td>200 €</td><td>236 €</td></tr>
      <tr><td>6x SSD NVMe 3.84TB Enterprise Class PCIe 5.0 Soft RAID</td><td>300 €</td><td>354 €</td></tr>
      <tr><td>4x SSD NVMe 7.68TB Enterprise Class PCIe 5.0 Soft RAID</td><td>320 €</td><td>420 €</td></tr>
      <tr><td>6x SSD NVMe 7.68TB Enterprise Class PCIe 5.0 Soft RAID</td><td>480 €</td><td>630 €</td></tr>
      <tr><td>6x SSD NVMe 15.36TB Enterprise Class Soft RAID</td><td>900 €</td><td>1176 €</td></tr>
      <tr class="server-header"><td>SCALE-a7 &#8211; 2026 &#8211; AMD EPYC 9755</td><td>809.99 €</td><td>829.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>128GB DDR5 ECC 4800MHz</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>192GB DDR5 ECC 4800MHz</td><td>40 €</td><td>100 €</td></tr>
      <tr><td>256GB DDR5 ECC 4800MHz</td><td>80 €</td><td>120 €</td></tr>
      <tr><td>384GB DDR5 ECC 4800MHz</td><td>160 €</td><td>280 €</td></tr>
      <tr><td>512GB DDR5 ECC 4800MHz</td><td>240 €</td><td>400 €</td></tr>
      <tr><td>768GB DDR5 ECC 4800MHz</td><td>192 €</td><td>700 €</td></tr>
      <tr><td>1TB DDR5 ECC 4800MHz</td><td>560 €</td><td>1368 €</td></tr>
      <tr><td>1.5TB DDR5 ECC 4800MHz</td><td>880 €</td><td>2152 €</td></tr>
      <tr><td>3TB DDR5 ECC 3600MHz</td><td>1840 €</td><td>4504 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Stockage</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class PCIe 5.0 Soft RAID</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>38 €</td><td>70 €</td></tr>
      <tr><td>2x SSD NVMe 3.84TB Enterprise Class PCIe 5.0 Soft RAID</td><td>100 €</td><td>118 €</td></tr>
      <tr><td>4x SSD NVMe 1.92TB Enterprise Class PCIe 5.0 Soft RAID</td><td>76 €</td><td>140 €</td></tr>
      <tr><td>2x SSD NVMe 7.68TB Enterprise Class PCIe 5.0 Soft RAID</td><td>160 €</td><td>210 €</td></tr>
      <tr><td>6x SSD NVMe 1.92TB Enterprise Class PCIe 5.0 Soft RAID</td><td>190 €</td><td>210 €</td></tr>
      <tr><td>4x SSD NVMe 3.84TB Enterprise Class PCIe 5.0 Soft RAID</td><td>200 €</td><td>236 €</td></tr>
      <tr><td>6x SSD NVMe 3.84TB Enterprise Class PCIe 5.0 Soft RAID</td><td>300 €</td><td>354 €</td></tr>
      <tr><td>4x SSD NVMe 7.68TB Enterprise Class PCIe 5.0 Soft RAID</td><td>320 €</td><td>420 €</td></tr>
      <tr><td>6x SSD NVMe 7.68TB Enterprise Class PCIe 5.0 Soft RAID</td><td>480 €</td><td>630 €</td></tr>
      <tr><td>6x SSD NVMe 15.36TB Enterprise Class Soft RAID</td><td>900 €</td><td>1176 €</td></tr>
      <tr class="server-header"><td>SCALE-a8 &#8211; 2026 &#8211; AMD EPYC 9965</td><td>869.99 €</td><td>899.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>128GB DDR5 ECC 4800MHz</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>192GB DDR5 ECC 4800MHz</td><td>40 €</td><td>100 €</td></tr>
      <tr><td>256GB DDR5 ECC 4800MHz</td><td>80 €</td><td>120 €</td></tr>
      <tr><td>384GB DDR5 ECC 4800MHz</td><td>160 €</td><td>280 €</td></tr>
      <tr><td>512GB DDR5 ECC 4800MHz</td><td>240 €</td><td>400 €</td></tr>
      <tr><td>768GB DDR5 ECC 4800MHz</td><td>192 €</td><td>700 €</td></tr>
      <tr><td>1TB DDR5 ECC 4800MHz</td><td>560 €</td><td>1368 €</td></tr>
      <tr><td>1.5TB DDR5 ECC 4800MHz</td><td>880 €</td><td>2152 €</td></tr>
      <tr><td>3TB DDR5 ECC 3600MHz</td><td>1840 €</td><td>4504 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Stockage</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class PCIe 5.0 Soft RAID</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>38 €</td><td>70 €</td></tr>
      <tr><td>2x SSD NVMe 3.84TB Enterprise Class PCIe 5.0 Soft RAID</td><td>100 €</td><td>118 €</td></tr>
      <tr><td>4x SSD NVMe 1.92TB Enterprise Class PCIe 5.0 Soft RAID</td><td>76 €</td><td>140 €</td></tr>
      <tr><td>2x SSD NVMe 7.68TB Enterprise Class PCIe 5.0 Soft RAID</td><td>160 €</td><td>210 €</td></tr>
      <tr><td>6x SSD NVMe 1.92TB Enterprise Class PCIe 5.0 Soft RAID</td><td>190 €</td><td>210 €</td></tr>
      <tr><td>4x SSD NVMe 3.84TB Enterprise Class PCIe 5.0 Soft RAID</td><td>200 €</td><td>236 €</td></tr>
      <tr><td>6x SSD NVMe 3.84TB Enterprise Class PCIe 5.0 Soft RAID</td><td>300 €</td><td>354 €</td></tr>
      <tr><td>4x SSD NVMe 7.68TB Enterprise Class PCIe 5.0 Soft RAID</td><td>320 €</td><td>420 €</td></tr>
      <tr><td>6x SSD NVMe 7.68TB Enterprise Class PCIe 5.0 Soft RAID</td><td>480 €</td><td>630 €</td></tr>
      <tr><td>6x SSD NVMe 15.36TB Enterprise Class Soft RAID</td><td>900 €</td><td>1176 €</td></tr>
      <tr class="server-header"><td>SCALE-a9 &#8211; 2026 &#8211; Dual AMD EPYC 9965</td><td>1349.99 €</td><td>1349.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>128GB DDR5 ECC 5600MHz</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>192GB DDR5 ECC 5600MHz</td><td>40 €</td><td>40 €</td></tr>
      <tr><td>256GB DDR5 ECC 5600MHz</td><td>80 €</td><td>80 €</td></tr>
      <tr><td>384GB DDR5 ECC 5600MHz</td><td>160 €</td><td>160 €</td></tr>
      <tr><td>512GB DDR5 ECC 5600MHz</td><td>240 €</td><td>240 €</td></tr>
      <tr><td>768GB DDR5 ECC 5600MHz</td><td>inclus</td><td>400 €</td></tr>
      <tr><td>1024GB DDR5 ECC 5600MHz</td><td>560 €</td><td>560 €</td></tr>
      <tr><td>1.5TB DDR5 ECC 5600MHz</td><td>880 €</td><td>880 €</td></tr>
      <tr><td>3TB DDR5 ECC 5600MHz</td><td>1840 €</td><td>1840 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Stockage</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class PCIe 5.0 Soft RAID</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>38 €</td><td>38 €</td></tr>
      <tr><td>2x SSD NVMe 3.84TB Enterprise Class PCIe 5.0 Soft RAID</td><td>100 €</td><td>100 €</td></tr>
      <tr><td>2x SSD NVMe 7.68TB Enterprise Class PCIe 5.0 Soft RAID</td><td>160 €</td><td>160 €</td></tr>
      <tr class="family-header"><td colspan="3">SCALE-i</td></tr>
      <tr class="server-header"><td>SCALE-i1 &#8211; 2024 &#8211; Intel Xeon Gold 6426Y</td><td>349.99 €</td><td>369.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>128GB DDR5 ECC 4800MHz</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>256GB DDR5 ECC 4800MHz</td><td>80 €</td><td>200 €</td></tr>
      <tr><td>512GB DDR5 ECC 4800MHz</td><td>240 €</td><td>440 €</td></tr>
      <tr><td>1TB DDR5 ECC 4800MHz</td><td>560 €</td><td>1000 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Stockage</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>52 €</td><td>70 €</td></tr>
      <tr><td>2x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>104 €</td><td>118 €</td></tr>
      <tr><td>4x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>104 €</td><td>140 €</td></tr>
      <tr><td>6x SSD NVMe 1.92TB Enterprise Class Soft RAID</td><td>156 €</td><td>210 €</td></tr>
      <tr><td>2x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>208 €</td><td>229 €</td></tr>
      <tr><td>4x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>208 €</td><td>236 €</td></tr>
      <tr><td>6x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>312 €</td><td>354 €</td></tr>
      <tr><td>4x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>416 €</td><td>458 €</td></tr>
      <tr><td>6x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>624 €</td><td>687 €</td></tr>
      <tr class="server-header"><td>SCALE-i2 &#8211; 2024 &#8211; Intel Xeon Gold 6442Y</td><td>379.99 €</td><td>389.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>128GB DDR5 ECC 4800MHz</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>256GB DDR5 ECC 4800MHz</td><td>80 €</td><td>200 €</td></tr>
      <tr><td>512GB DDR5 ECC 4800MHz</td><td>240 €</td><td>440 €</td></tr>
      <tr><td>1TB DDR5 ECC 4800MHz</td><td>560 €</td><td>1000 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Stockage</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>52 €</td><td>70 €</td></tr>
      <tr><td>2x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>104 €</td><td>118 €</td></tr>
      <tr><td>4x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>104 €</td><td>140 €</td></tr>
      <tr><td>6x SSD NVMe 1.92TB Enterprise Class Soft RAID</td><td>156 €</td><td>210 €</td></tr>
      <tr><td>2x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>208 €</td><td>229 €</td></tr>
      <tr><td>4x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>208 €</td><td>236 €</td></tr>
      <tr><td>6x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>312 €</td><td>354 €</td></tr>
      <tr><td>4x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>416 €</td><td>458 €</td></tr>
      <tr><td>6x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>624 €</td><td>687 €</td></tr>
      <tr class="server-header"><td>SCALE-i3 &#8211; 2024 &#8211; Intel Xeon Gold 6438M</td><td>409.99 €</td><td>449.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>128GB DDR5 ECC 4800MHz</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>256GB DDR5 ECC 4800MHz</td><td>80 €</td><td>200 €</td></tr>
      <tr><td>512GB DDR5 ECC 4800MHz</td><td>240 €</td><td>440 €</td></tr>
      <tr><td>1TB DDR5 ECC 4800MHz</td><td>560 €</td><td>1000 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Stockage</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>52 €</td><td>70 €</td></tr>
      <tr><td>2x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>104 €</td><td>118 €</td></tr>
      <tr><td>4x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>104 €</td><td>140 €</td></tr>
      <tr><td>6x SSD NVMe 1.92TB Enterprise Class Soft RAID</td><td>156 €</td><td>210 €</td></tr>
      <tr><td>2x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>208 €</td><td>229 €</td></tr>
      <tr><td>4x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>208 €</td><td>236 €</td></tr>
      <tr><td>6x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>312 €</td><td>354 €</td></tr>
      <tr><td>4x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>416 €</td><td>458 €</td></tr>
      <tr><td>6x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>624 €</td><td>687 €</td></tr>
      <tr class="family-header"><td colspan="3">SCALE-GPU</td></tr>
      <tr class="server-header"><td>SCALE-GPU-1 &#8211; 2024 &#8211; AMD EPYC GENOA 9354</td><td>969.99 €</td><td>969.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>192GB DDR5 ECC 4800MHz</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>384GB DDR5 ECC 4800MHz</td><td>120 €</td><td>120 €</td></tr>
      <tr><td>768GB DDR5 ECC 4800MHz</td><td>240 €</td><td>240 €</td></tr>
      <tr><td>RAM 1,1TB DDR5 ECC 4800MHz</td><td>420 €</td><td>420 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Stockage</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>52 €</td><td>52 €</td></tr>
      <tr><td>2x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>104 €</td><td>104 €</td></tr>
      <tr><td>2x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>208 €</td><td>208 €</td></tr>
      <tr class="server-header"><td>SCALE-GPU-2 &#8211; 2024 &#8211; AMD EPYC GENOA 9454</td><td>999.99 €</td><td>999.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>192GB DDR5 ECC 4800MHz</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>384GB DDR5 ECC 4800MHz</td><td>120 €</td><td>120 €</td></tr>
      <tr><td>768GB DDR5 ECC 4800MHz</td><td>240 €</td><td>240 €</td></tr>
      <tr><td>RAM 1,1TB DDR5 ECC 4800MHz</td><td>420 €</td><td>420 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Stockage</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>52 €</td><td>52 €</td></tr>
      <tr><td>2x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>104 €</td><td>104 €</td></tr>
      <tr><td>2x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>208 €</td><td>208 €</td></tr>
      <tr class="server-header"><td>SCALE-GPU-3 &#8211; 2024 &#8211; AMD EPYC GENOA 9554</td><td>1029.99 €</td><td>1029.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>192GB DDR5 ECC 4800MHz</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>384GB DDR5 ECC 4800MHz</td><td>120 €</td><td>120 €</td></tr>
      <tr><td>768GB DDR5 ECC 4800MHz</td><td>240 €</td><td>240 €</td></tr>
      <tr><td>RAM 1,1TB DDR5 ECC 4800MHz</td><td>420 €</td><td>420 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Stockage</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>2x SSD NVMe 1.92TB Datacenter Class Soft RAID</td><td>52 €</td><td>52 €</td></tr>
      <tr><td>2x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>104 €</td><td>104 €</td></tr>
      <tr><td>2x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>208 €</td><td>208 €</td></tr>
      <tr class="family-header"><td colspan="3">HGR</td></tr>
      <tr class="server-header"><td>HGR-AI-2 &#8211; 2024 &#8211; DUAL AMD EPYC 9354</td><td>2969.99 €</td><td>2969.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>384GB DDR5 ECC 4800MHz</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>512GB DDR5 ECC 4800MHz</td><td>64 €</td><td>74 €</td></tr>
      <tr><td>768GB DDR5 ECC 4800MHz</td><td>400 €</td><td>360 €</td></tr>
      <tr><td>RAM 2304GB DDR5 ECC 4800MHz</td><td>960 €</td><td>2208 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Stockage</td></tr>
      <tr><td>2x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>4x SSD NVMe 3.84TB Datacenter Class Soft RAID</td><td>88 €</td><td>118 €</td></tr>
      <tr><td>2x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>150 €</td><td>165 €</td></tr>
      <tr><td>4x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>300 €</td><td>330 €</td></tr>
      <tr><td>2x SSD NVMe 15.36TB Enterprise Class Soft RAID</td><td>308 €</td><td>339 €</td></tr>
      <tr><td>4x SSD NVMe 15.36TB Enterprise Class Soft RAID</td><td>616 €</td><td>680 €</td></tr>
      <tr class="server-header"><td>HGR-HCI-a1 &#8211; 2024 &#8211; DUAL AMD EPYC 9254</td><td>999.99 €</td><td>1119.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>256GB DDR5 ECC 4800MHz</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>512GB DDR5 ECC 4800MHz</td><td>128 €</td><td>240 €</td></tr>
      <tr><td>1TB DDR5 ECC 4800MHz</td><td>384 €</td><td>800 €</td></tr>
      <tr><td>1.5TB DDR5 ECC 4800MHz</td><td>512 €</td><td>1472 €</td></tr>
      <tr><td>RAM 2304GB DDR5 ECC 4800MHz</td><td>1024 €</td><td>2408 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Stockage</td></tr>
      <tr><td>6x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>12x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>264 €</td><td>354 €</td></tr>
      <tr><td>18x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>528 €</td><td>708 €</td></tr>
      <tr><td>24x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>792 €</td><td>1062 €</td></tr>
      <tr class="server-header"><td>HGR-HCI-a2 &#8211; 2024 &#8211; DUAL AMD EPYC 9354</td><td>1139.99 €</td><td>1274.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>384GB DDR5 ECC 4800MHz</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>384GB DDR5 ECC 4800MHz</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>512GB DDR5 ECC 4800MHz</td><td>64 €</td><td>74 €</td></tr>
      <tr><td>768GB DDR5 ECC 4800MHz</td><td>400 €</td><td>360 €</td></tr>
      <tr><td>1TB DDR5 ECC 4800MHz</td><td>320 €</td><td>600 €</td></tr>
      <tr><td>1TB DDR5 ECC 4800MHz</td><td>320 €</td><td>600 €</td></tr>
      <tr><td>1.5TB DDR5 ECC 4800MHz</td><td>384 €</td><td>1272 €</td></tr>
      <tr><td>1.5TB DDR5 ECC 4800MHz</td><td>384 €</td><td>1272 €</td></tr>
      <tr><td>RAM 2304GB DDR5 ECC 4800MHz</td><td>960 €</td><td>2208 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Stockage</td></tr>
      <tr><td>6x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>12x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>264 €</td><td>354 €</td></tr>
      <tr><td>18x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>528 €</td><td>708 €</td></tr>
      <tr><td>24x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>792 €</td><td>1062 €</td></tr>
      <tr class="server-header"><td>HGR-HCI-i1 &#8211; 2024 &#8211; DUAL INTEL XEON GOLD 5515+</td><td>849.99 €</td><td>949.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>256GB DDR5 ECC 4800MHz</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>512GB DDR5 ECC 4800MHz</td><td>128 €</td><td>240 €</td></tr>
      <tr><td>1TB DDR5 ECC 4800MHz</td><td>384 €</td><td>800 €</td></tr>
      <tr><td>1.5TB DDR5 ECC 4800MHz</td><td>512 €</td><td>1472 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Stockage</td></tr>
      <tr><td>2x SSD NVMe 960GB Datacenter Class Soft RAID</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>6x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>12x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>264 €</td><td>354 €</td></tr>
      <tr><td>18x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>528 €</td><td>708 €</td></tr>
      <tr><td>24x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>792 €</td><td>1062 €</td></tr>
      <tr class="server-header"><td>HGR-HCI-i2 &#8211; 2024 &#8211; DUAL INTEL XEON GOLD 6526Y</td><td>929.99 €</td><td>1039.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>256GB DDR5 ECC 4800MHz</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>512GB DDR5 ECC 4800MHz</td><td>128 €</td><td>240 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Stockage</td></tr>
      <tr><td>6x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>12x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>264 €</td><td>354 €</td></tr>
      <tr><td>18x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>528 €</td><td>708 €</td></tr>
      <tr><td>24x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>792 €</td><td>1062 €</td></tr>
      <tr class="server-header"><td>HGR-HCI-i3 &#8211; 2024 &#8211; DUAL INTEL XEON GOLD 6542Y</td><td>999.99 €</td><td>1119.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>256GB DDR5 ECC 4800MHz</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>512GB DDR5 ECC 4800MHz</td><td>128 €</td><td>240 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Stockage</td></tr>
      <tr><td>6x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>12x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>264 €</td><td>354 €</td></tr>
      <tr><td>18x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>528 €</td><td>708 €</td></tr>
      <tr><td>24x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>792 €</td><td>1062 €</td></tr>
      <tr class="server-header"><td>HGR-HCI-i4 &#8211; 2024 &#8211; DUAL INTEL XEON GOLD 6554S</td><td>1079.99 €</td><td>1209.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>256GB DDR5 ECC 4800MHz</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>512GB DDR5 ECC 4800MHz</td><td>128 €</td><td>240 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Stockage</td></tr>
      <tr><td>6x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>12x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>264 €</td><td>354 €</td></tr>
      <tr><td>18x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>528 €</td><td>708 €</td></tr>
      <tr><td>24x SSD NVMe 3.84TB Enterprise Class Soft RAID</td><td>792 €</td><td>1062 €</td></tr>
      <tr class="server-header"><td>HGR-SAP-1 &#8211; 2024 &#8211; DUAL INTEL XEON GOLD 6226R</td><td>1011.99 €</td><td>1254.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>192GB DDR4 ECC 2933MHz</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>384GB DDR4 ECC 2933MHz</td><td>96 €</td><td>216 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Stockage</td></tr>
      <tr><td>6x SSD SAS 3.84TB Enterprise Class Hard RAID</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>2x SSD SATA 480GB</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>12x SSD SAS 3.84TB Enterprise Class Hard RAID</td><td>264 €</td><td>354 €</td></tr>
      <tr><td>24x SSD SAS 3.84TB Enterprise Class Hard RAID</td><td>792 €</td><td>1062 €</td></tr>
      <tr class="server-header"><td>HGR-SAP-2 &#8211; 2024 &#8211; DUAL INTEL XEON GOLD 6242R</td><td>1121.99 €</td><td>1391.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>384GB DDR4 ECC 2933MHz</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>RAM 768GB DDR4 ECC 2933MHz</td><td>192 €</td><td>312 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Stockage</td></tr>
      <tr><td>6x SSD SAS 3.84TB Enterprise Class Hard RAID</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>2x SSD SATA 480GB</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>12x SSD SAS 3.84TB Enterprise Class Hard RAID</td><td>264 €</td><td>354 €</td></tr>
      <tr><td>24x SSD SAS 3.84TB Enterprise Class Hard RAID</td><td>792 €</td><td>1062 €</td></tr>
      <tr class="server-header"><td>HGR-SAP-3 &#8211; 2024 &#8211; DUAL INTEL XEON GOLD 6248R</td><td>1231.99 €</td><td>1527.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>RAM 768GB DDR4 ECC 2933MHz</td><td>192 €</td><td>inclus</td></tr>
      <tr><td>RAM 1.5TB DDR4 ECC 2933MHz</td><td>384 €</td><td>1032 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Stockage</td></tr>
      <tr><td>6x SSD SAS 3.84TB Enterprise Class Hard RAID</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>2x SSD SATA 480GB</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>12x SSD SAS 3.84TB Enterprise Class Hard RAID</td><td>264 €</td><td>354 €</td></tr>
      <tr><td>24x SSD SAS 3.84TB Enterprise Class Hard RAID</td><td>792 €</td><td>1062 €</td></tr>
      <tr class="server-header"><td>HGR-SDS-1 &#8211; 2024 &#8211; DUAL INTEL XEON GOLD 5515+</td><td>999.99 €</td><td>1119.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Stockage</td></tr>
      <tr><td>6x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>6x SSD NVMe 15.36TB Enterprise Class Soft RAID</td><td>475 €</td><td>546 €</td></tr>
      <tr><td>12x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>450 €</td><td>630 €</td></tr>
      <tr><td>18x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>900 €</td><td>1260 €</td></tr>
      <tr><td>12x SSD NVMe 15.36TB Enterprise Class Soft RAID</td><td>1399 €</td><td>1722 €</td></tr>
      <tr><td>24x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>1350 €</td><td>1890 €</td></tr>
      <tr><td>18x SSD NVMe 15.36TB Enterprise Class Soft RAID</td><td>2323 €</td><td>2898 €</td></tr>
      <tr><td>24x SSD NVMe 15.36TB Enterprise Class Soft RAID</td><td>3247 €</td><td>4074 €</td></tr>
      <tr class="server-header"><td>HGR-SDS-2 &#8211; 2024 &#8211; DUAL INTEL XEON GOLD 6542Y</td><td>1149.99 €</td><td>1289.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Stockage</td></tr>
      <tr><td>6x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>6x SSD NVMe 15.36TB Enterprise Class Soft RAID</td><td>475 €</td><td>546 €</td></tr>
      <tr><td>12x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>450 €</td><td>630 €</td></tr>
      <tr><td>18x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>900 €</td><td>1260 €</td></tr>
      <tr><td>12x SSD NVMe 15.36TB Enterprise Class Soft RAID</td><td>1399 €</td><td>1722 €</td></tr>
      <tr><td>24x SSD NVMe 7.68TB Enterprise Class Soft RAID</td><td>1350 €</td><td>1890 €</td></tr>
      <tr><td>18x SSD NVMe 15.36TB Enterprise Class Soft RAID</td><td>2323 €</td><td>2898 €</td></tr>
      <tr><td>24x SSD NVMe 15.36TB Enterprise Class Soft RAID</td><td>3247 €</td><td>4074 €</td></tr>
      <tr class="server-header"><td>HGR-STOR-1 &#8211; 2024 &#8211; INTEL XEON GOLD 6554S</td><td>1199.99 €</td><td>1399.99 €</td></tr>
      <tr class="option-subheader"><td colspan="3">RAM</td></tr>
      <tr><td>128GB DDR5 ECC 4800MHz</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>256GB DDR5 ECC 4800MHz</td><td>64 €</td><td>200 €</td></tr>
      <tr><td>512GB DDR5 ECC 4800MHz</td><td>192 €</td><td>440 €</td></tr>
      <tr><td>768GB DDR5 ECC 4800MHz</td><td>400 €</td><td>760 €</td></tr>
      <tr><td>768GB DDR5 ECC 4800MHz</td><td>320 €</td><td>760 €</td></tr>
      <tr class="option-subheader"><td colspan="3">Stockage</td></tr>
      <tr><td>24x HDD SAS 22TB Enterprise Class Soft RAID</td><td>inclus</td><td>inclus</td></tr>
      <tr><td>24x HDD SAS 22TB + 2x SSD NVMe 3.84TB High perf. cache Enterprise Class Soft RAID</td><td>88 €</td><td>118 €</td></tr>
      <tr><td>24x HDD SAS 22TB Enterprise Class Hard RAID</td><td>66 €</td><td>120 €</td></tr>
      <tr><td>24x HDD SAS 22TB + 2x SSD NVMe 7.68TB High perf. cache Enterprise Class Soft RAID</td><td>150 €</td><td>210 €</td></tr>
      <tr><td>24x HDD SAS 22TB + 2x SSD NVMe 3.84TB High perf. cache Enterprise Class Hard RAID</td><td>154 €</td><td>238 €</td></tr>
      <tr><td>24x HDD SAS 22TB + 2x SSD NVMe 7.68TB High perf. cache Enterprise Class Hard RAID</td><td>216 €</td><td>330 €</td></tr>
      <tr><td>24x HDD SAS 22TB + 2x SSD NVMe 15.36TB High perf. cache Enterprise Class Soft RAID</td><td>308 €</td><td>392 €</td></tr>
      <tr><td>24x HDD SAS 22TB + 2x SSD NVMe 15.36TB High perf. cache Enterprise Class Hard RAID</td><td>374 €</td><td>512 €</td></tr>
      <tr><td>36x HDD SAS 22TB Enterprise Class Soft RAID</td><td>384 €</td><td>516 €</td></tr>
      <tr><td>36x HDD SAS 22TB + 2x SSD NVMe 3.84TB High perf. cache Enterprise Class Soft RAID</td><td>472 €</td><td>634 €</td></tr>
      <tr><td>36x HDD SAS 22TB Enterprise Class Hard RAID</td><td>450 €</td><td>696 €</td></tr>
      <tr><td>36x HDD SAS 22TB + 2x SSD NVMe 7.68TB High perf. cache Enterprise Class Soft RAID</td><td>534 €</td><td>726 €</td></tr>
      <tr><td>36x HDD SAS 22TB + 2x SSD NVMe 3.84TB High perf. cache Enterprise Class Hard RAID</td><td>538 €</td><td>814 €</td></tr>
      <tr><td>36x HDD SAS 22TB + 2x SSD NVMe 7.68TB High perf. cache Enterprise Class Hard RAID</td><td>600 €</td><td>906 €</td></tr>
      <tr><td>36x HDD SAS 22TB + 2x SSD NVMe 15.36TB High perf. cache Enterprise Class Soft RAID</td><td>692 €</td><td>908 €</td></tr>
      <tr><td>36x HDD SAS 22TB + 2x SSD NVMe 15.36TB High perf. cache Enterprise Class Hard RAID</td><td>758 €</td><td>1088 €</td></tr>
    </tbody>
  </table>
</div>
</body>
</html>


<div class='table-container'>
  <h2>IPs</h2>
  <table class="dataframe">
    <thead>
      <tr style="text-align: right;">
        <th>Reference</th>
        <th>Ancien prix public (HT / mois)</th>
        <th>Nouveau prix public (HT / mois)</th>
      </tr>
    </thead>
    <tbody>
      <tr><td>Additional IPv4</td><td>1.50 €</td><td>2.00 €</td></tr>
    </tbody>
  </table>
</div>



<div class='table-container'>
  <h2>VPS</h2>
  <table class="dataframe">
    <thead>
      <tr style="text-align: right;">
        <th>Family</th>
        <th>Reference</th>
        <th>Commit</th>
        <th>Ancien prix public (HT / mois)</th>
        <th>Nouveau prix public (HT / mois)</th>
      </tr>
    </thead>
    <tbody>
      <tr><td>VPS 2026</td><td>VPS-1</td><td>Monthly</td><td>4.49 €</td><td>6.49 €</td></tr>
      <tr><td>VPS 2026</td><td>VPS-2</td><td>Monthly</td><td>6.99 €</td><td>9.99 €</td></tr>
      <tr><td>VPS 2026</td><td>VPS-3</td><td>Monthly</td><td>13.99 €</td><td>19.99 €</td></tr>
      <tr><td>VPS 2026</td><td>VPS-4</td><td>Monthly</td><td>24.99 €</td><td>36.99 €</td></tr>
      <tr><td>VPS 2026</td><td>VPS-5</td><td>Monthly</td><td>36.99 €</td><td>54.99 €</td></tr>
      <tr><td>VPS 2026</td><td>VPS-6</td><td>Monthly</td><td>48.99 €</td><td>72.99 €</td></tr>
      <tr><td>VPS 2026</td><td>VPSLZ-1</td><td>Monthly</td><td>5.49 €</td><td>7.49 €</td></tr>
      <tr><td>VPS 2026</td><td>VPS-1</td><td>prepayment 6 months</td><td>25.56 €</td><td>36.99 €</td></tr>
      <tr><td>VPS 2026</td><td>VPS-2</td><td>prepayment 6 months</td><td>39.84 €</td><td>56.94 €</td></tr>
      <tr><td>VPS 2026</td><td>VPS-3</td><td>prepayment 6 months</td><td>79.74 €</td><td>113.94 €</td></tr>
      <tr><td>VPS 2026</td><td>VPS-4</td><td>prepayment 6 months</td><td>142.44 €</td><td>210.84 €</td></tr>
      <tr><td>VPS 2026</td><td>VPS-5</td><td>prepayment 6 months</td><td>210.84 €</td><td>313.44 €</td></tr>
      <tr><td>VPS 2026</td><td>VPS-6</td><td>prepayment 6 months</td><td>279.24 €</td><td>416.04 €</td></tr>
      <tr><td>VPS 2026</td><td>VPSLZ-1</td><td>prepayment 6 months</td><td>31.26 €</td><td>42.69 €</td></tr>
      <tr><td>VPS 2026</td><td>VPS-1</td><td>prepayment 12 months</td><td>45.72 €</td><td>66.19 €</td></tr>
      <tr><td>VPS 2026</td><td>VPS-2</td><td>prepayment 12 months</td><td>71.28 €</td><td>101.89 €</td></tr>
      <tr><td>VPS 2026</td><td>VPS-3</td><td>prepayment 12 months</td><td>142.68 €</td><td>203.89 €</td></tr>
      <tr><td>VPS 2026</td><td>VPS-4</td><td>prepayment 12 months</td><td>254.88 €</td><td>377.29 €</td></tr>
      <tr><td>VPS 2026</td><td>VPS-5</td><td>prepayment 12 months</td><td>377.28 €</td><td>560.89 €</td></tr>
      <tr><td>VPS 2026</td><td>VPS-6</td><td>prepayment 12 months</td><td>499.68 €</td><td>744.49 €</td></tr>
      <tr><td>VPS 2026</td><td>VPSLZ-1</td><td>prepayment 12 months</td><td>55.92 €</td><td>76.39 €</td></tr>
    </tbody>
  </table>
</div>

</body>
</html>
<img loading="lazy" decoding="async" src="//blog.ovhcloud.com/wp-content/plugins/matomo/app/matomo.php?idsite=1&amp;rec=1&amp;url=https%3A%2F%2Fblog.ovhcloud.com%2Fevolutions-tarifaires-de-public-cloud-bare-metal-et-vps-chez-ovhcloud%2F&amp;action_name=%C3%89volutions%20tarifaires%20de%20Public%20Cloud%2C%20Bare%20Metal%20et%20VPS%20chez%20OVHcloud&amp;urlref=https%3A%2F%2Fblog.ovhcloud.com%2Ffeed%2F" style="border:0;width:0;height:0" width="0" height="0" alt="" />]]></content:encoded>
					
		
		
			</item>
		<item>
		<title>Reference Architecture: Custom metric autoscaling for LLM inference with vLLM on OVHcloud AI Deploy and observability using MKS</title>
		<link>https://blog.ovhcloud.com/reference-architecture-custom-metric-autoscaling-for-llm-inference-with-vllm-on-ovhcloud-ai-deploy-and-observability-using-mks/</link>
		
		<dc:creator><![CDATA[Eléa Petton]]></dc:creator>
		<pubDate>Tue, 10 Feb 2026 08:51:11 +0000</pubDate>
				<category><![CDATA[OVHcloud Engineering]]></category>
		<category><![CDATA[AI]]></category>
		<category><![CDATA[AI Deploy]]></category>
		<category><![CDATA[Kubernetes]]></category>
		<category><![CDATA[LLM]]></category>
		<category><![CDATA[MKS]]></category>
		<category><![CDATA[Open Source]]></category>
		<category><![CDATA[OVHcloud]]></category>
		<category><![CDATA[prometheus]]></category>
		<category><![CDATA[Public Cloud]]></category>
		<guid isPermaLink="false">https://blog.ovhcloud.com/?p=30203</guid>

					<description><![CDATA[Take your LLM (Large Language Model) deployment to production level with comprehensive custom autoscaling configuration and advanced vLLM metrics observability. This reference architecture describes a comprehensive solution for deploying, autoscaling and monitoring vLLM-based LLM workloads on OVHcloud infrastructure. It combinesAI Deploy, used for model serving with custom metric autoscaling, and Managed Kubernetes Service (MKS), which [&#8230;]<img src="//blog.ovhcloud.com/wp-content/plugins/matomo/app/matomo.php?idsite=1&amp;rec=1&amp;url=https%3A%2F%2Fblog.ovhcloud.com%2Freference-architecture-custom-metric-autoscaling-for-llm-inference-with-vllm-on-ovhcloud-ai-deploy-and-observability-using-mks%2F&amp;action_name=Reference%20Architecture%3A%20Custom%20metric%20autoscaling%20for%20LLM%20inference%20with%20vLLM%20on%20OVHcloud%20AI%20Deploy%20and%20observability%20using%20MKS&amp;urlref=https%3A%2F%2Fblog.ovhcloud.com%2Ffeed%2F" style="border:0;width:0;height:0" width="0" height="0" alt="" />]]></description>
										<content:encoded><![CDATA[
<p><em><strong>Take your LLM (Large Language Model) deployment to production level with comprehensive custom autoscaling configuration and advanced vLLM metrics observability.</strong></em></p>



<figure class="wp-block-image aligncenter size-large"><img loading="lazy" decoding="async" width="1024" height="538" src="https://blog.ovhcloud.com/wp-content/uploads/2026/02/3-1024x538.jpg" alt="" class="wp-image-30579" srcset="https://blog.ovhcloud.com/wp-content/uploads/2026/02/3-1024x538.jpg 1024w, https://blog.ovhcloud.com/wp-content/uploads/2026/02/3-300x158.jpg 300w, https://blog.ovhcloud.com/wp-content/uploads/2026/02/3-768x403.jpg 768w, https://blog.ovhcloud.com/wp-content/uploads/2026/02/3.jpg 1200w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /><figcaption class="wp-element-caption"><em>vLLM metrics monitoring and observability based on OVHcloud infrastructure</em></figcaption></figure>



<p>This reference architecture describes a comprehensive solution for <strong>deploying, autoscaling and monitoring vLLM-based LLM workloads</strong> on OVHcloud infrastructure. It combines<strong>AI Deploy</strong>, used for <strong>model serving with custom metric autoscaling</strong>, and <strong>Managed Kubernetes Service (MKS)</strong>, which hosts the monitoring and observability stack.</p>



<p>By leveraging <strong>application-level Prometheus metrics exposed by vLLM</strong>, AI Deploy can automatically scale inference replicas based on real workload demand, ensuring <strong>high availability, consistent performance under load and efficient GPU utilisation</strong>. This autoscaling mechanism allows the platform to react dynamically to traffic spikes while maintaining predictable latency for end users.</p>



<p>On top of this scalable inference layer, the monitoring architecture provides <strong>observability</strong> through <strong>Prometheus</strong>, <strong>Grafana</strong> and Alertmanager. It enables real-time performance monitoring, capacity planning, and operational insights, while ensuring <strong>full data sovereignty</strong> for organisations running Large Language Models (LLMs) in production environments.</p>



<p><strong>What are the key benefits</strong>?</p>



<ul class="wp-block-list">
<li><strong>Cost-effective</strong>: Leverage managed services to minimise operational overhead</li>



<li><strong>Real-time observability</strong>: Track Time-to-First-Token (TTFT), throughput, and resource utilisation</li>



<li><strong>Sovereign infrastructure</strong>: All metrics and data remain within European datacentres</li>



<li><strong>Production-ready</strong>: Persistent storage, high availability, and automated monitoring</li>
</ul>



<h2 class="wp-block-heading">Context</h2>



<h3 class="wp-block-heading">AI Deploy</h3>



<p>OVHcloud AI Deploy is a<strong>&nbsp;Container as a Service</strong>&nbsp;(CaaS) platform designed to help you deploy, manage and scale AI models. It provides a solution that allows you to optimally deploy your applications/APIs based on Machine Learning (ML), Deep Learning (DL) or Large Language Models (LLMs).</p>



<p><strong>Key points to keep in mind</strong>:</p>



<ul class="wp-block-list">
<li><strong>Easy to use:</strong>&nbsp;Bring your own custom Docker image and deploy it in a command line or a few clicks surely</li>



<li><strong>High-performance computing:</strong>&nbsp;A complete range of GPUs available (H100, A100, V100S, L40S and L4)</li>



<li><strong>Scalability and flexibility:</strong>&nbsp;Supports automatic scaling, allowing your model to effectively handle fluctuating workloads</li>



<li><strong>Cost-efficient:</strong>&nbsp;Billing per minute, no surcharges</li>
</ul>



<h3 class="wp-block-heading">Managed Kubernetes Service</h3>



<p><strong>OVHcloud MKS</strong> is a fully managed Kubernetes platform designed to help you deploy, operate, and scale containerised applications in production. It provides a secure and reliable Kubernetes environment without the operational overhead of managing the control plane.</p>



<p><strong>What should you keep in mind?</strong></p>



<ul class="wp-block-list">
<li><strong>Cost-efficient</strong>: Only pay for worker nodes and consumed resources, with no additional charge for the Kubernetes control plane</li>



<li><strong>Fully managed Kubernetes</strong>: Certified upstream Kubernetes with automated control plane management, upgrades and high availability</li>



<li><strong>Production-ready by design</strong>: Built-in integrations with OVHcloud Load Balancers, networking and persistent storage</li>



<li><strong>Scalability and flexibility</strong>: Easily scale workloads and node pools to match application demand</li>



<li><strong>Open and portable</strong>: Based on standard Kubernetes APIs, enabling seamless integration with open-source ecosystems and avoiding vendor lock-in</li>
</ul>



<p>In the following guide, all services are deployed within the&nbsp;<strong>OVHcloud Public Cloud</strong>.</p>



<h2 class="wp-block-heading">Overview of the architecture</h2>



<p>This reference architecture describes a <strong>complete, secure and scalable solution</strong> to:</p>



<ul class="wp-block-list">
<li>Deploy an LLM with vLLM and <strong>AI Deploy</strong>, benefiting from automatic scaling based on custom metrics to ensure high service availability &#8211; vLLM exposes <code><mark class="has-inline-color has-ast-global-color-0-color"><strong>/metrics</strong></mark></code> via its public HTTPS endpoint on AI Deploy</li>



<li>Collect, store and visualise these vLLM metrics using Prometheus and Grafana on <strong>MKS</strong></li>
</ul>



<figure class="wp-block-image aligncenter size-full"><img loading="lazy" decoding="async" width="1200" height="630" src="https://blog.ovhcloud.com/wp-content/uploads/2026/02/1.jpg" alt="" class="wp-image-30578" srcset="https://blog.ovhcloud.com/wp-content/uploads/2026/02/1.jpg 1200w, https://blog.ovhcloud.com/wp-content/uploads/2026/02/1-300x158.jpg 300w, https://blog.ovhcloud.com/wp-content/uploads/2026/02/1-1024x538.jpg 1024w, https://blog.ovhcloud.com/wp-content/uploads/2026/02/1-768x403.jpg 768w" sizes="auto, (max-width: 1200px) 100vw, 1200px" /><figcaption class="wp-element-caption"><em>vLLM metrics monitoring and observability architecture overview</em></figcaption></figure>



<p>Here you will find the main components of the architecture. The solution comprises three main layers:</p>



<ol class="wp-block-list">
<li><strong>Model serving layer</strong> with AI Deploy
<ul class="wp-block-list">
<li>vLLM containers running on top of GPUs for LLM inference</li>



<li>vLLM inference server exposing Prometheus metrics</li>



<li>Automatic scaling based on custom metrics to ensure high availability</li>



<li>HTTPS endpoints with Bearer token authentication</li>
</ul>
</li>



<li><strong>Monitoring and observability infrastructure</strong> using Kubernetes
<ul class="wp-block-list">
<li>Prometheus for metrics collection and storage</li>



<li>Grafana for visualisation and dashboards</li>



<li>Persistent volume storage for long-term retention</li>
</ul>
</li>



<li><strong>Network layer</strong>
<ul class="wp-block-list">
<li>Secure HTTPS communication between components</li>



<li>OVHcloud LoadBalancer for external access</li>
</ul>
</li>
</ol>



<p>To go further, some prerequisites must be checked!</p>



<h2 class="wp-block-heading">Prerequisites</h2>



<p>Before you begin, ensure you have:</p>



<ul class="wp-block-list">
<li>An&nbsp;<strong>OVHcloud Public Cloud</strong>&nbsp;account</li>



<li>An&nbsp;<strong>OpenStack user</strong>&nbsp;with the<a href="https://help.ovhcloud.com/csm/en-gb-public-cloud-ai-users?id=kb_article_view&amp;sysparm_article=KB0048170" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer"> </a><strong><code><mark class="has-inline-color has-ast-global-color-0-color">Administrator</mark></code></strong> role</li>



<li><strong>ovhai CLI available</strong> &#8211;&nbsp;<em>install the&nbsp;<a href="https://help.ovhcloud.com/csm/en-gb-public-cloud-ai-cli-install-client?id=kb_article_view&amp;sysparm_article=KB0047844" target="_blank" rel="noreferrer noopener nofollow external" data-wpel-link="external">ovhai CLI</a></em></li>



<li>A <strong>Hugging Face access</strong> &#8211; <em>create a&nbsp;<a href="https://huggingface.co/" target="_blank" rel="noreferrer noopener nofollow external" data-wpel-link="external">Hugging Face account</a>&nbsp;and generate an&nbsp;<a href="https://huggingface.co/settings/tokens" target="_blank" rel="noreferrer noopener nofollow external" data-wpel-link="external">access token</a></em></li>



<li><code><strong><mark class="has-inline-color has-ast-global-color-0-color">kubectl</mark></strong></code> installed and <code><strong><mark class="has-inline-color has-ast-global-color-0-color">helm</mark></strong></code> installed (at least version 3.x)</li>
</ul>



<p><strong>🚀 Now you have all the ingredients for our recipe, it’s time to deploy the Ministral 14B using AI Deploy and vLLM Docker container!</strong></p>



<h2 class="wp-block-heading">Architecture guide: From autoscaling to observability for LLMs served by vLLM</h2>



<p>Let’s set up and deploy this architecture!</p>



<figure class="wp-block-image aligncenter size-large"><img loading="lazy" decoding="async" width="1024" height="538" src="https://blog.ovhcloud.com/wp-content/uploads/2026/02/2-1024x538.jpg" alt="" class="wp-image-30580" srcset="https://blog.ovhcloud.com/wp-content/uploads/2026/02/2-1024x538.jpg 1024w, https://blog.ovhcloud.com/wp-content/uploads/2026/02/2-300x158.jpg 300w, https://blog.ovhcloud.com/wp-content/uploads/2026/02/2-768x403.jpg 768w, https://blog.ovhcloud.com/wp-content/uploads/2026/02/2.jpg 1200w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /><figcaption class="wp-element-caption"><em>Overview of the deployment workflow</em></figcaption></figure>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p><strong>✅ <em>Note</em></strong></p>



<p><strong><em>In this example, <a href="https://huggingface.co/mistralai/Ministral-3-14B-Instruct-2512" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">mistralai/Ministral-3-14B-Instruct-2512</a> is used. Choose the open-source model of your choice and follow the same steps, adapting the model slug (from Hugging Face), the versions and the GPU(s) flavour.</em></strong></p>
</blockquote>



<p><em>Remember that all of the following steps can be automated using OVHcloud APIs!</em></p>



<h3 class="wp-block-heading">Step 1 &#8211; Manage access tokens</h3>



<p>Before introducing the monitoring stack, this architecture starts with the <strong>deployment of the <strong>Ministral 3 14B</strong> on OVHcloud AI Deploy</strong>, configured to <strong>autoscale based on custom Prometheus metrics exposed by vLLM itself</strong>.</p>



<p>Export your&nbsp;<a href="https://huggingface.co/settings/tokens" target="_blank" rel="noreferrer noopener nofollow external" data-wpel-link="external">Hugging Face token</a>.</p>



<pre class="wp-block-code"><code class="">export MY_HF_TOKEN=hf_xxxxxxxxxxxxxxxxxxxx</code></pre>



<p><a href="https://help.ovhcloud.com/csm/en-gb-public-cloud-ai-cli-app-token?id=kb_article_view&amp;sysparm_article=KB0035280" target="_blank" rel="noreferrer noopener nofollow external" data-wpel-link="external">Create a Bearer token</a>&nbsp;to access your AI Deploy app once it&#8217;s been deployed.</p>



<pre class="wp-block-code"><code class="">ovhai token create --role operator ai_deploy_token=my_operator_token</code></pre>



<p>Returning the following output:</p>



<p><code><strong>Id: 47292486-fb98-4a5b-8451-600895597a2b<br>Created At: 20-01-26 11:53:05<br>Updated At: 20-01-26 11:53:05<br>Spec:<br>Name: ai_deploy_token=my_operator_token<br>Role: AiTrainingOperator<br>Label Selector:<br>Status:<br>Value: XXXXXXXXXXXXXXXXXXXXXXXXXXXXXX<br>Version: 1</strong></code></p>



<p>You can now store and export your access token:</p>



<pre class="wp-block-code"><code class="">export MY_OVHAI_ACCESS_TOKEN=XXXXXXXXXXXXXXXXXXXXXXXXXXXXXX</code></pre>



<h3 class="wp-block-heading">Step 2 &#8211; LLM deployment using AI Deploy</h3>



<p>Before introducing the monitoring stack, this architecture starts with the <strong>deployment of the <strong>Ministral 3 14B</strong> on OVHcloud AI Deploy</strong>, configured to <strong>autoscale based on custom Prometheus metrics exposed by vLLM itself</strong>.</p>



<h4 class="wp-block-heading">1. Define the targeted vLLM metric for autoscaling</h4>



<p>Before proceeding with the deployment of the <strong>Ministral 3 14B</strong> endpoint, you have to choose the metric you want to use as the trigger for scaling.</p>



<p>Instead of relying solely on CPU/RAM utilisation, AI Deploy allows autoscaling decisions to be driven by <strong>application-level signals</strong>.</p>



<p>To do this, you can consult the <a href="https://docs.vllm.ai/en/latest/design/metrics/#v1-metrics" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">metrics exposed by vLLM</a>.</p>



<p>In this example, you can use a basic metric such as <code><mark class="has-inline-color has-ast-global-color-0-color"><strong>vllm:num_requests_running</strong></mark></code> to scale the number of replicas based on <strong>real inference load</strong>.</p>



<p>This enables:</p>



<ul class="wp-block-list">
<li>Faster reaction to traffic spikes</li>



<li>Better GPU utilisation</li>



<li>Reduced inference latency under load</li>



<li>Cost-efficient scaling</li>
</ul>



<p>Finally, the configuration chosen for scaling this application is as follows:</p>



<figure class="wp-block-table"><table class="has-fixed-layout"><thead><tr><th>Parameter</th><th>Value</th><th>Description</th></tr></thead><tbody><tr><td>Metric source</td><td><code>/metrics</code></td><td>vLLM Prometheus endpoint</td></tr><tr><td>Metric name</td><td><code>vllm:num_requests_running</code></td><td>Number of in-flight requests</td></tr><tr><td>Aggregation</td><td><code>AVERAGE</code></td><td>Mean across replicas</td></tr><tr><td>Target value</td><td><code>50</code></td><td>Desired load per replica</td></tr><tr><td>Min replicas</td><td><code>1</code></td><td>Baseline capacity</td></tr><tr><td>Max replicas</td><td><code>3</code></td><td>Burst capacity</td></tr></tbody></table></figure>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p><strong>✅ <em>Note</em></strong></p>



<p><em><strong>You can choose the metric that best suits your use case. You can also apply a patch to your AI Deploy deployment at any time to change the target metric for scaling</strong></em>.</p>
</blockquote>



<p>When the <strong>average number of running requests exceeds 50</strong>, AI Deploy automatically provisions <strong>additional GPU-backed replicas</strong>.</p>



<h4 class="wp-block-heading">2. Deploy Ministral 3 14B using AI Deploy</h4>



<p>Now you can deploy the LLM using the <strong><code>ovhai</code> CLI</strong>.</p>



<p>Key elements necessary for proper functioning:</p>



<ul class="wp-block-list">
<li>GPU-based inference: <strong><code><mark class="has-inline-color has-ast-global-color-0-color">1 x H100</mark></code></strong></li>



<li>vLLM OpenAI-compatible Docker image: <a href="https://hub.docker.com/r/vllm/vllm-openai/tags" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer"><strong><code><mark class="has-inline-color has-ast-global-color-0-color">vllm/vllm-openai:v0.13.0</mark></code></strong></a></li>



<li>Custom autoscaling rules based on Prometheus metrics: <code><strong><mark class="has-inline-color has-ast-global-color-0-color">vllm:num_requests_running</mark></strong></code></li>
</ul>



<p>Below is the reference command used to deploy the <strong><a href="https://huggingface.co/mistralai/Ministral-3-14B-Instruct-2512" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">mistralai/Ministral-3-14B-Instruct-2512</a></strong>:</p>



<pre class="wp-block-code"><code class="">ovhai app run \<br>  --name vllm-ministral-14B-autoscaling-custom-metric \<br>  --default-http-port 8000 \<br>  --label ai_deploy_token=my_operator_token \<br>  --gpu 1 \<br>  --flavor h100-1-gpu \<br>  -e OUTLINES_CACHE_DIR=/tmp/.outlines \<br>  -e HF_TOKEN=$MY_HF_TOKEN \<br>  -e HF_HOME=/hub \<br>  -e HF_DATASETS_TRUST_REMOTE_CODE=1 \<br>  -e HF_HUB_ENABLE_HF_TRANSFER=0 \<br>  -v standalone:/hub:rw \<br>  -v standalone:/workspace:rw \<br>  --liveness-probe-path /health \<br>  --liveness-probe-port 8000 \<br>  --liveness-initial-delay-seconds 300 \<br>  --probe-path /v1/models \<br>  --probe-port 8000 \<br>  --initial-delay-seconds 300 \<br>  --auto-min-replicas 1 \<br>  --auto-max-replicas 3 \<br>  --auto-custom-api-url "http://&lt;SELF&gt;:8000/metrics" \<br>  --auto-custom-metric-format PROMETHEUS \<br>  --auto-custom-value-location vllm:num_requests_running \<br>  --auto-custom-target-value 50 \<br>  --auto-custom-metric-aggregation-type AVERAGE \<br>  vllm/vllm-openai:v0.13.0 \<br>  -- bash -c "python3 -m vllm.entrypoints.openai.api_server \<br>    --model mistralai/Ministral-3-14B-Instruct-2512 \<br>    --tokenizer_mode mistral \<br>    --load_format mistral \<br>    --config_format mistral \<br>    --enable-auto-tool-choice \<br>    --tool-call-parser mistral \<br>    --enable-prefix-caching"</code></pre>



<p>How to understand the different parameters of this command?</p>



<h5 class="wp-block-heading"><strong>a. Start your AI Deploy app</strong></h5>



<p>Launch a new app using&nbsp;<a href="https://help.ovhcloud.com/csm/en-gb-public-cloud-ai-cli-install-client?id=kb_article_view&amp;sysparm_article=KB0047844" target="_blank" rel="noreferrer noopener nofollow external" data-wpel-link="external">ovhai CLI</a>&nbsp;and name it.</p>



<p><code><strong>ovhai app run --name vllm-ministral-14B-autoscaling-custom-metric</strong></code></p>



<h5 class="wp-block-heading"><strong>b. Define access</strong></h5>



<p>Define the HTTP API port and restrict access to your token.</p>



<p><strong><code>--default-http-port 8000</code><br><code>--label ai_deploy_token=my_operator_token</code></strong></p>



<h5 class="wp-block-heading"><strong>c. Configure GPU resources</strong></h5>



<p>Specify the hardware type (<code><strong>h100-1-gpu</strong></code>), which refers to an&nbsp;<strong>NVIDIA H100 GPU</strong>&nbsp;and the number (<strong>1</strong>).</p>



<p><code><strong>--gpu 1<br>--flavor h100-1-gpu</strong></code></p>



<p><strong><mark>⚠️WARNING!</mark></strong>&nbsp;For this model, one H100 is sufficient, but if you want to deploy another model, you will need to check which GPU you need. Note that you can also access L40S and A100 GPUs for your LLM deployment.</p>



<h5 class="wp-block-heading"><strong>d. Set up environment variables</strong></h5>



<p>Configure caching for the&nbsp;<strong>Outlines library</strong>&nbsp;(used for efficient text generation):</p>



<p><code><strong>-e OUTLINES_CACHE_DIR=/tmp/.outlines</strong></code></p>



<p>Pass the&nbsp;<strong>Hugging Face token</strong>&nbsp;(<code>$MY_HF_TOKEN</code>) for model authentication and download:</p>



<p><code><strong>-e HF_TOKEN=$MY_HF_TOKEN</strong></code></p>



<p>Set the&nbsp;<strong>Hugging Face cache directory</strong>&nbsp;to&nbsp;<code>/hub</code>&nbsp;(where models will be stored):</p>



<p><code><strong>-e HF_HOME=/hub</strong></code></p>



<p>Allow execution of&nbsp;<strong>custom remote code</strong>&nbsp;from Hugging Face datasets (required for some model behaviours):</p>



<p><code><strong>-e HF_DATASETS_TRUST_REMOTE_CODE=1</strong></code></p>



<p>Disable&nbsp;<strong>Hugging Face Hub transfer acceleration</strong>&nbsp;(to use standard model downloading):</p>



<p><code><strong>-e HF_HUB_ENABLE_HF_TRANSFER=0</strong></code></p>



<h5 class="wp-block-heading"><strong>e. Mount persistent volumes</strong></h5>



<p>Mount&nbsp;<strong>two persistent storage volumes</strong>:</p>



<ol class="wp-block-list">
<li><code>/hub</code>&nbsp;→ Stores Hugging Face model files</li>



<li><code>/workspace</code>&nbsp;→ Main working directory</li>
</ol>



<p>The&nbsp;<code>rw</code>&nbsp;flag means&nbsp;<strong>read-write access</strong>.</p>



<p><code><strong>-v standalone:/hub:rw<br>-v standalone:/workspace:rw</strong></code></p>



<h5 class="wp-block-heading"><strong>f. Health checks and readiness</strong></h5>



<p>Configure <strong>liveness and readiness probes</strong>:</p>



<ol class="wp-block-list">
<li><code>/health</code> verifies the container is alive</li>



<li><code>/v1/models</code> confirms the model is loaded and ready to serve requests</li>
</ol>



<p>The long initial delays (300 seconds) can be reduced; they correspond to the startup time of vLLM and the loading of the model on the GPU.</p>



<p><code><strong>--liveness-probe-path /health<br>--liveness-probe-port 8000<br>--liveness-initial-delay-seconds 300<br><br>--probe-path /v1/models<br>--probe-port 8000<br>--initial-delay-seconds 300</strong></code></p>



<h5 class="wp-block-heading"><strong>g. Autoscaling configuration (custom metrics)</strong></h5>



<p>First set the minimum and maximum number of replicas.</p>



<p><strong><code>--auto-min-replicas 1<br>--auto-max-replicas 3</code></strong></p>



<p>This guarantees basic availability (one replica always up) while allowing for peak capacity.</p>



<p>Then enable autoscaling based on application-level metrics exposed by vLLM.</p>



<p><strong><code>--auto-custom-api-url "http://&lt;SELF&gt;:8000/metrics"<br>--auto-custom-metric-format PROMETHEUS<br>--auto-custom-value-location vllm:num_requests_running<br>--auto-custom-target-value 50<br>--auto-custom-metric-aggregation-type AVERAGE</code></strong></p>



<p>AI Deploy:</p>



<ul class="wp-block-list">
<li>Scrapes the local <mark class="has-inline-color has-ast-global-color-0-color"><strong><code>/metrics</code></strong></mark> endpoint</li>



<li>Parses Prometheus-formatted metrics</li>



<li>Extracts the <strong><mark class="has-inline-color has-ast-global-color-0-color"><code>vllm:num_requests_running</code></mark></strong> gauge</li>



<li>Computes the average value across replicas</li>
</ul>



<p>Scaling behaviour:</p>



<ul class="wp-block-list">
<li>When the average number of in-flight requests exceeds <strong><code><mark class="has-inline-color has-ast-global-color-0-color">50</mark></code></strong>, AI Deploy adds replicas</li>



<li>When load decreases, replicas are scaled down</li>
</ul>



<p>This approach ensures high availability and predictable latency under fluctuating traffic.</p>



<h5 class="wp-block-heading"><strong>h. Choose the target Docker image and the startup command</strong></h5>



<p>Use the official <strong><a href="https://hub.docker.com/r/vllm/vllm-openai/tags" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">vLLM OpenAI-compatible Docker image</a></strong>.</p>



<p><strong><code>vllm/vllm-openai:v0.13.0</code></strong></p>



<p>Finally, run the model inside the container using a Python command to launch the vLLM API server:</p>



<ul class="wp-block-list">
<li><strong><code>python3 -m vllm.entrypoints.openai.api_server</code></strong>&nbsp;→ Starts the OpenAI-compatible vLLM API server</li>



<li><strong><code>--model mistralai/Ministral-3-14B-Instruct-2512</code></strong>&nbsp;→ Loads the&nbsp;<strong>Ministral 3 14B</strong>&nbsp;model from Hugging Face</li>



<li><strong><code>--tokenizer_mode mistral</code></strong>&nbsp;→ Uses the&nbsp;<strong>Mistral tokenizer</strong></li>



<li><strong><code>--load_format mistral</code></strong>&nbsp;→ Uses Mistral’s model loading format</li>



<li><strong><code>--config_format mistral</code></strong>&nbsp;→ Ensures the model configuration follows Mistral’s standard</li>



<li><code><strong>--enable-auto-tool-choice </strong></code>→ Automatic call of tools if necessary (function/tool call)</li>



<li><strong><code>--tool-call-parser mistral </code></strong>→ Tool calling support</li>



<li><strong><code>--enable-prefix-caching</code></strong> → Prefix caching for improved throughput and reduced latency</li>
</ul>



<p>You can now launch this command using <strong>ovhai CLI</strong>.</p>



<h4 class="wp-block-heading">3. Check AI Deploy app status</h4>



<p>You can now check if your&nbsp;<strong>AI Deploy</strong>&nbsp;app is alive:</p>



<pre class="wp-block-code"><code class="">ovhai app get &lt;your_vllm_app_id&gt;</code></pre>



<p><strong>Is your app in&nbsp;<code>RUNNING</code>&nbsp;status?</strong>&nbsp;Perfect! You can check in the logs that the server is started:</p>



<pre class="wp-block-code"><code class="">ovhai app logs &lt;your_vllm_app_id&gt;</code></pre>



<p><strong><mark>⚠️WARNING!</mark></strong>&nbsp;This step may take a little time as the LLM must be loaded.</p>



<h4 class="wp-block-heading">4. Test that the deployment is functional</h4>



<p>First you can request and send a prompt to the LLM. Launch the following query by asking the question of your choice:</p>



<pre class="wp-block-code"><code class="">curl https://&lt;your_vllm_app_id&gt;.app.gra.ai.cloud.ovh.net/v1/chat/completions \<br>  -H "Authorization: Bearer $MY_OVHAI_ACCESS_TOKEN" \<br>  -H "Content-Type: application/json" \<br>  -d '{<br>    "model": "mistralai/Ministral-3-14B-Instruct-2512",<br>    "messages": [<br>      {"role": "system", "content": "You are a helpful assistant."},<br>      {"role": "user", "content": "Give me the name of OVHcloud’s founder."}<br>    ],<br>    "stream": false<br>  }'</code></pre>



<p>You can also verify access to vLLM metrics.</p>



<pre class="wp-block-code"><code class="">curl -H "Authorization: Bearer $MY_OVHAI_ACCESS_TOKEN" \<br>  https://&lt;your_vllm_app_id&gt;.app.gra.ai.cloud.ovh.net/metrics</code></pre>



<p>If both tests show that the model deployment is functional and you receive 200 HTTP responses, you are ready to move on to the next step!</p>



<p>The next step is to set up the observability and monitoring stack. This autoscaling mechanism is <strong>fully independent</strong> from Prometheus used for observability:</p>



<ul class="wp-block-list">
<li>AI Deploy queries the local <strong><mark class="has-inline-color has-ast-global-color-0-color"><code>/metrics</code></mark></strong> endpoint internally</li>



<li>Prometheus scrapes the <strong>same metrics endpoint</strong> externally for monitoring, dashboards and potentially alerting</li>
</ul>



<p>This ensures:</p>



<ul class="wp-block-list">
<li>A single source of truth for metrics</li>



<li>No duplication of exporters</li>



<li>Consistent signals for scaling and observability</li>
</ul>



<h3 class="wp-block-heading">Step 3 &#8211; Create an MKS cluster</h3>



<p>From <a href="https://manager.eu.ovhcloud.com/#/hub/" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">OVHcloud Control Panel</a>, create a Kubernetes cluster using the <strong>MKS</strong>.</p>



<p>Consider using the following configuration for the current use case:</p>



<ul class="wp-block-list">
<li><strong>Location</strong>: GRA ( Gravelines) &#8211; <em>you can select the same region as for AI Deploy</em></li>



<li><strong>Network</strong>: Public</li>



<li><strong>Node pool</strong> :
<ul class="wp-block-list">
<li>Flavour : <code><strong><mark class="has-inline-color has-ast-global-color-0-color">b2-15</mark></strong></code> (or something similar)</li>



<li>Number of nodes: <strong><code><mark class="has-inline-color has-ast-global-color-0-color">3</mark></code></strong></li>



<li>Autoscaling : <strong><code><mark class="has-inline-color has-ast-global-color-0-color">OFF</mark></code></strong></li>
</ul>
</li>



<li><strong>Name your node pool:</strong> <strong><mark class="has-inline-color has-ast-global-color-0-color"><code>monitoring</code></mark></strong></li>
</ul>



<p>You should see your cluster (e.g. <code><mark class="has-inline-color has-ast-global-color-0-color"><strong>prometheus-vllm-metrics-ai-deploy</strong></mark></code>) in the list, along with the following information:</p>



<figure class="wp-block-image aligncenter size-large"><img loading="lazy" decoding="async" width="1024" height="632" src="https://blog.ovhcloud.com/wp-content/uploads/2026/01/image-3-1024x632.png" alt="" class="wp-image-30242" srcset="https://blog.ovhcloud.com/wp-content/uploads/2026/01/image-3-1024x632.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2026/01/image-3-300x185.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2026/01/image-3-768x474.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2026/01/image-3-1536x948.png 1536w, https://blog.ovhcloud.com/wp-content/uploads/2026/01/image-3-2048x1264.png 2048w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p>If the status is green with the <strong><mark style="color:#00d084" class="has-inline-color"><code>OK</code></mark></strong> label, you can proceed to the next step.</p>



<h3 class="wp-block-heading">Step 4 &#8211; Configure Kubernetes access</h3>



<p>Download your <strong>kubeconfig file</strong> from the OVHcloud Control Panel and configure <strong><code><mark class="has-inline-color has-ast-global-color-0-color">kubectl</mark></code></strong>:</p>



<pre class="wp-block-code"><code class=""># configure kubectl with your MKS cluster<br>export KUBECONFIG=/path/to/your/kubeconfig-xxxxxx.yml<br><br># verify cluster connectivity<br>kubectl cluster-info<br>kubectl get nodes</code></pre>



<p>Now,- you can create the <strong><mark class="has-inline-color has-ast-global-color-0-color"><code>values-prometheus.yaml</code></mark></strong> file:</p>



<pre class="wp-block-code"><code class=""># general configuration<br>nameOverride: "monitoring"<br>fullnameOverride: "monitoring"<br><br># Prometheus configuration<br>prometheus:<br>  prometheusSpec:<br>    # data retention (15d)<br>    retention: 15d<br>    <br>    # scrape interval (15s)<br>    scrapeInterval: 15s<br>    <br>    # persistent storage (required for production deployment)<br>    storageSpec:<br>      volumeClaimTemplate:<br>        spec:<br>          storageClassName: csi-cinder-high-speed  # OVHcloud storage<br>          accessModes: ["ReadWriteOnce"]<br>          resources:<br>            requests:<br>              storage: 50Gi  # (can be modified according to your needs)<br>    <br>    # scrape vLLM metrics from your AI Deploy instance (Ministral 3 14B)<br>    additionalScrapeConfigs:<br>      - job_name: 'vllm-ministral'<br>        scheme: https<br>        metrics_path: '/metrics'<br>        scrape_interval: 15s<br>        scrape_timeout: 10s<br>        <br>        # authentication using AI Deploy Bearer token stored Kubernetes Secret<br>        bearer_token_file: /etc/prometheus/secrets/vllm-auth-token/token<br>        static_configs:<br>          - targets:<br>              - '&lt;APP_ID&gt;.app.gra.ai.cloud.ovh.net'  # /!\ REPLACE THE &lt;APP_ID&gt; by yours /!\<br>            labels:<br>              service: 'vllm'<br>              model: 'ministral'<br>              environment: 'production'<br>        <br>        # TLS configuration<br>        tls_config:<br>          insecure_skip_verify: false<br>    <br>    # kube-prometheus-stack mounts the secret under /etc/prometheus/secrets/ and makes it accessible to Prometheus<br>    secrets:<br>      - vllm-auth-token<br><br># Grafana configuration (visualization layer)<br>grafana:<br>  enabled: true<br>  <br>  # disable automatic datasource provisioning<br>  sidecar:<br>    datasources:<br>      enabled: false<br>  <br>  # persistent dashboards<br>  persistence:<br>    enabled: true<br>    storageClassName: csi-cinder-high-speed<br>    size: 10Gi<br>  <br>  # /!\ DEFINE ADMIN PASSWORD - REPLACE "test" BY YOURS /!\<br>  adminPassword: "test"<br>  <br>  # access via OVHcloud LoadBalancer (public IP and managed LB)<br>  service:<br>    type: LoadBalancer<br>    port: 80<br>    annotations:<br>      # optional : limiter l'accès à certaines IPs<br>      # service.beta.kubernetes.io/ovh-loadbalancer-allowed-sources: "1.2.3.4/32"<br>  <br># alertmanager (optional but recommended for production)<br>alertmanager:<br>  enabled: true<br>  <br>  alertmanagerSpec:<br>    storage:<br>      volumeClaimTemplate:<br>        spec:<br>          storageClassName: csi-cinder-high-speed<br>          accessModes: ["ReadWriteOnce"]<br>          resources:<br>            requests:<br>              storage: 10Gi<br><br># cluster observability components<br>nodeExporter:<br>  enabled: true<br>  <br>kubeStateMetrics:<br>  enabled: true</code></pre>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p><strong>✅ <em>Note</em></strong></p>



<p><strong><em>On OVHcloud MKS, persistent storage is handled automatically through the Cinder CSI driver. When a PersistentVolumeClaim (PVC) references a supported <code>storageClassName</code> such as <code>csi-cinder-high-speed</code>, OVHcloud dynamically provisions the underlying Block Storage volume and attaches it to the node running the pod. This enables stateful components like Prometheus, Alertmanager and Grafana to persist data reliably without any manual volume management, making the architecture fully cloud-native and operationally simple.</em></strong></p>
</blockquote>



<p>Then create the <strong><code><mark class="has-inline-color has-ast-global-color-0-color">monitoring</mark></code></strong> namespace:</p>



<pre class="wp-block-code"><code class=""># create namespace<br>kubectl create namespace monitoring<br><br># verify creation<br>kubectl get namespaces | grep monitoring</code></pre>



<p>Finally,  configure the Bearer token secret to access vLLM metrics.</p>



<pre class="wp-block-code"><code class=""># create bearer token secret<br>kubectl create secret generic vllm-auth-token \<br>  --from-literal=token='"$MY_OVHAI_ACCESS_TOKEN"' \<br>  -n monitoring<br><br># verify secret creation<br>kubectl get secret vllm-auth-token -n monitoring<br><br># test token (optional)<br>kubectl get secret vllm-auth-token -n monitoring \<br>  -o jsonpath='{.data.token}' | base64 -d </code></pre>



<p>Right, if everything is working, let&#8217;s move on to deployment.</p>



<h3 class="wp-block-heading">Step 5 &#8211; Deploy Prometheus stack</h3>



<p>Add the Prometheus Helm repository and install the monitoring stack. The deployment creates:</p>



<ul class="wp-block-list">
<li>Prometheus StatefulSet with persistent storage</li>



<li>Grafana deployment with LoadBalancer access</li>



<li>Alertmanager for future alert configuration (optional)</li>



<li>Supporting components (node exporters, kube-state-metrics)</li>
</ul>



<pre class="wp-block-code"><code class=""># add Helm repository<br>helm repo add prometheus-community \<br>  https://prometheus-community.github.io/helm-charts<br>helm repo update<br><br># install monitoring stack<br>helm install monitoring prometheus-community/kube-prometheus-stack \<br>  --namespace monitoring \<br>  --values values-prometheus.yaml \<br>  --wait</code></pre>



<p>Then you can retrieve the LoadBalancer IP address to access Grafana:</p>



<pre class="wp-block-code"><code class="">kubectl get svc -n monitoring monitoring-grafana</code></pre>



<p>Finally, open your browser to <code><strong><mark class="has-inline-color has-ast-global-color-0-color">http://&lt;EXTERNAL-IP&gt;</mark></strong></code> and login with:</p>



<ul class="wp-block-list">
<li><strong>Username</strong>: <code><mark class="has-inline-color has-ast-global-color-0-color"><strong>admin</strong></mark></code></li>



<li><strong>Password</strong>: as configured in your <code><strong><mark class="has-inline-color has-ast-global-color-0-color">values-prometheus.yaml</mark></strong></code> file</li>
</ul>



<h3 class="wp-block-heading">Step 6 &#8211; Create Grafana dashboards</h3>



<p>In this step, you will be able to access Grafana interface and add your Prometheus as a new data source, then create a complete dashboard with different vLLM metrics.</p>



<h4 class="wp-block-heading">1. Add a new data source in Grafana</h4>



<p>First of all, create a new Prometheus connection inside Grafana:</p>



<ul class="wp-block-list">
<li>Navigate to <strong><mark class="has-inline-color has-ast-global-color-0-color"><code>Connections</code></mark></strong> → <strong><mark class="has-inline-color has-ast-global-color-0-color"><code>Data sources</code></mark></strong> → <strong><code><mark class="has-inline-color has-ast-global-color-0-color">Add data source</mark></code></strong></li>



<li>Select <strong>Prometheus</strong></li>



<li>Configure URL: <code><strong><mark class="has-inline-color has-ast-global-color-0-color">http://monitoring-prometheus:9090</mark></strong></code></li>



<li>Click <strong>Save &amp; test</strong></li>
</ul>



<figure class="wp-block-image size-large"><img loading="lazy" decoding="async" width="1024" height="609" src="https://blog.ovhcloud.com/wp-content/uploads/2026/01/image-4-1024x609.png" alt="" class="wp-image-30247" srcset="https://blog.ovhcloud.com/wp-content/uploads/2026/01/image-4-1024x609.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2026/01/image-4-300x178.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2026/01/image-4-768x457.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2026/01/image-4-1536x913.png 1536w, https://blog.ovhcloud.com/wp-content/uploads/2026/01/image-4-2048x1218.png 2048w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p>Now that your Prometheus has been configured as a new data source, you can create your Grafana dashboard.</p>



<h4 class="wp-block-heading">2. Create your monitoring dashboard</h4>



<p>To begin with, you can use the following pre-configured Grafana dashboard by downloading this JSON file locally:</p>





<p>In the left-hand menu, select <strong><code><mark class="has-inline-color has-ast-global-color-0-color">Dashboard</mark></code></strong>:</p>



<ol class="wp-block-list">
<li>Navigate to <strong><code><mark class="has-inline-color has-ast-global-color-0-color">Dashboards</mark></code></strong> → <strong><code><mark class="has-inline-color has-ast-global-color-0-color">Import</mark></code></strong></li>



<li>Upload the provided dashboard JSON</li>



<li>Select <strong>Prometheus</strong> as datasource</li>



<li>Click <strong>Import</strong> and select the <strong><code><mark class="has-inline-color has-ast-global-color-0-color">vLLM-metrics-grafana-monitoring.json</mark></code></strong> file</li>
</ol>



<figure class="wp-block-image size-large"><img loading="lazy" decoding="async" width="1024" height="449" src="https://blog.ovhcloud.com/wp-content/uploads/2026/01/image-6-1024x449.png" alt="" class="wp-image-30250" srcset="https://blog.ovhcloud.com/wp-content/uploads/2026/01/image-6-1024x449.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2026/01/image-6-300x131.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2026/01/image-6-768x337.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2026/01/image-6-1536x673.png 1536w, https://blog.ovhcloud.com/wp-content/uploads/2026/01/image-6-2048x897.png 2048w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p>The dashboard provides real-time visibility for <strong>Ministral 3 14B</strong> deployed with vLLM container and OVHcloud AI Deploy.</p>



<p>You can now track:</p>



<ul class="wp-block-list">
<li><strong>Performance metrics</strong>: TTFT, inter-token latency, end-to-end latency</li>



<li><strong>Throughput indicators</strong>: Requests per second, token generation rates</li>



<li><strong>Resource utilisation</strong>: KV cache usage, active/waiting requests</li>



<li><strong>Capacity indicators</strong>: Queue depth, preemption rates</li>
</ul>



<figure class="wp-block-image size-large"><img loading="lazy" decoding="async" width="1024" height="540" src="https://blog.ovhcloud.com/wp-content/uploads/2026/01/image-7-1024x540.png" alt="" class="wp-image-30253" srcset="https://blog.ovhcloud.com/wp-content/uploads/2026/01/image-7-1024x540.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2026/01/image-7-300x158.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2026/01/image-7-768x405.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2026/01/image-7-1536x811.png 1536w, https://blog.ovhcloud.com/wp-content/uploads/2026/01/image-7-2048x1081.png 2048w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p>Here are the key metrics tracked and displayed in the Grafana dashboard:</p>



<figure class="wp-block-table"><table class="has-fixed-layout"><thead><tr><th>Metric Category</th><th>Prometheus Metric</th><th>Description</th><th>Use case</th></tr></thead><tbody><tr><td><strong>Latency</strong></td><td><code>vllm:time_to_first_token_seconds</code></td><td>Time until first token generation</td><td>User experience monitoring</td></tr><tr><td><strong>Latency</strong></td><td><code>vllm:inter_token_latency_seconds</code></td><td>Time between tokens</td><td>Throughput optimisation</td></tr><tr><td><strong>Latency</strong></td><td><code>vllm:e2e_request_latency_seconds</code></td><td>End-to-end request time</td><td>SLA monitoring</td></tr><tr><td><strong>Throughput</strong></td><td><code>vllm:request_success_total</code></td><td>Successful requests counter</td><td>Capacity planning</td></tr><tr><td><strong>Resource</strong></td><td><code>vllm:kv_cache_usage_perc</code></td><td>KV cache memory usage</td><td>Memory management</td></tr><tr><td><strong>Queue</strong></td><td><code>vllm:num_requests_running</code></td><td>Active requests</td><td>Load monitoring</td></tr><tr><td><strong>Queue</strong></td><td><code>vllm:num_requests_waiting</code></td><td>Queued requests</td><td>Overload detection</td></tr><tr><td><strong>Capacity</strong></td><td><code>vllm:num_preemptions_total</code></td><td>Request preemptions</td><td>Peak load indicator</td></tr><tr><td><strong>Tokens</strong></td><td><code>vllm:prompt_tokens_total</code></td><td>Input tokens processed</td><td>Usage analytics</td></tr><tr><td><strong>Tokens</strong></td><td><code>vllm:generation_tokens_total</code></td><td>Output tokens generated</td><td>Cost tracking</td></tr></tbody></table></figure>



<p>Well done, you now have at your disposal:</p>



<ul class="wp-block-list">
<li>An endpoint of the Ministral 3 14B model deployed with vLLM thanks to <strong>OVHcloud AI Deploy</strong> and its autoscaling strategies based on custom metrics</li>



<li>Prometheus for metrics collection and Grafana for visualisation/dashboards thanks to <strong>OVHcloud MKS</strong></li>
</ul>



<p><strong>But how can you check that everything will work when the load increases?</strong></p>



<h3 class="wp-block-heading">Step 7 &#8211; Test autoscaling and real-time visualisation</h3>



<p>The first objective here is to force AI Deploy to:</p>



<ul class="wp-block-list">
<li>Increase <code>vllm:num_requests_running</code></li>



<li>&#8216;Saturate&#8217; a single replica</li>



<li>Trigger the <strong>scale up</strong></li>



<li>Observe replica increase + latency drop</li>
</ul>



<h4 class="wp-block-heading">1. Autoscaling testing strategy</h4>



<p>The goal is to combine:</p>



<ul class="wp-block-list">
<li><strong>High concurrency</strong></li>



<li><strong>Long prompts</strong> (KVcache heavy)</li>



<li><strong>Long generations</strong></li>



<li><strong>Bursty load</strong></li>
</ul>



<p>This is what vLLM autoscaling actually reacts to.</p>



<p>To do so, a Python code can simulate the expected behaviour:</p>



<pre class="wp-block-code"><code class="">import time<br>import threading<br>import random<br>from statistics import mean<br>from openai import OpenAI<br>from tqdm import tqdm<br><br>APP_URL = "https://&lt;APP_ID&gt;.app.gra.ai.cloud.ovh.net/v1" # /!\ REPLACE THE &lt;APP_ID&gt; by yours /!\<br>MODEL = "mistralai/Ministral-3-14B-Instruct-2512"<br>API_KEY = $MY_OVHAI_ACCESS_TOKEN<br><br>CONCURRENT_WORKERS = 500          # concurrency (main scaling trigger)<br>REQUESTS_PER_WORKER = 25<br>MAX_TOKENS = 768                  # generation pressure<br><br># some random prompts<br>SHORT_PROMPTS = [<br>    "Summarize the theory of relativity.",<br>    "Explain what a transformer model is.",<br>    "What is Kubernetes autoscaling?"<br>]<br><br>MEDIUM_PROMPTS = [<br>    "Explain how attention mechanisms work in transformer-based models, including self-attention and multi-head attention.",<br>    "Describe how vLLM manages KV cache and why it impacts inference performance."<br>]<br><br>LONG_PROMPTS = [<br>    "Write a very detailed technical explanation of how large language models perform inference, "<br>    "including tokenization, embedding lookup, transformer layers, attention computation, KV cache usage, "<br>    "GPU memory management, and how batching affects latency and throughput. Use examples.",<br>]<br><br>PROMPT_POOL = (<br>    SHORT_PROMPTS * 2 +<br>    MEDIUM_PROMPTS * 4 +<br>    LONG_PROMPTS * 6    # bias toward long prompts<br>)<br><br># openai compliance<br>client = OpenAI(<br>    base_url=APP_URL,<br>    api_key=API_KEY,<br>)<br><br># basic metrics<br>latencies = []<br>errors = 0<br>lock = threading.Lock()<br><br># worker<br>def worker(worker_id):<br>    global errors<br>    for _ in range(REQUESTS_PER_WORKER):<br>        prompt = random.choice(PROMPT_POOL)<br><br>        start = time.time()<br>        try:<br>            client.chat.completions.create(<br>                model=MODEL,<br>                messages=[{"role": "user", "content": prompt}],<br>                max_tokens=MAX_TOKENS,<br>                temperature=0.7,<br>            )<br>            elapsed = time.time() - start<br><br>            with lock:<br>                latencies.append(elapsed)<br><br>        except Exception as e:<br>            with lock:<br>                errors += 1<br><br># run<br>threads = []<br>start_time = time.time()<br><br>print("Starting autoscaling stress test...")<br>print(f"Concurrency: {CONCURRENT_WORKERS}")<br>print(f"Total requests: {CONCURRENT_WORKERS * REQUESTS_PER_WORKER}")<br><br>for i in range(CONCURRENT_WORKERS):<br>    t = threading.Thread(target=worker, args=(i,))<br>    t.start()<br>    threads.append(t)<br><br>for t in threads:<br>    t.join()<br><br>total_time = time.time() - start_time<br><br># results<br>print("\n=== AUTOSCALING BENCH RESULTS ===")<br>print(f"Total requests sent: {len(latencies) + errors}")<br>print(f"Successful requests: {len(latencies)}")<br>print(f"Errors: {errors}")<br>print(f"Total wall time: {total_time:.2f}s")<br><br>if latencies:<br>    print(f"Avg latency: {mean(latencies):.2f}s")<br>    print(f"Min latency: {min(latencies):.2f}s")<br>    print(f"Max latency: {max(latencies):.2f}s")<br>    print(f"Throughput: {len(latencies)/total_time:.2f} req/s")</code></pre>



<p><strong>How can you verify that autoscaling is working and that the load is being handled correctly without latency skyrocketing?</strong></p>



<h4 class="wp-block-heading">2. Hardware and platform-level monitoring</h4>



<p>First, <strong>AI Deploy Grafana</strong> answers <strong>&#8216;What resources are being used and how many replicas exist?</strong>&#8216;.</p>



<p>GPU utilisation, GPU memory, CPU, RAM and replica count are monitored through <strong>OVHcloud AI Deploy Grafana</strong> (monitoring URL), which exposes infrastructure and runtime metrics for the AI Deploy application. This layer provides visibility into <strong>resource saturation and scaling events</strong> managed by the AI Deploy platform itself.</p>



<p>Access it using the following URL (do not forget to replace <code><mark class="has-inline-color has-ast-global-color-0-color"><strong>&lt;APP_ID&gt;</strong></mark></code> by yours): <strong><code>https://monitoring.gra.ai.cloud.ovh.net/d/app/app-monitoring?var-app=</code><mark class="has-inline-color has-ast-global-color-0-color"><code>&lt;APP_ID&gt;</code></mark><code>&amp;orgId=1</code></strong></p>



<p>For example, check GPU/RAM metrics:</p>



<figure class="wp-block-image size-large"><img loading="lazy" decoding="async" width="1024" height="540" src="https://blog.ovhcloud.com/wp-content/uploads/2026/01/image-8-1024x540.png" alt="" class="wp-image-30260" srcset="https://blog.ovhcloud.com/wp-content/uploads/2026/01/image-8-1024x540.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2026/01/image-8-300x158.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2026/01/image-8-768x405.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2026/01/image-8-1536x811.png 1536w, https://blog.ovhcloud.com/wp-content/uploads/2026/01/image-8-2048x1081.png 2048w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p>You can also monitor scale ups and downs in real time, as well as information on HTTP calls and much more!</p>



<figure class="wp-block-image size-large"><img loading="lazy" decoding="async" width="1024" height="540" src="https://blog.ovhcloud.com/wp-content/uploads/2026/01/image-9-1024x540.png" alt="" class="wp-image-30261" srcset="https://blog.ovhcloud.com/wp-content/uploads/2026/01/image-9-1024x540.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2026/01/image-9-300x158.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2026/01/image-9-768x405.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2026/01/image-9-1536x811.png 1536w, https://blog.ovhcloud.com/wp-content/uploads/2026/01/image-9-2048x1081.png 2048w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<h4 class="wp-block-heading">3. Software and application-level monitoring</h4>



<p>Next the combination of MKS + Prometheus + Grafana answers <strong>&#8216;How the inference engine behaves internally&#8217;</strong>.</p>



<p>In fact, vLLM internal metrics (request concurrency, token throughput, latency indicators, KV cache pressure, etc.) are collected via the <strong>vLLM <code>/metrics</code> endpoint</strong> and scraped by <strong>Prometheus running on OVHcloud MKS</strong>, then visualised in a <strong>dedicated Grafana instance</strong>. This layer focuses on <strong>model behaviour and inference performance</strong>.</p>



<p>Find all these metrics via (just replace <strong><code><mark class="has-inline-color has-ast-global-color-0-color">&lt;EXTERNAL-IP&gt;</mark></code></strong>): <strong><code>http://<mark class="has-inline-color has-ast-global-color-0-color">&lt;EXTERNAL-IP&gt;</mark>/d/vllm-ministral-monitoring/ministral-14b-vllm-metrics-monitoring?orgId=1</code></strong></p>



<p>Find key metrics such as TTF, etc:</p>



<figure class="wp-block-image size-large"><img loading="lazy" decoding="async" width="1024" height="540" src="https://blog.ovhcloud.com/wp-content/uploads/2026/01/image-10-1024x540.png" alt="" class="wp-image-30263" srcset="https://blog.ovhcloud.com/wp-content/uploads/2026/01/image-10-1024x540.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2026/01/image-10-300x158.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2026/01/image-10-768x405.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2026/01/image-10-1536x811.png 1536w, https://blog.ovhcloud.com/wp-content/uploads/2026/01/image-10-2048x1081.png 2048w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p>You can also find some information about <strong>&#8216;Model load and throughput&#8217;</strong>:</p>



<figure class="wp-block-image size-large"><img loading="lazy" decoding="async" width="1024" height="540" src="https://blog.ovhcloud.com/wp-content/uploads/2026/01/image-11-1024x540.png" alt="" class="wp-image-30264" srcset="https://blog.ovhcloud.com/wp-content/uploads/2026/01/image-11-1024x540.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2026/01/image-11-300x158.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2026/01/image-11-768x405.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2026/01/image-11-1536x811.png 1536w, https://blog.ovhcloud.com/wp-content/uploads/2026/01/image-11-2048x1081.png 2048w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p>To go further and add even more metrics, you can refer to the vLLM documentation on &#8216;<a href="https://docs.vllm.ai/en/v0.7.2/getting_started/examples/prometheus_grafana.html" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">Prometheus and Grafana</a>&#8216;.</p>



<h2 class="wp-block-heading">Conclusion</h2>



<p>This reference architecture provides a scalable, and production-ready approach for deploying LLM inference on OVHcloud using <strong>AI Deploy</strong> and the <a href="https://help.ovhcloud.com/csm/en-gb-public-cloud-ai-deploy-apps-deployments?id=kb_article_view&amp;sysparm_article=KB0047997#advanced-custom-metrics-for-autoscaling" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">autoscaling on custom metric feature</a>.</p>



<p>OVHcloud <strong>MKS</strong> is dedicated to running Prometheus and Grafana, enabling secure scraping and visualisation of <strong>vLLM internal metrics</strong> exposed via the <strong><mark class="has-inline-color has-ast-global-color-0-color"><code>/metrics</code> </mark></strong>endpoint.</p>



<p>By scraping vLLM metrics securely from AI Deploy into Prometheus and exposing them through Grafana, the architecture provides full visibility into model behaviour, performance and load, enabling informed scaling analysis, troubleshooting and capacity planning in production environments.</p>
<img loading="lazy" decoding="async" src="//blog.ovhcloud.com/wp-content/plugins/matomo/app/matomo.php?idsite=1&amp;rec=1&amp;url=https%3A%2F%2Fblog.ovhcloud.com%2Freference-architecture-custom-metric-autoscaling-for-llm-inference-with-vllm-on-ovhcloud-ai-deploy-and-observability-using-mks%2F&amp;action_name=Reference%20Architecture%3A%20Custom%20metric%20autoscaling%20for%20LLM%20inference%20with%20vLLM%20on%20OVHcloud%20AI%20Deploy%20and%20observability%20using%20MKS&amp;urlref=https%3A%2F%2Fblog.ovhcloud.com%2Ffeed%2F" style="border:0;width:0;height:0" width="0" height="0" alt="" />]]></content:encoded>
					
		
		
			</item>
		<item>
		<title>Reference Architecture: build a sovereign n8n RAG workflow for AI agent using OVHcloud Public Cloud solutions</title>
		<link>https://blog.ovhcloud.com/reference-architecture-build-a-sovereign-n8n-rag-workflow-for-ai-agent-using-ovhcloud-public-cloud-solutions/</link>
		
		<dc:creator><![CDATA[Eléa Petton]]></dc:creator>
		<pubDate>Tue, 27 Jan 2026 13:12:03 +0000</pubDate>
				<category><![CDATA[OVHcloud Engineering]]></category>
		<category><![CDATA[AI]]></category>
		<category><![CDATA[AI Deploy]]></category>
		<category><![CDATA[AI Endpoints]]></category>
		<category><![CDATA[LLM]]></category>
		<category><![CDATA[Managed Database]]></category>
		<category><![CDATA[n8n]]></category>
		<category><![CDATA[Object Storage]]></category>
		<category><![CDATA[Open Source]]></category>
		<category><![CDATA[OVHcloud]]></category>
		<category><![CDATA[Public Cloud]]></category>
		<category><![CDATA[RAG]]></category>
		<category><![CDATA[S3]]></category>
		<guid isPermaLink="false">https://blog.ovhcloud.com/?p=29694</guid>

					<description><![CDATA[What if an n8n workflow, deployed in a&#160;sovereign environment, saved you time while giving you peace of mind? From document ingestion to targeted response generation, n8n acts as the conductor of your RAG pipeline without compromising data protection. In the current landscape of AI agents and knowledge assistants, connecting your internal documentation with&#160;Large Language Models&#160;(LLMs) [&#8230;]<img src="//blog.ovhcloud.com/wp-content/plugins/matomo/app/matomo.php?idsite=1&amp;rec=1&amp;url=https%3A%2F%2Fblog.ovhcloud.com%2Freference-architecture-build-a-sovereign-n8n-rag-workflow-for-ai-agent-using-ovhcloud-public-cloud-solutions%2F&amp;action_name=Reference%20Architecture%3A%20build%20a%20sovereign%20n8n%20RAG%20workflow%20for%20AI%20agent%20using%20OVHcloud%20Public%20Cloud%20solutions&amp;urlref=https%3A%2F%2Fblog.ovhcloud.com%2Ffeed%2F" style="border:0;width:0;height:0" width="0" height="0" alt="" />]]></description>
										<content:encoded><![CDATA[
<p><em><em>What if an n8n workflow, deployed in a&nbsp;</em><strong><em>sovereign environment</em></strong><em>, saved you time while giving you peace of mind? From document ingestion to targeted response generation, n8n acts as the conductor of your RAG pipeline without compromising data protection.</em></em></p>



<figure class="wp-block-image aligncenter size-large"><img loading="lazy" decoding="async" width="1024" height="576" src="https://blog.ovhcloud.com/wp-content/uploads/2025/11/ref-archi-n8n-rag-1024x576.jpg" alt="" class="wp-image-30002" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/11/ref-archi-n8n-rag-1024x576.jpg 1024w, https://blog.ovhcloud.com/wp-content/uploads/2025/11/ref-archi-n8n-rag-300x169.jpg 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/11/ref-archi-n8n-rag-768x432.jpg 768w, https://blog.ovhcloud.com/wp-content/uploads/2025/11/ref-archi-n8n-rag-1536x864.jpg 1536w, https://blog.ovhcloud.com/wp-content/uploads/2025/11/ref-archi-n8n-rag.jpg 1920w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /><figcaption class="wp-element-caption"><em>n8n workflow overview</em></figcaption></figure>



<p>In the current landscape of AI agents and knowledge assistants, connecting your internal documentation with&nbsp;<strong>Large Language Models</strong>&nbsp;(LLMs) is becoming a strategic differentiator.</p>



<p><strong>How?</strong>&nbsp;By building&nbsp;<strong>Agentic RAG systems</strong>&nbsp;capable of retrieving, reasoning, and acting autonomously based on external knowledge.</p>



<p>To make this possible, engineers need a way to connect&nbsp;<strong>retrieval pipelines (RAG)</strong>&nbsp;with&nbsp;<strong>tool-based orchestration</strong>.</p>



<p>This article outlines a&nbsp;<strong>reference architecture</strong>&nbsp;for building a&nbsp;<strong>fully automated RAG pipeline orchestrated by n8n</strong>, leveraging&nbsp;<strong>OVHcloud AI Endpoints</strong>&nbsp;and&nbsp;<strong>PostgreSQL with pgvector</strong>&nbsp;as core components.</p>



<p>The final result will be a system that automatically ingests Markdown documentation from&nbsp;<strong>Object Storage</strong>, creates embeddings with OVHcloud’s&nbsp;<strong>BGE-M3</strong>&nbsp;model available on AI Endpoints, and stores them in a&nbsp;<strong>Managed Database PostgreSQL</strong>&nbsp;with pgvector extension.</p>



<p>Lastly, you’ll be able to build an AI Agent that lets you chat with an LLM (<strong>GPT-OSS-120B</strong>&nbsp;on AI Endpoints). This agent, utilising the RAG implementation carried out upstream, will be an expert on OVHcloud products.</p>



<p>You can further improve the process by using an&nbsp;<strong>LLM guard</strong>&nbsp;to protect the questions sent to the LLM, and set up a chat memory to use conversation history for higher response quality.</p>



<p><strong>But what about n8n?</strong></p>



<p><a href="https://n8n.io/" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer"><strong>n8n</strong></a>, the open-source workflow automation tool,&nbsp;offers many benefits and connects seamlessly with over&nbsp;<strong>300</strong>&nbsp;APIs, apps, and services:</p>



<ul class="wp-block-list">
<li><strong>Open-source</strong>: n8n is a 100% self-hostable solution, which means you retain full data control;</li>



<li><strong>Flexible</strong>: combines low-code nodes and custom JavaScript/Python logic;</li>



<li><strong>AI-ready</strong>: includes useful integrations for LangChain, OpenAI, and embedding support capabilities;</li>



<li><strong>Composable</strong>: enables simple connections between data, APIs, and models in minutes;</li>



<li><strong>Sovereign by design</strong>: compliant with privacy-sensitive or regulated sectors.</li>
</ul>



<p>This reference architecture serves as a blueprint for building a sovereign, scalable Retrieval Augmented Generation (<strong>RAG</strong>) platform using&nbsp;<strong>n8n</strong>&nbsp;and&nbsp;<strong>OVHcloud Public Cloud</strong>&nbsp;solutions.</p>



<p>This setup shows how to orchestrate data ingestion, generate embedding, and enable conversational AI by combining&nbsp;<strong>OVHcloud Object Storage</strong>,&nbsp;<strong>Managed Databases with PostgreSQL</strong>,&nbsp;<strong>AI Endpoints</strong>&nbsp;and&nbsp;<strong>AI Deploy</strong>.<strong>The result?</strong>&nbsp;An AI environment that is fully integrated, protects privacy, and is exclusively hosted on <strong>OVHcloud’s European infrastructure</strong>.</p>



<h2 class="wp-block-heading">Overview of the n8n workflow architecture for RAG </h2>



<p>The workflow involves the following steps:</p>



<ul class="wp-block-list">
<li><strong>Ingestion:</strong>&nbsp;documentation in markdown format is fetched from <strong>OVHcloud Object Storage (S3);</strong></li>



<li><strong>Preprocessing:</strong> n8n cleans and normalises the text, removing YAML front-matter and encoding noise;</li>



<li><strong>Vectorisation:</strong>&nbsp;Each document is embedded using the <strong>BGE-M3</strong> model, which is available via <strong>OVHcloud AI Endpoints;</strong></li>



<li><strong>Persistence:</strong> vectors and metadata are stored in <strong>OVHcloud PostgreSQL Managed Database</strong> using pgvector;</li>



<li><strong>Retrieval:</strong> when a user sends a query, n8n triggers a <strong>LangChain Agent</strong> that retrieves relevant chunks from the database;</li>



<li><strong>Reasoning and actions:</strong>&nbsp;The <strong>AI Agent node</strong> combines LLM reasoning, memory, and tool usage to generate a contextual response or trigger downstream actions (Slack reply, Notion update, API call, etc.).</li>
</ul>



<p>In this tutorial, all services are deployed within the <strong>OVHcloud Public Cloud</strong>.</p>



<h2 class="wp-block-heading">Prerequisites</h2>



<p>Before you start, double-check that you have:</p>



<ul class="wp-block-list">
<li>an <strong>OVHcloud Public Cloud</strong> account</li>



<li>an <strong>OpenStack user</strong> with the <a href="https://help.ovhcloud.com/csm/en-gb-public-cloud-ai-users?id=kb_article_view&amp;sysparm_article=KB0048170" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">&nbsp;following roles</a>:
<ul class="wp-block-list">
<li>Administrator</li>



<li>AI Operator</li>



<li>Object Storage Operator</li>
</ul>
</li>



<li>An <strong>API key</strong> for <a href="https://help.ovhcloud.com/csm/en-gb-public-cloud-ai-endpoints-getting-started?id=kb_article_view&amp;sysparm_article=KB0065401" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">AI Endpoints</a></li>



<li><strong>ovhai CLI available</strong> – <em>install the </em><a href="https://help.ovhcloud.com/csm/en-gb-public-cloud-ai-cli-install-client?id=kb_article_view&amp;sysparm_article=KB0047844" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer"><em>ovhai CLI</em></a></li>



<li><strong>Hugging Face access</strong> – <em>create a </em><a href="https://huggingface.co/" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer"><em>Hugging Face account</em></a><em> and generate an </em><a href="https://huggingface.co/settings/tokens" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer"><em>access token</em></a></li>
</ul>



<p><strong>🚀 Now that you have everything you need, you can start building your n8n workflow!</strong></p>



<h2 class="wp-block-heading">Architecture guide: n8n agentic RAG workflow</h2>



<p>You’re all set to configure and deploy your n8n workflow</p>



<p>⚙️<em> Keep in mind that the following steps can be completed using OVHcloud APIs!</em></p>



<h3 class="wp-block-heading">Step 1 &#8211; Build the RAG data ingestion pipeline</h3>



<p>This first step involves building the foundation of the entire RAG workflow by preparing the elements you need:</p>



<ul class="wp-block-list">
<li>n8n deployment</li>



<li>Object Storage bucket creation</li>



<li>PostgreSQL database creation</li>



<li>and more</li>
</ul>



<p>Remember to set up the proper credentials in n8n so the different elements can connect and function.</p>



<h4 class="wp-block-heading">1. Deploy n8n on OVHcloud VPS</h4>



<p>OVHcloud provides <a href="https://www.ovhcloud.com/en-gb/vps/vps-n8n/" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer"><strong>VPS solutions compatible with n8n</strong></a><strong>.</strong> Get a ready-to-use virtual server with <strong>pre-installed n8n </strong>and start building automation workflows without manual setup. With plans ranging from <strong>6 vCores&nbsp;/&nbsp;12 GB RAM</strong> to <strong>24 vCores&nbsp;/&nbsp;96 GB RAM</strong>, you can choose the capacity that suits your workload.</p>



<p><strong>How to set up n8n on a VPS?</strong></p>



<p>Setting up n8n on an OVHcloud VPS generally involves:</p>



<ul class="wp-block-list">
<li>Choosing and provisioning your OVHcloud VPS plan;</li>



<li>Connecting to your server via SSH and carrying out the initial server configuration, which includes updating the OS;</li>



<li>Installing n8n, typically with Docker (recommended for ease of management and updates), or npm by following this <a href="https://help.ovhcloud.com/csm/en-gb-vps-install-n8n?id=kb_article_view&amp;sysparm_article=KB0072179" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">guide</a>;</li>



<li>Configuring n8n with a domain name, SSL certificate for HTTPS, and any necessary environment variables for databases or settings.</li>
</ul>



<p>While OVHcloud provides a robust VPS platform, you can find detailed n8n installation guides in the <a href="https://docs.n8n.io/hosting/installation/docker/" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">official n8n documentation</a>.</p>



<p>Once the configuration is complete, you can configure the database and bucket in Object Storage.</p>



<h4 class="wp-block-heading">2. Create Object Storage bucket</h4>



<p>First, you have to set up your data source. Here you can store all your documentation in an S3-compatible <a href="https://www.ovhcloud.com/en-gb/public-cloud/object-storage/" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">Object Storage</a> bucket.</p>



<p>Here, assume that all the documentation files are in Markdown format.</p>



<p>From <strong>OVHcloud Control Panel</strong>, create a new Object Storage container with <strong>S3-compatible API </strong>solution; follow this <a href="https://help.ovhcloud.com/csm/en-gb-public-cloud-storage-s3-getting-started-object-storage?id=kb_article_view&amp;sysparm_article=KB0034674" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">guide</a>.</p>



<p>When the bucket is ready, add your Markdown documentation to it.</p>



<figure class="wp-block-image size-large"><img loading="lazy" decoding="async" width="1024" height="580" src="https://blog.ovhcloud.com/wp-content/uploads/2025/10/ref-archi-n8n-1024x580.png" alt="" class="wp-image-29733" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/10/ref-archi-n8n-1024x580.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/ref-archi-n8n-300x170.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/ref-archi-n8n-768x435.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/ref-archi-n8n-1536x870.png 1536w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/ref-archi-n8n-2048x1160.png 2048w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p><strong>Note:</strong>&nbsp;For this tutorial, we’re using the various OVHcloud product documentation available in Open-Source on the GitHub repository maintained by OVHcloud members.</p>



<p><em>Click this </em><a href="https://github.com/ovh/docs.git" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer"><em>link</em></a><em> to access the repository.</em></p>
</blockquote>
</blockquote>



<p>How do you do that? Extract all the <a href="http://guide.en-gb.md" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer"><strong>guide.en-gb.md</strong></a> files from the GitHub repository and rename each one to match its parent folder.</p>



<p>Example: the documentation about ovhai cli installation <code><strong>docs/pages/public_cloud/ai_machine_learning/cli_10_howto_install_cli/</strong></code><a href="http://guide.en-gb.md" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer"><strong>guide.en-gb.md</strong></a> is stored in <strong>ovhcloud-products-documentation-md</strong> bucket as <a href="http://cli_10_howto_install_cli.md" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer"><strong>cli_10_howto_install_cli.md</strong></a></p>



<p>You should get an overview that looks like this:</p>



<figure class="wp-block-image size-large"><img loading="lazy" decoding="async" width="1024" height="580" src="https://blog.ovhcloud.com/wp-content/uploads/2025/10/ref-archi-n8n-1-1024x580.png" alt="" class="wp-image-29735" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/10/ref-archi-n8n-1-1024x580.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/ref-archi-n8n-1-300x170.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/ref-archi-n8n-1-768x435.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/ref-archi-n8n-1-1536x870.png 1536w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/ref-archi-n8n-1-2048x1160.png 2048w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p>Keep the following elements and create a new credential in n8n named <code><strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">OVHcloud S3 gra credentials</mark></strong></code>:</p>



<ul class="wp-block-list">
<li>S3 Endpoint: <a href="https://s3.gra.io.cloud.ovh.net/" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer"><strong><code><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">https://s3.gra.io.cloud.ovh.net/</mark></code></strong></a></li>



<li>Region: <strong><code><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">gra</mark></code></strong></li>



<li>Access Key ID: <strong><code>&lt;your_object_storage_user_access_key&gt;</code></strong></li>



<li>Secret Access Key: <strong><code>&lt;your_pbject_storage_user_secret_key&gt;</code></strong></li>
</ul>



<figure class="wp-block-image size-large"><img loading="lazy" decoding="async" width="1024" height="580" src="https://blog.ovhcloud.com/wp-content/uploads/2025/10/ref-archi-n8n-2-1024x580.png" alt="" class="wp-image-29736" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/10/ref-archi-n8n-2-1024x580.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/ref-archi-n8n-2-300x170.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/ref-archi-n8n-2-768x435.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/ref-archi-n8n-2-1536x870.png 1536w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/ref-archi-n8n-2-2048x1160.png 2048w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p>Then, create a new n8n node by selecting&nbsp;<strong>S3</strong>, then&nbsp;<strong>Get Multiple Files</strong>.<br>Configure this node as follows:</p>



<figure class="wp-block-image size-large"><img loading="lazy" decoding="async" width="1024" height="580" src="https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-15-a-16.20.47-1024x580.png" alt="" class="wp-image-29740" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-15-a-16.20.47-1024x580.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-15-a-16.20.47-300x170.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-15-a-16.20.47-768x435.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-15-a-16.20.47-1536x870.png 1536w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-15-a-16.20.47-2048x1160.png 2048w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p>Connect the node to the previous one before moving on to the next step.</p>



<figure class="wp-block-image size-large"><img loading="lazy" decoding="async" width="1024" height="580" src="https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-15-a-16.18.00-1024x580.png" alt="" class="wp-image-29741" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-15-a-16.18.00-1024x580.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-15-a-16.18.00-300x170.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-15-a-16.18.00-768x435.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-15-a-16.18.00-1536x870.png 1536w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-15-a-16.18.00-2048x1160.png 2048w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p>With the first phase done, you can now configure the vector DB.</p>



<h4 class="wp-block-heading">3. Configure PostgreSQL Managed DB (pgvector)</h4>



<p>In this step, you can set up the vector database that lets you store the embeddings generated from your documents.</p>



<p>How? By using OVHcloud’s managed databases, a pgvector extension of&nbsp;<a href="https://www.ovhcloud.com/en-gb/public-cloud/postgresql/" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">PostgreSQL</a>. Go to your OVHcloud Control Panel and follow the steps.</p>



<p>1. Navigate to&nbsp;<strong>Databases &amp; Analytics &gt; Databases</strong></p>



<p><strong>2. Create a new database and select&nbsp;<em>PostgreSQL</em>&nbsp;and a datacenter location</strong></p>



<figure class="wp-block-image size-large"><img loading="lazy" decoding="async" width="1024" height="580" src="https://blog.ovhcloud.com/wp-content/uploads/2025/10/4-1024x580.png" alt="" class="wp-image-29758" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/10/4-1024x580.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/4-300x170.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/4-768x435.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/4-1536x870.png 1536w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/4-2048x1160.png 2048w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p><strong>3. Select&nbsp;<em>Production</em>&nbsp;plan and&nbsp;<em>Instance type</em></strong></p>



<figure class="wp-block-image size-large"><img loading="lazy" decoding="async" width="1024" height="580" src="https://blog.ovhcloud.com/wp-content/uploads/2025/10/5-1024x580.png" alt="" class="wp-image-29759" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/10/5-1024x580.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/5-300x170.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/5-768x435.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/5-1536x870.png 1536w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/5-2048x1160.png 2048w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p><strong>4. Reset the user password and save it</strong></p>



<figure class="wp-block-image size-large"><img loading="lazy" decoding="async" width="1024" height="580" src="https://blog.ovhcloud.com/wp-content/uploads/2025/10/ref-archi-n8n-1-1-1024x580.png" alt="" class="wp-image-29762" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/10/ref-archi-n8n-1-1-1024x580.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/ref-archi-n8n-1-1-300x170.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/ref-archi-n8n-1-1-768x435.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/ref-archi-n8n-1-1-1536x870.png 1536w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/ref-archi-n8n-1-1-2048x1160.png 2048w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p><strong>5. Whitelist the IP of your n8n instance as follows</strong></p>



<figure class="wp-block-image size-large"><img loading="lazy" decoding="async" width="1024" height="580" src="https://blog.ovhcloud.com/wp-content/uploads/2025/10/7-1024x580.png" alt="" class="wp-image-29761" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/10/7-1024x580.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/7-300x170.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/7-768x435.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/7-1536x870.png 1536w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/7-2048x1160.png 2048w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p><strong>6. Take note of te following parameters</strong></p>



<figure class="wp-block-image size-large"><img loading="lazy" decoding="async" width="1024" height="580" src="https://blog.ovhcloud.com/wp-content/uploads/2025/10/6-1024x580.png" alt="" class="wp-image-29760" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/10/6-1024x580.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/6-300x170.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/6-768x435.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/6-1536x870.png 1536w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/6-2048x1160.png 2048w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p>Make a note of this information and create a new credential in n8n named&nbsp;<strong>OVHcloud PGvector credentials</strong>:</p>



<ul class="wp-block-list">
<li>Host:<strong>&nbsp;<code><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">&lt;db_hostname&gt;</mark></code></strong></li>



<li>Database:&nbsp;<strong>defaultdb</strong></li>



<li>User:&nbsp;<code>avnadmin</code></li>



<li>Password:&nbsp;<code><strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">&lt;db_password&gt;</mark></strong></code></li>



<li>Port:&nbsp;<strong>20184</strong></li>
</ul>



<p>Consider&nbsp;<code>enabling</code>&nbsp;the&nbsp;<strong>Ignore SSL Issues (Insecure)</strong>&nbsp;button as needed and setting the&nbsp;<strong>Maximum Number of Connections</strong>&nbsp;value to&nbsp;<strong><code><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">1000</mark></code></strong>.</p>



<figure class="wp-block-image"><img loading="lazy" decoding="async" width="1024" height="580" src="https://blog.ovhcloud.com/wp-content/uploads/2025/10/8-1024x580.png" alt="" class="wp-image-29763" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/10/8-1024x580.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/8-300x170.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/8-768x435.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/8-1536x870.png 1536w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/8-2048x1160.png 2048w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p>✅ You’re now connected to the database! But what about the PGvector extension?</p>



<p>Add a PosgreSQL node in your n8n workflow&nbsp;<code><strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">Execute a SQL query</mark></strong></code>,&nbsp;and create the extension through an SQL query, which should look like this:</p>



<pre class="wp-block-code"><code class="">-- drop table as needed<br>DROP TABLE IF EXISTS md_embeddings;<br><br>-- activate pgvector<br>CREATE EXTENSION IF NOT EXISTS vector;<br><br>-- create table<br>CREATE TABLE md_embeddings (<br>    id SERIAL PRIMARY KEY,<br>    text TEXT,<br>    embedding vector(1024),<br>    metadata JSONB<br>);</code></pre>



<p>You should get this n8n node:</p>



<figure class="wp-block-image size-large"><img loading="lazy" decoding="async" width="1024" height="580" src="https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-16-a-14.43.39-1024x580.png" alt="" class="wp-image-29752" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-16-a-14.43.39-1024x580.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-16-a-14.43.39-300x170.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-16-a-14.43.39-768x435.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-16-a-14.43.39-1536x870.png 1536w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-16-a-14.43.39-2048x1160.png 2048w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p>Finally, you can create a new table and name it&nbsp;<code><strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">md_embeddings</mark></strong></code>&nbsp;using this node. Create a&nbsp;<code><strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">Stop and Error</mark></strong></code>&nbsp;node if you run into errors setting up the table.</p>



<figure class="wp-block-image size-large"><img loading="lazy" decoding="async" width="1024" height="580" src="https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-16-a-14.51.45-1024x580.png" alt="" class="wp-image-29753" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-16-a-14.51.45-1024x580.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-16-a-14.51.45-300x170.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-16-a-14.51.45-768x435.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-16-a-14.51.45-1536x870.png 1536w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-16-a-14.51.45-2048x1160.png 2048w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p>All set! Your vector DB is prepped and ready for data! Keep in mind, you still need an&nbsp;<strong>embeddings model</strong> for the RAG data ingestion pipeline.</p>



<h4 class="wp-block-heading">4. Access to OVHcloud AI Endpoints</h4>



<p><strong>OVHcloud AI Endpoints</strong>&nbsp;is a managed service that provides&nbsp;<strong>ready-to-use APIs for AI models</strong>, including&nbsp;<strong>LLM, CodeLLM, embeddings, Speech-to-Text, and image models</strong>&nbsp;hosted within OVHcloud’s European infrastructure.</p>



<p>To vectorise the various documents in Markdown format, you have to select an embedding model:&nbsp;<a href="https://endpoints.ai.cloud.ovh.net/models/bge-m3" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer"><strong>BGE-M3</strong></a>.</p>



<p>Usually, your AI Endpoints API key should already be created. If not, head to the AI Endpoints menu in your OVHcloud Control Panel to generate a new API key.</p>



<figure class="wp-block-image size-large"><img loading="lazy" decoding="async" width="1024" height="580" src="https://blog.ovhcloud.com/wp-content/uploads/2025/10/ref-archi-n8n-3-1-1024x580.png" alt="" class="wp-image-29775" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/10/ref-archi-n8n-3-1-1024x580.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/ref-archi-n8n-3-1-300x170.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/ref-archi-n8n-3-1-768x435.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/ref-archi-n8n-3-1-1536x870.png 1536w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/ref-archi-n8n-3-1-2048x1160.png 2048w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p>Once this is done, you can create new OpenAI credentials in your n8n.</p>



<p>Why do I need OpenAI credentials? Because <strong>AI Endpoints API&nbsp;</strong>is fully compatible with OpenAI’s, integrating it is simple and ensures the&nbsp;<strong>sovereignty of your data.</strong></p>



<p>How? Thanks to a single endpoint&nbsp;<a href="https://oai.endpoints.kepler.ai.cloud.ovh.net/v1" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer"><strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color"><code>https://oai.endpoints.kepler.ai.cloud.ovh.net/v1</code></mark></strong></a>, you can request the different AI Endpoints models.</p>



<figure class="wp-block-image size-large"><img loading="lazy" decoding="async" width="1024" height="580" src="https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-17-a-11.45.33-1024x580.png" alt="" class="wp-image-29776" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-17-a-11.45.33-1024x580.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-17-a-11.45.33-300x170.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-17-a-11.45.33-768x435.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-17-a-11.45.33-1536x870.png 1536w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-17-a-11.45.33-2048x1160.png 2048w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p>This means you can create a new n8n node by selecting&nbsp;<strong>Postgres PGVector Store</strong>&nbsp;and&nbsp;<strong>Add documents to Vector Store</strong>.<br>Set up this node as shown below:</p>



<figure class="wp-block-image size-large"><img loading="lazy" decoding="async" width="1024" height="580" src="https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-17-a-11.50.24-1024x580.png" alt="" class="wp-image-29781" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-17-a-11.50.24-1024x580.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-17-a-11.50.24-300x170.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-17-a-11.50.24-768x435.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-17-a-11.50.24-1536x870.png 1536w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-17-a-11.50.24-2048x1160.png 2048w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p>Then configure the <strong>Data Loader</strong> with a custom text splitting and a JSON type.</p>



<figure class="wp-block-image size-large"><img loading="lazy" decoding="async" width="1024" height="580" src="https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-17-a-11.50.38-1-1024x580.png" alt="" class="wp-image-29780" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-17-a-11.50.38-1-1024x580.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-17-a-11.50.38-1-300x170.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-17-a-11.50.38-1-768x435.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-17-a-11.50.38-1-1536x870.png 1536w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-17-a-11.50.38-1-2048x1160.png 2048w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p>For the text splitter, here are some options:</p>



<figure class="wp-block-image size-large"><img loading="lazy" decoding="async" width="1024" height="580" src="https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-17-a-12.02.43-1024x580.png" alt="" class="wp-image-29786" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-17-a-12.02.43-1024x580.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-17-a-12.02.43-300x170.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-17-a-12.02.43-768x435.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-17-a-12.02.43-1536x870.png 1536w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-17-a-12.02.43-2048x1160.png 2048w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p>To finish, select the&nbsp;<strong>BGE-M3</strong> embedding model from the model list and set the&nbsp;<strong>Dimensions</strong> to 1024.</p>



<figure class="wp-block-image size-large"><img loading="lazy" decoding="async" width="1024" height="580" src="https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-17-a-11.50.51-1024x580.png" alt="" class="wp-image-29784" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-17-a-11.50.51-1024x580.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-17-a-11.50.51-300x170.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-17-a-11.50.51-768x435.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-17-a-11.50.51-1536x870.png 1536w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-17-a-11.50.51-2048x1160.png 2048w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p>You now have everything you need to build the ingestion pipeline.</p>



<h4 class="wp-block-heading">5. Set up the ingestion pipeline loop</h4>



<p>To make use of a fully automated document ingestion and vectorisation pipeline, you have to integrate some specific nodes, mainly:</p>



<ul class="wp-block-list">
<li>a <strong><code><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">Loop Over Items</mark></code></strong> that downloads each markdown file one by one so that it can be vectorised;</li>



<li>a <code><strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">Code in JavaScript</mark></strong></code> that counts the number of files processed, which subsequently determines the number of requests sent to the embedding model;</li>



<li>an <code><strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">If</mark></strong></code> condition that allows you to check when the 400 requests have been reached;</li>



<li>a <code><strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">Wait</mark></strong></code> node that pauses after every 400 requests to avoid getting rate-limited;</li>



<li>an S3 block <code><strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">Download a file</mark></strong></code> to download each markdown;</li>



<li>another <code><strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">Code in JavaScript</mark></strong></code> to extract and process text from Markdown files by cleaning and removing special characters before sending it to the embeddings model;</li>



<li>a PostgreSQL node to <code><strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">Execute a SQL</mark></strong></code> query to check that the table contains vectors after the process (loop) is complete.</li>
</ul>



<h5 class="wp-block-heading">5.1. Create a loop to process each documentation file</h5>



<p>Begin by creating a <code><strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">Loop Over Items</mark></strong></code> to process all the Markdown files one at a time. Set the <strong>batch size</strong> to <strong><code><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">1</mark></code></strong> in this loop.</p>



<figure class="wp-block-image size-large"><img loading="lazy" decoding="async" width="1024" height="580" src="https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-10.50.13-1024x580.png" alt="" class="wp-image-29788" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-10.50.13-1024x580.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-10.50.13-300x170.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-10.50.13-768x435.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-10.50.13-1536x870.png 1536w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-10.50.13-2048x1160.png 2048w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p>Add the <strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color"><code>Loop</code></mark></strong> statement right after the S3 <strong><code><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">Get Many Files</mark></code></strong> node as shown below:</p>



<figure class="wp-block-image size-large"><img loading="lazy" decoding="async" width="1024" height="580" src="https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-11.30.00-1024x580.png" alt="" class="wp-image-29797" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-11.30.00-1024x580.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-11.30.00-300x170.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-11.30.00-768x435.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-11.30.00-1536x870.png 1536w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-11.30.00-2048x1160.png 2048w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p>Time to put the loop’s content into action!</p>



<h5 class="wp-block-heading">5.2. Count the number of files using a code snippet</h5>



<p>Next, choose the <code><strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">Code in JavaScript</mark></strong></code> node from the list to see how many files have been processed. Set “Run Once for Each Item” <code><strong>Mode</strong></code> and “JavaScript” code <strong>Language</strong>, then add the following code snippet to the designated block.</p>



<pre class="wp-block-code"><code class="">// simple counter per item<br>const counter = $runIndex + 1;<br><br>return {<br>  counter<br>};</code></pre>



<figure class="wp-block-image size-large"><img loading="lazy" decoding="async" width="1024" height="580" src="https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-11.05.47-1024x580.png" alt="" class="wp-image-29792" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-11.05.47-1024x580.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-11.05.47-300x170.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-11.05.47-768x435.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-11.05.47-1536x870.png 1536w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-11.05.47-2048x1160.png 2048w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p>Make sure this code snippet is included in the loop.</p>



<figure class="wp-block-image size-large"><img loading="lazy" decoding="async" width="1024" height="580" src="https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-11.33.57-1024x580.png" alt="" class="wp-image-29798" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-11.33.57-1024x580.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-11.33.57-300x170.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-11.33.57-768x435.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-11.33.57-1536x870.png 1536w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-11.33.57-2048x1160.png 2048w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p>You can start adding the <mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color"><strong><code>if</code></strong></mark> part to the loop now.</p>



<h5 class="wp-block-heading">5.3. Add a condition that applies a rule every 400 requests</h5>



<p>Here, you need to create an <code><strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">If</mark></strong></code> node and add the following condition, which you have set as an expression.</p>



<pre class="wp-block-code"><code class="">{{ (Number($json["counter"]) % 400) === 0 }}</code></pre>



<figure class="wp-block-image size-large"><img loading="lazy" decoding="async" width="1024" height="580" src="https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-11.11.42-1024x580.png" alt="" class="wp-image-29794" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-11.11.42-1024x580.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-11.11.42-300x170.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-11.11.42-768x435.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-11.11.42-1536x870.png 1536w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-11.11.42-2048x1160.png 2048w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p>Add it immediately after counting the files:</p>



<figure class="wp-block-image size-large"><img loading="lazy" decoding="async" width="1024" height="580" src="https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-11.44.10-1024x580.png" alt="" class="wp-image-29800" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-11.44.10-1024x580.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-11.44.10-300x170.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-11.44.10-768x435.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-11.44.10-1536x870.png 1536w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-11.44.10-2048x1160.png 2048w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p>If this condition <code><strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">is true</mark></strong></code>, trigger the <code><strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">Wait</mark></strong></code> node.</p>



<h5 class="wp-block-heading">5.4. Insert a pause after each set of 400 requests</h5>



<p>Then insert a <code><strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">Wait</mark></strong></code> node to pause for a few seconds before resuming. You can insert <code><strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">Resume</mark></strong></code> “After Time Interval” and set the <code><strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">Wait Amount</mark></strong></code> to “60:00” seconds.</p>



<figure class="wp-block-image size-large"><img loading="lazy" decoding="async" width="1024" height="580" src="https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-11.23.39-1024x580.png" alt="" class="wp-image-29796" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-11.23.39-1024x580.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-11.23.39-300x170.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-11.23.39-768x435.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-11.23.39-1536x870.png 1536w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-11.23.39-2048x1160.png 2048w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p>Link it to the <code><strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">If</mark></strong></code> condition when this is <strong>True</strong>.</p>



<figure class="wp-block-image size-large"><img loading="lazy" decoding="async" width="1024" height="580" src="https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-11.45.08-1024x580.png" alt="" class="wp-image-29801" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-11.45.08-1024x580.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-11.45.08-300x170.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-11.45.08-768x435.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-11.45.08-1536x870.png 1536w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-11.45.08-2048x1160.png 2048w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p>Next, you can go ahead and download the Markdown file, and then process it.</p>



<h5 class="wp-block-heading">5.5. Launch documentation download</h5>



<p>To do this, create a new <code><strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">Download a file</mark></strong></code> S3 node and configure it with this File Key expression:</p>



<pre class="wp-block-code"><code class="">{{ $('Process each documentation file').item.json.Key }}</code></pre>



<figure class="wp-block-image size-large"><img loading="lazy" decoding="async" width="1024" height="580" src="https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-16.42.12-1024x580.png" alt="" class="wp-image-29804" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-16.42.12-1024x580.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-16.42.12-300x170.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-16.42.12-768x435.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-16.42.12-1536x870.png 1536w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-16.42.12-2048x1160.png 2048w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p>Want to connect it?  That’s easy, link it to the output of the <code><strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">Wait</mark></strong></code> and <code><strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">If</mark></strong></code> statements when the ‘if’ statement returns <strong>False</strong>; this will allow the file to be processed only if the rate limit is not exceeded.</p>



<figure class="wp-block-image size-large"><img loading="lazy" decoding="async" width="1024" height="580" src="https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-16.49.05-1024x580.png" alt="" class="wp-image-29805" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-16.49.05-1024x580.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-16.49.05-300x170.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-16.49.05-768x435.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-16.49.05-1536x870.png 1536w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-16.49.05-2048x1160.png 2048w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p>You’re almost done! Now you need to extract and process the text from the Markdown files – clean and remove any special characters before sending it to the embedding model.</p>



<h5 class="wp-block-heading">5.6 Clean Markdown text content</h5>



<p>Next, create another <code><strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">Code in JavaScript</mark></strong></code> to process text from Markdown files:</p>



<pre class="wp-block-code"><code class="">// extract binary content<br>const binary = $input.item.binary.data;<br><br>// decoding into clean UTF-8 text<br>let text = Buffer.from(binary.data, 'base64').toString('utf8');<br><br>// cleaning - remove non-printable characters<br>text = text<br>  .replace(/[^\x09\x0A\x0D\x20-\x7EÀ-ÿ€£¥•–—‘’“”«»©®™°±§¶÷×]/g, ' ')<br>  .replace(/\s{2,}/g, ' ')<br>  .trim();<br><br>// check lenght<br>if (text.length &gt; 14000) {<br>  text = text.slice(0, 14000);<br>}<br><br>return [{<br>  text,<br>  fileName: binary.fileName,<br>  mimeType: binary.mimeType<br>}];</code></pre>



<p>Select the <em>“Run Once for Each Item”</em> <code><strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">Mode</mark></strong></code> and place the previous code in the dedicated JavaScript block.</p>



<figure class="wp-block-image size-large"><img loading="lazy" decoding="async" width="1024" height="580" src="https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-17.02.04-1024x580.png" alt="" class="wp-image-29806" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-17.02.04-1024x580.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-17.02.04-300x170.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-17.02.04-768x435.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-17.02.04-1536x870.png 1536w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-17.02.04-2048x1160.png 2048w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p>To finish, check that the output text has been sent to the document vectorisation system, which was set up in <strong>Step 3 – Configure PostgreSQL Managed DB (pgvector)</strong>.</p>



<figure class="wp-block-image size-large"><img loading="lazy" decoding="async" width="1024" height="580" src="https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-17.15.45-1024x580.png" alt="" class="wp-image-29808" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-17.15.45-1024x580.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-17.15.45-300x170.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-17.15.45-768x435.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-17.15.45-1536x870.png 1536w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-17.15.45-2048x1160.png 2048w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p>How do I confirm that the table contains all elements after vectorisation?</p>



<h5 class="wp-block-heading">5.7 Double-check that the documents are in the table</h5>



<p>To confirm that your RAG system is working, make sure your vector database has different vectors; use a PostgreSQL node with <code><strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">Execute a SQL query</mark></strong></code> in your n8n workflow.</p>



<p>Then, run the following query:</p>



<pre class="wp-block-code"><code class="">-- count the number of elements<br>SELECT COUNT(*) FROM md_embeddings;</code></pre>



<figure class="wp-block-image size-large"><img loading="lazy" decoding="async" width="1024" height="580" src="https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-20.28.49-1024x580.png" alt="" class="wp-image-29818" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-20.28.49-1024x580.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-20.28.49-300x170.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-20.28.49-768x435.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-20.28.49-1536x870.png 1536w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-20-a-20.28.49-2048x1160.png 2048w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p>Next, link this element to the <code><strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">Done</mark></strong></code> section of your <strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">Loop</mark></strong>, so the elements are counted when the process is complete.</p>



<figure class="wp-block-image size-large"><img loading="lazy" decoding="async" width="1024" height="580" src="https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-17-a-11.14.41-1024x580.png" alt="" class="wp-image-29773" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-17-a-11.14.41-1024x580.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-17-a-11.14.41-300x170.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-17-a-11.14.41-768x435.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-17-a-11.14.41-1536x870.png 1536w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-17-a-11.14.41-2048x1160.png 2048w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p>Congrats! You can now run the workflow to begin ingesting documents.</p>



<p>Click the <code><strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">Execute workflow</mark></strong></code> button and wait until the vectorization process is complete.</p>



<figure class="wp-block-image size-large"><img loading="lazy" decoding="async" width="1024" height="580" src="https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-21-a-11.41.52-1024x580.png" alt="" class="wp-image-29823" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-21-a-11.41.52-1024x580.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-21-a-11.41.52-300x170.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-21-a-11.41.52-768x435.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-21-a-11.41.52-1536x870.png 1536w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-21-a-11.41.52-2048x1160.png 2048w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p>Remember, everything should be green when it’s finished ✅.</p>



<h3 class="wp-block-heading">Step 2 – RAG chatbot</h3>



<p>With the data ingestion and vectorisation steps completed, you can now begin implementing your AI agent.</p>



<p>This involves building a <strong>RAG-based AI Agent</strong>&nbsp;by simply starting a chat with an LLM.</p>



<h4 class="wp-block-heading">1. Set up the chat box to start a conversation</h4>



<p>First, configure your AI Agent based on the RAG system, and add a new node in the same n8n workflow: <code><strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">Chat Trigger</mark></strong></code>.</p>



<figure class="wp-block-image size-large"><img loading="lazy" decoding="async" width="1024" height="580" src="https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-21-a-16.31.24-1024x580.png" alt="" class="wp-image-29834" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-21-a-16.31.24-1024x580.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-21-a-16.31.24-300x170.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-21-a-16.31.24-768x435.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-21-a-16.31.24-1536x870.png 1536w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-21-a-16.31.24-2048x1160.png 2048w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p>This node will allow you to interact directly with your AI agent! But before that, you need to check that your message is safe.</p>



<p>This node will allow you to interact directly with your AI agent! But before that, you need to check that your message is secure.</p>



<h4 class="wp-block-heading">2. Set up your LLM Guard with AI Deploy</h4>



<p>To check whether a message is secure or not, use an LLM Guard.</p>



<p><strong>What’s an LLM Guard?</strong>&nbsp;This is a safety and control layer that sits between users and an LLM, or between the LLM and an external connection. Its main goal is to filter, monitor, and enforce rules on what goes into or comes out of the model 🔐.</p>



<p>You can use <a href="file:///Users/jdutse/Downloads/www.ovhcloud.com/en-gb/public-cloud/ai-deploy" data-wpel-link="internal">AI Deploy</a> from OVHcloud to deploy your desired LLM guard. With a single command line, this AI solution lets you deploy a Hugging Face model using vLLM Docker containers.</p>



<p>For more details, please refer to this <a href="https://blog.ovhcloud.com/mistral-small-24b-served-with-vllm-and-ai-deploy-one-command-to-deploy-llm/" data-wpel-link="internal">blog</a>.</p>



<p>For the use case covered in this article, you can use the open-source model <strong>meta-llama/Llama-Guard-3-8B</strong> available on <a href="https://huggingface.co/meta-llama/Llama-Guard-3-8B" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">Hugging Face</a>.</p>



<h5 class="wp-block-heading">2.1 Create a Bearer token to request your custom AI Deploy endpoint</h5>



<p><a href="https://help.ovhcloud.com/csm/en-gb-public-cloud-ai-cli-app-token?id=kb_article_view&amp;sysparm_article=KB0035280" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">Create a token</a> to access your AI Deploy app once it’s deployed.</p>



<pre class="wp-block-code"><code class="">ovhai token create --role operator ai_deploy_token=my_operator_token</code></pre>



<p>The following output is returned:</p>



<p><code><strong>Id: 47292486-fb98-4a5b-8451-600895597a2b<br>Created At: 20-10-25 8:53:05<br>Updated At: 20-10-25 8:53:05<br>Spec:<br>Name: ai_deploy_token=my_operator_token<br>Role: AiTrainingOperator<br>Label Selector:<br>Status:<br>Value: XXXXXXXXXXXXXXXXXXXXXXXXXXXXXX<br>Version: 1</strong></code></p>



<p>You can now store and export your access token to add it as a new credential in n8n.</p>



<pre class="wp-block-code"><code class="">export MY_OVHAI_ACCESS_TOKEN=XXXXXXXXXXXXXXXXXXXXXXXXXXXXXX</code></pre>



<h5 class="wp-block-heading">2.1 Start Llama Guard 3 model with AI Deploy</h5>



<p>Using <code><strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">ovhai</mark></strong></code> CLI, launch the following command and vLLM start inference server.</p>



<pre class="wp-block-code"><code class="">ovhai app run \<br>	--name vllm-llama-guard3 \<br>        --default-http-port 8000 \<br>        --gpu 1 \<br>	--flavor l40s-1-gpu \<br>        --label ai_deploy_token=my_operator_token \<br>	--env OUTLINES_CACHE_DIR=/tmp/.outlines \<br>	--env HF_TOKEN=$MY_HF_TOKEN \<br>	--env HF_HOME=/hub \<br>	--env HF_DATASETS_TRUST_REMOTE_CODE=1 \<br>	--env HF_HUB_ENABLE_HF_TRANSFER=0 \<br>	--volume standalone:/workspace:RW \<br>	--volume standalone:/hub:RW \<br>	vllm/vllm-openai:v0.10.1.1 \<br>	-- bash -c python3 -m vllm.entrypoints.openai.api_server                       <br>                           --model meta-llama/Llama-Guard-3-8B \                     <br>                           --tensor-parallel-size 1 \                     <br>                           --dtype bfloat16</code></pre>



<p><em>Full command explained:</em></p>



<ul class="wp-block-list">
<li><code><strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">ovhai app run</mark></strong></code></li>
</ul>



<p>This is the core command to&nbsp;<strong>run an app</strong>&nbsp;using the&nbsp;<strong>OVHcloud AI Deploy</strong>&nbsp;platform.</p>



<ul class="wp-block-list">
<li><code><strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">--name vllm-llama-guard3</mark></strong></code></li>
</ul>



<p>Sets a&nbsp;<strong>custom name</strong>&nbsp;for the job. For example,&nbsp;<code>vllm-llama-guard3</code>.</p>



<ul class="wp-block-list">
<li><code><strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">--default-http-port 8000</mark></strong></code></li>
</ul>



<p>Exposes&nbsp;<strong>port 8000</strong>&nbsp;as the default HTTP endpoint. vLLM server typically runs on port 8000.</p>



<ul class="wp-block-list">
<li><strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color"><code>--gpu&nbsp;</code>1</mark></strong></li>



<li><strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color"><code>--flavor l40s-1-gpu</code></mark></strong></li>
</ul>



<p>Allocates&nbsp;<strong>1 GPU L40S</strong>&nbsp;for the app. You can adjust the GPU type and number depending on the model you have to deploy.</p>



<ul class="wp-block-list">
<li><code><strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">--volume standalone:/workspace:RW</mark></strong></code></li>



<li><code><strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">--volume standalone:/hub:RW</mark></strong></code></li>
</ul>



<p>Mounts&nbsp;<strong>two persistent storage volumes</strong>: <strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color"><code>/workspace</code></mark></strong> which is the main working directory and <code><strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">/hub</mark></strong></code>&nbsp;to store Hugging Face model files.</p>



<ul class="wp-block-list">
<li><code><strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">--env OUTLINES_CACHE_DIR=/tmp/.outlines</mark></strong></code></li>



<li><strong><code><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">--env HF_TOKEN=$MY_HF_TOKEN</mark></code></strong></li>



<li><code><strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">--env HF_HOME=/hub</mark></strong></code></li>



<li><code><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color"><strong>--env HF_DATASETS_TRUST_REMOTE_CODE=1</strong></mark></code></li>



<li><code><strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">--env HF_HUB_ENABLE_HF_TRANSFER=0</mark></strong></code></li>
</ul>



<p>These are Hugging Face&nbsp;<strong>environment variables</strong> you have to set. Please export your Hugging Face access token as environment variable before starting the app: <code><strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">export MY_HF_TOKEN=***********</mark></strong></code></p>



<ul class="wp-block-list">
<li><code><strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">vllm/vllm-openai:v0.10.1.1</mark></strong></code></li>
</ul>



<p>Use the&nbsp;<strong><code>v<strong><code>llm/vllm-openai</code></strong></code></strong>&nbsp;Docker image (a pre-configured vLLM OpenAI API server).</p>



<ul class="wp-block-list">
<li><code><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color"><strong>-- bash -c python3 -m vllm.entrypoints.openai.api_server                       <br>                    --model meta-llama/Llama-Guard-3-8B \                     <br>                    --tensor-parallel-size 1 \                     <br>                    --dtype bfloat16</strong></mark></code></li>
</ul>



<p>Finally, run a<strong>&nbsp;bash shell</strong>&nbsp;inside the container and executes a Python command to launch the vLLM API server.</p>



<h5 class="wp-block-heading">2.2 Check to confirm your AI Deploy app is RUNNING</h5>



<p>Replace the <code><strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">&lt;app_id></mark></strong></code> by yours.</p>



<pre class="wp-block-code"><code class="">ovhai app get &lt;app_id&gt;</code></pre>



<p>You should get:</p>



<p><code>History:<br>DATE STATE<br>20-1O-25 09:58:00 QUEUED<br>20-10-25 09:58:01 INITIALIZING<br>04-04-25 09:58:07 PENDING<br>04-04-25 10:03:10&nbsp;<strong>RUNNING</strong><br>Info:<br>Message: App is running</code></p>



<h5 class="wp-block-heading">2.3 Create a new n8n credential with AI Deploy app URL and Bearer access token</h5>



<p>First, using your <code><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color"><strong>&lt;app_id></strong></mark></code>, retrieve your AI Deploy app URL.</p>



<pre class="wp-block-code"><code class="">ovhai app get <span style="background-color: initial; font-family: inherit; font-size: inherit; text-align: initial; font-weight: inherit;">&lt;app_id&gt;</span> -o json | jq '.status.url' -r</code></pre>



<p>Then, create a new OpenAI credential from your n8n workflow, using your AI Deploy URL and the Bearer token as an API key.</p>



<figure class="wp-block-image size-large"><img loading="lazy" decoding="async" width="1024" height="580" src="https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-21-a-16.49.14-1024x580.png" alt="" class="wp-image-29837" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-21-a-16.49.14-1024x580.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-21-a-16.49.14-300x170.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-21-a-16.49.14-768x435.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-21-a-16.49.14-1536x870.png 1536w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-21-a-16.49.14-2048x1160.png 2048w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p>Don&#8217;t forget to replace <strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color"><code>6e10e6a5-2862-4c82-8c08-26c458ca12c7</code></mark></strong> with your <span style="background-color: initial; font-family: inherit; font-size: inherit; text-align: initial; font-weight: inherit;"><strong><code><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">&lt;app_id></mark></code></strong></span>.</p>



<h5 class="wp-block-heading">2.4 Create the LLM Guard node in n8n workflow</h5>



<p>Create a new <strong>OpenAI node</strong> to <strong>Message a model</strong> and select the new AI Deploy credential for LLM Guard usage.</p>



<p>Next, create the prompt as follows:</p>



<pre class="wp-block-code"><code class="">{{ $('Chat with the OVHcloud product expert').item.json.chatInput }}</code></pre>



<figure class="wp-block-image size-large"><img loading="lazy" decoding="async" width="1024" height="580" src="https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-21-a-18.09.43-1024x580.png" alt="" class="wp-image-29840" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-21-a-18.09.43-1024x580.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-21-a-18.09.43-300x170.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-21-a-18.09.43-768x435.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-21-a-18.09.43-1536x870.png 1536w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-21-a-18.09.43-2048x1160.png 2048w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p>Then, use an <code><strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">If</mark></strong></code> node to determine if the scenario is <strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color"><code>safe</code></mark></strong> or <strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color"><code>unsafe</code></mark></strong>:</p>



<figure class="wp-block-image size-large"><img loading="lazy" decoding="async" width="1024" height="580" src="https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-21-a-18.25.29-1024x580.png" alt="" class="wp-image-29842" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-21-a-18.25.29-1024x580.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-21-a-18.25.29-300x170.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-21-a-18.25.29-768x435.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-21-a-18.25.29-1536x870.png 1536w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-21-a-18.25.29-2048x1160.png 2048w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p>If the message is <code><strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">unsafe</mark></strong></code>, send an error message right away to stop the workflow.</p>



<figure class="wp-block-image size-large"><img loading="lazy" decoding="async" width="1024" height="580" src="https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-21-a-18.26.49-1024x580.png" alt="" class="wp-image-29843" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-21-a-18.26.49-1024x580.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-21-a-18.26.49-300x170.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-21-a-18.26.49-768x435.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-21-a-18.26.49-1536x870.png 1536w, https://blog.ovhcloud.com/wp-content/uploads/2025/10/Capture-decran-2025-10-21-a-18.26.49-2048x1160.png 2048w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p>But if the message is <code><strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">safe</mark></strong></code>, you can send the request to the AI Agent without issues 🔐.</p>



<h4 class="wp-block-heading">3. Set up AI Agent</h4>



<p>The&nbsp;<strong>AI Agent</strong>&nbsp;node in&nbsp;<strong>n8n</strong>&nbsp;acts as an intelligent orchestration layer that combines&nbsp;<strong>LLMs, memory, and external tools</strong>&nbsp;within an automated workflow.</p>



<p>It allows you to:</p>



<ul class="wp-block-list">
<li>Connect a <strong>Large Language Model</strong> using APIs (e.g., LLMs from AI Endpoints);</li>



<li>Use <strong>tools</strong> such as HTTP requests, databases, or RAG retrievers so the agent can take actions or fetch real information;</li>



<li>Maintain <strong>conversational memory</strong> via PostgreSQL databases;</li>



<li>Integrate directly with chat platforms (e.g., Slack, Teams) for interactive assistants (optional).</li>
</ul>



<p>Simply put, n8n becomes an&nbsp;<strong>agentic automation framework</strong>, enabling LLMs to not only provide answers, but also think, choose, and perform actions.</p>



<p>Please note that you can change and customise this n8n <code><strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">AI Agent</mark></strong></code> node to fit your use cases, using features like function calling or structured output. This is the most basic configuration for the given use case. You can go even further with different agents.</p>



<p>🧑‍💻&nbsp;<strong>How do I implement this RAG?</strong></p>



<p>First, create an <code><strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">AI Agent</mark></strong></code> node in <strong>n8n</strong> as follows:</p>



<figure class="wp-block-image size-large"><img loading="lazy" decoding="async" width="1024" height="580" src="https://blog.ovhcloud.com/wp-content/uploads/2025/11/image-1024x580.png" alt="" class="wp-image-29933" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/11/image-1024x580.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2025/11/image-300x170.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/11/image-768x435.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2025/11/image-1536x870.png 1536w, https://blog.ovhcloud.com/wp-content/uploads/2025/11/image-2048x1160.png 2048w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p>Then, a series of steps are required, the first of which is creating prompts.</p>



<h5 class="wp-block-heading">3.1 Create prompts</h5>



<p>In the AI Agent node on your n8n workflow, edit the user and system prompts.</p>



<p>Begin by creating the&nbsp;<strong>prompt</strong>,&nbsp;which is also the&nbsp;<strong>user message</strong>:</p>



<pre class="wp-block-code"><code class="">{{ $('Chat with the OVHcloud product expert').item.json.chatInput }}</code></pre>



<p>Then create the <strong>System Message</strong> as shown below:</p>



<pre class="wp-block-code"><code class="">You have access to a retriever tool connected to a knowledge base.  <br>Before answering, always search for relevant documents using the retriever tool.  <br>Use the retrieved context to answer accurately.  <br>If no relevant documents are found, say that you have no information about it.</code></pre>



<p>You should get a configuration like this:</p>



<figure class="wp-block-image size-large"><img loading="lazy" decoding="async" width="1024" height="580" src="https://blog.ovhcloud.com/wp-content/uploads/2025/11/image-1-1024x580.png" alt="" class="wp-image-29935" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/11/image-1-1024x580.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2025/11/image-1-300x170.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/11/image-1-768x435.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2025/11/image-1-1536x870.png 1536w, https://blog.ovhcloud.com/wp-content/uploads/2025/11/image-1-2048x1160.png 2048w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p>🤔 Well, an LLM is now needed for this to work!</p>



<h5 class="wp-block-heading">3.2 Select LLM using AI Endpoints API</h5>



<p>First, add an <strong>OpenAI Chat Model</strong> node, and then set it as the <code><strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">Chat Model</mark></strong></code> for your agent.</p>



<figure class="wp-block-image size-large"><img loading="lazy" decoding="async" width="1024" height="580" src="https://blog.ovhcloud.com/wp-content/uploads/2025/11/image-3-1024x580.png" alt="" class="wp-image-29939" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/11/image-3-1024x580.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2025/11/image-3-300x170.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/11/image-3-768x435.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2025/11/image-3-1536x870.png 1536w, https://blog.ovhcloud.com/wp-content/uploads/2025/11/image-3-2048x1160.png 2048w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p>Next, select one of the&nbsp;<a href="https://www.ovhcloud.com/en/public-cloud/ai-endpoints/" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">OVHcloud AI Endpoints</a>&nbsp;from the list provided, because they are compatible with Open AI APIs.</p>



<p>✅ <strong>How?</strong> By using the right API <a href="https://oai.endpoints.kepler.ai.cloud.ovh.net/v1" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer"><strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color"><code>https://oai.endpoints.kepler.ai.cloud.ovh.net/v1</code></mark></strong></a></p>



<figure class="wp-block-image size-large"><img loading="lazy" decoding="async" width="1024" height="580" src="https://blog.ovhcloud.com/wp-content/uploads/2025/11/image-2-1024x580.png" alt="" class="wp-image-29936" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/11/image-2-1024x580.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2025/11/image-2-300x170.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/11/image-2-768x435.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2025/11/image-2-1536x870.png 1536w, https://blog.ovhcloud.com/wp-content/uploads/2025/11/image-2-2048x1160.png 2048w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p>The <a href="https://www.ovhcloud.com/en/public-cloud/ai-endpoints/catalog/gpt-oss-120b/" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer"><strong>GPT OSS 120B</strong></a> model has been selected for this use case. Other models, such as Llama, Mistral, and Qwen, are also available.</p>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p><mark style="background-color:#fcb900" class="has-inline-color">⚠️ <strong>WARNING</strong> ⚠️</mark></p>



<p>If you are using a recent version of n8n, you will likely encounter the <strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color"><code>/responses</code></mark></strong> issue (linked to OpenAI compatibility). To resolve this, you will need to disable the button <strong><code><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">Use Responses API</mark></code></strong> and everything will work correctly</p>
</blockquote>



<figure class="wp-block-image aligncenter size-full is-resized"><img loading="lazy" decoding="async" width="829" height="675" src="https://blog.ovhcloud.com/wp-content/uploads/2026/01/02_44_08-1.jpg" alt="" class="wp-image-30352" style="aspect-ratio:1.2281554640124863;width:409px;height:auto" srcset="https://blog.ovhcloud.com/wp-content/uploads/2026/01/02_44_08-1.jpg 829w, https://blog.ovhcloud.com/wp-content/uploads/2026/01/02_44_08-1-300x244.jpg 300w, https://blog.ovhcloud.com/wp-content/uploads/2026/01/02_44_08-1-768x625.jpg 768w" sizes="auto, (max-width: 829px) 100vw, 829px" /><figcaption class="wp-element-caption"><em>Tips to fix /responses issue</em></figcaption></figure>



<p>Your LLM is now set to answer your questions! Don’t forget, it needs access to the knowledge base.</p>



<h5 class="wp-block-heading">3.3 Connect the knowledge base to the RAG retriever</h5>



<p>As usual, the first step is to create an n8n node called <code><strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">PGVector Vector Store nod</mark>e</strong></code> and enter your PGvector credentials.</p>



<figure class="wp-block-image size-large"><img loading="lazy" decoding="async" width="1024" height="580" src="https://blog.ovhcloud.com/wp-content/uploads/2025/11/image-4-1024x580.png" alt="" class="wp-image-29943" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/11/image-4-1024x580.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2025/11/image-4-300x170.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/11/image-4-768x435.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2025/11/image-4-1536x870.png 1536w, https://blog.ovhcloud.com/wp-content/uploads/2025/11/image-4-2048x1160.png 2048w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p>Next, link this element to the <strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color"><code>Tools</code></mark></strong> section of the AI Agent node.</p>



<figure class="wp-block-image size-large"><img loading="lazy" decoding="async" width="1024" height="580" src="https://blog.ovhcloud.com/wp-content/uploads/2025/11/image-5-1024x580.png" alt="" class="wp-image-29944" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/11/image-5-1024x580.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2025/11/image-5-300x170.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/11/image-5-768x435.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2025/11/image-5-1536x870.png 1536w, https://blog.ovhcloud.com/wp-content/uploads/2025/11/image-5-2048x1160.png 2048w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p>Remember to connect your PG vector database so that the retriever can access the previously generated embeddings. Here’s an overview of what you’ll get.</p>



<figure class="wp-block-image size-large"><img loading="lazy" decoding="async" width="1024" height="580" src="https://blog.ovhcloud.com/wp-content/uploads/2025/11/image-6-1024x580.png" alt="" class="wp-image-29945" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/11/image-6-1024x580.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2025/11/image-6-300x170.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/11/image-6-768x435.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2025/11/image-6-1536x870.png 1536w, https://blog.ovhcloud.com/wp-content/uploads/2025/11/image-6-2048x1160.png 2048w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p>⏳Nearly done! The final step is to add the database memory.</p>



<h5 class="wp-block-heading">3.4 Manage conversation history with database memory</h5>



<p>Creating&nbsp;<strong>Database Memory</strong>&nbsp;node in n8n (PostgreSQL) lets you link it to your AI Agent, so it can store and retrieve past conversation history. This enables the model to remember and use context from multiple interactions.</p>



<figure class="wp-block-image size-large"><img loading="lazy" decoding="async" width="1024" height="580" src="https://blog.ovhcloud.com/wp-content/uploads/2025/11/image-7-1024x580.png" alt="" class="wp-image-29946" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/11/image-7-1024x580.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2025/11/image-7-300x170.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/11/image-7-768x435.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2025/11/image-7-1536x870.png 1536w, https://blog.ovhcloud.com/wp-content/uploads/2025/11/image-7-2048x1160.png 2048w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p>So link this PostgreSQL database to the <code><strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color">Memory</mark></strong></code> section of your AI agent.</p>



<figure class="wp-block-image size-large"><img loading="lazy" decoding="async" width="1024" height="580" src="https://blog.ovhcloud.com/wp-content/uploads/2025/11/image-8-1024x580.png" alt="" class="wp-image-29947" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/11/image-8-1024x580.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2025/11/image-8-300x170.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/11/image-8-768x435.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2025/11/image-8-1536x870.png 1536w, https://blog.ovhcloud.com/wp-content/uploads/2025/11/image-8-2048x1160.png 2048w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p>Congrats! 🥳 Your&nbsp;<strong>n8n RAG workflow</strong>&nbsp;is now complete. Ready to test it?</p>



<h4 class="wp-block-heading">4. Make the most of your automated workflow</h4>



<p>Want to try it? It’s easy!</p>



<p>By clicking the orange <strong><mark style="background-color:rgba(0, 0, 0, 0)" class="has-inline-color has-ast-global-color-0-color"><code>Open chat</code></mark></strong> button, you can ask the AI agent questions about OVHcloud products, particularly where you need technical assistance.</p>



<figure class="wp-block-video"><video height="1660" style="aspect-ratio: 2930 / 1660;" width="2930" controls src="https://blog.ovhcloud.com/wp-content/uploads/2025/11/video-n8n1.mp4"></video></figure>



<p>For example, you can ask the LLM about rate limits in OVHcloud AI Endpoints and get the information in seconds.</p>



<figure class="wp-block-video"><video height="1660" style="aspect-ratio: 2930 / 1660;" width="2930" controls src="https://blog.ovhcloud.com/wp-content/uploads/2025/11/video-n8n2.mp4"></video></figure>



<p>You can now build your own autonomous RAG system using OVHcloud Public Cloud, suited for a wide range of applications.</p>



<h2 class="wp-block-heading">What’s next?</h2>



<p>To sum up, this reference architecture provides a guide on using&nbsp;<strong>n8n</strong> with&nbsp;<strong>OVHcloud AI Endpoints</strong>,&nbsp;<strong>AI Deploy</strong>,&nbsp;<strong>Object Storage</strong>, and&nbsp;<strong>PostgreSQL + pgvector</strong> to build a fully controlled, autonomous&nbsp;<strong>RAG AI system</strong>.</p>



<p>Teams can build scalable AI assistants that work securely and independently in their cloud environment by orchestrating ingestion, embedding generation, vector storage, retrieval, and LLM safety check, and reasoning within a single workflow.</p>



<p>With the core architecture in place, you can add more features to improve the capabilities and robustness of your agentic RAG system:</p>



<ul class="wp-block-list">
<li>Web search</li>



<li>Images with OCR</li>



<li>Audio files transcribed using the Whisper model</li>
</ul>



<p>This delivers an extensive knowledge base and a wider variety of use cases!</p>
<img loading="lazy" decoding="async" src="//blog.ovhcloud.com/wp-content/plugins/matomo/app/matomo.php?idsite=1&amp;rec=1&amp;url=https%3A%2F%2Fblog.ovhcloud.com%2Freference-architecture-build-a-sovereign-n8n-rag-workflow-for-ai-agent-using-ovhcloud-public-cloud-solutions%2F&amp;action_name=Reference%20Architecture%3A%20build%20a%20sovereign%20n8n%20RAG%20workflow%20for%20AI%20agent%20using%20OVHcloud%20Public%20Cloud%20solutions&amp;urlref=https%3A%2F%2Fblog.ovhcloud.com%2Ffeed%2F" style="border:0;width:0;height:0" width="0" height="0" alt="" />]]></content:encoded>
					
		
		<enclosure url="https://blog.ovhcloud.com/wp-content/uploads/2025/11/video-n8n1.mp4" length="11190376" type="video/mp4" />
<enclosure url="https://blog.ovhcloud.com/wp-content/uploads/2025/11/video-n8n2.mp4" length="9881210" type="video/mp4" />

			</item>
		<item>
		<title>Safety first: Detect harmful texts using an AI safeguard agent</title>
		<link>https://blog.ovhcloud.com/safety-first-detect-harmful-texts-using-an-ai-safeguard-agent/</link>
		
		<dc:creator><![CDATA[Alexandre Movsessian]]></dc:creator>
		<pubDate>Thu, 22 Jan 2026 10:46:11 +0000</pubDate>
				<category><![CDATA[Deploy & Scale]]></category>
		<category><![CDATA[OVHcloud Engineering]]></category>
		<category><![CDATA[AI]]></category>
		<category><![CDATA[Machine learning]]></category>
		<guid isPermaLink="false">https://blog.ovhcloud.com/?p=30185</guid>

					<description><![CDATA[This article explains how to use the Qwen 3 Guard safeguard models provided by OVHCloud. Using this guide, you can analyse and moderate texts for LLM applications, chat platforms, customer support systems, or any other text-based services requiring safe and compliant interactions. Our focus will be on written content, such as conversations or plain text. [&#8230;]<img src="//blog.ovhcloud.com/wp-content/plugins/matomo/app/matomo.php?idsite=1&amp;rec=1&amp;url=https%3A%2F%2Fblog.ovhcloud.com%2Fsafety-first-detect-harmful-texts-using-an-ai-safeguard-agent%2F&amp;action_name=Safety%20first%3A%20Detect%20harmful%20texts%20using%20an%20AI%20safeguard%20agent&amp;urlref=https%3A%2F%2Fblog.ovhcloud.com%2Ffeed%2F" style="border:0;width:0;height:0" width="0" height="0" alt="" />]]></description>
										<content:encoded><![CDATA[
<figure class="wp-block-image size-full"><img loading="lazy" decoding="async" width="981" height="463" src="https://blog.ovhcloud.com/wp-content/uploads/2026/01/image.png" alt="" class="wp-image-30187" srcset="https://blog.ovhcloud.com/wp-content/uploads/2026/01/image.png 981w, https://blog.ovhcloud.com/wp-content/uploads/2026/01/image-300x142.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2026/01/image-768x362.png 768w" sizes="auto, (max-width: 981px) 100vw, 981px" /></figure>



<p class="has-text-align-left"><strong>This article explains how to use the Qwen 3 Guard safeguard models provided by OVHCloud.</strong></p>



<p>Using this guide, you can analyse and moderate texts for LLM applications, chat platforms, customer support systems, or any other text-based services requiring safe and compliant interactions.</p>



<p>Our focus will be on written content, such as conversations or plain text. Although image moderators exist, they won’t be covered here.</p>



<h2 class="wp-block-heading"><strong>Introduction</strong></h2>



<p><strong><br></strong>As <strong>Large Language Models</strong> (LLMs) continue to grow, access to information has become more seamless, but this ease of access makes it easier to generate, and be exposed to, harmful or toxic content.</p>



<p>LLMs can be prompted with malicious queries (e.g., “How do I make a bomb?”) and some models might comply by generating potentially dangerous responses. This risk is particularly concerning given the widespread availability of LLMs, to both minors and malicious actors alike.</p>



<p>To combat this, LLM providers train their models to reject toxic prompts, and integrate safety features to prevent the creation of harmful content. Even so, users often craft ‘<strong>jailbreaks</strong>’, which are specific prompts designed to get around these safety measures.</p>



<p>As a result, providers have created <strong>specialised safeguard models</strong> to find and remove toxic content in writing.</p>



<h1 class="wp-block-heading">What is toxicity?</h1>



<p>Toxicity is inherently difficult to define, as perceptions vary depending on factors such as individual sensitivity, cultural background, age, and personal experience.</p>



<p>Perceptions of content can vary widely. For example, some users may find certain jokes offensive, while others consider them perfectly acceptable. Similarly, roleplaying with an AI chat may be enjoyable for some, yet deemed inappropriate by others depending on the context.</p>



<p>Furthermore, each moderation system focuses on different categories of harmful content, based on the specific data and instructions it was trained on. For instance, models developed in the United States tend to be highly sensitive to hate speech, political content, and other related categories.</p>



<p>Because jailbreak attempts are a fairly new issue, existing moderation models often fail to address them.</p>



<p>Below are the toxicity categories for the Qwen 3 Guard models:</p>



<figure class="wp-block-table"><table class="has-fixed-layout"><tbody><tr><td><strong>Name</strong></td><td><strong>Description</strong></td></tr><tr><td><em>Violent</em></td><td>Content that provides detailed instructions, methods, or advice on how to commit acts of violence, including the manufacture, acquisition, or use of weapons. Also includes depictions of violence.</td></tr><tr><td><em>Nonviolent illegal acts</em></td><td>Content providing guidance or advice for nonviolent criminal activities like hacking, unauthorised drug manufacturing, or theft.</td></tr><tr><td><em>Sexual content or sexual acts</em></td><td>Content with sexual depictions, references, or descriptions of people. Also includes content with explicit sexual imagery, references, or descriptions of illegal or unethical sexual acts, such as rape, bestiality, incest, and sexual slavery.</td></tr><tr><td><em>Personally identifiable information</em></td><td>Content that shares or discloses sensitive personal identifying information, with authorisation, such as name, ID number, address, phone number, medical records, financial details, and account passwords, etc.</td></tr><tr><td><em>Suicide &amp; self-harm</em></td><td>Content advocating, directly encouraging, or detailing methods for self-harm, suicide, or dangerous activities that could lead to serious injury or death.</td></tr><tr><td><em>Unethical acts</em></td><td>Any immoral or unethical content or acts, including but not limited to bias, discrimination, stereotype, injustice, hate speech, offensive language, harassment, insults, threat, defamation, extremism, misinformation regarding ethics, and other behaviours that, while not illegal, are still considered unethical.</td></tr><tr><td><em>Politically sensitive topics</em></td><td>The deliberate creation or spread of false information about government actions, historical events, or public figures that is demonstrably untrue and poses risk of public deception or social harm.</td></tr><tr><td><em>Copyright violation</em></td><td>Content that includes unauthorised reproduction, distribution, public display, or derivative use of copyrighted materials, such as novels, scripts, lyrics, and other legally protected creative works, without the copyright holder’s clear consent.</td></tr><tr><td><em>Jailbreak</em></td><td>Content that explicitly attempts to override the model&#8217;s system prompt or model conditioning.</td></tr></tbody></table></figure>



<p>These categories are <strong>not mutually exclusive</strong>. A text may very well contain both Unethical Acts and Violence, for example. Most notably, jailbreaks often include another kind of toxic query as it is designed to bypass security guardrails. The Qwen 3 Guard moderator, however, will only return one category.</p>



<p>These categories were arbitrarily chosen by Qwen 3 Guard creators; they can’t be changed, but <strong>you may choose to ignore some</strong> depending on your use case.</p>



<h1 class="wp-block-heading">Metrics</h1>



<p><em>Attack</em>: An attack refers to any attempt to produce harmful or toxic content. This is either a prompt crafted to make an LLM generate harmful output, or just a user’s toxic message in a chat system.</p>



<p><em>Attack Success Rates (ASR)</em>: This is a metric used to assess the effectiveness of a moderation system. It represents the <strong>proportion of attacks that successfully bypass the moderator</strong> and go undetected. A lower ASR indicates a more robust moderation system.</p>



<p><em>False positive</em>: A false positive occurs when benign, nontoxic content is incorrectly flagged as harmful by the moderator.</p>



<p><em>False Positive Rate (FPR)</em>: The FPR measures how often a moderation system misclassifies safe content as toxic. It complements the ASR by reflecting the <strong>model’s ability to correctly allow harmless content through</strong>. A lower FPR indicates better reliability.</p>



<h2 class="wp-block-heading">Qwen 3 Guard</h2>



<p>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; Qwen 3 Guard was launched in October 2025 by Qwen, Alibaba’s AI team. After extensive testing and evaluation, we found this model to be the most effective in safeguarding content.</p>



<p>Besides being efficient, Qwen 3 Guard can detect toxicity across nine categories, including jailbreak attempts, a feature that isn’t common in safeguard models.</p>



<p>It also provides explanations by specifying the exact category detected.</p>



<h3 class="wp-block-heading">Specs</h3>



<ul class="wp-block-list">
<li>Base model: Qwen 3</li>



<li>Flavours: 0.6B, 4B, 8B</li>



<li>Context size: 32,768 tokens</li>



<li>Languages: English, French and 117 other languages and dialects</li>



<li>Tasks:<ul><li>Detection of toxicity in raw text</li></ul><ul><li>Detection of toxicity in LLM dialogue</li></ul><ul><li>Detection of answer refusal (LLM dialogue only)</li></ul>
<ul class="wp-block-list">
<li>Classification of toxicity</li>
</ul>
</li>
</ul>



<h3 class="wp-block-heading">Availability</h3>



<p><a href="https://www.ovhcloud.com/en/public-cloud/ai-endpoints/catalog" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">https://www.ovhcloud.com/en/public-cloud/ai-endpoints/catalog</a></p>



<p>There are two flavours of Qwen 3 Guard available on OVHCloud:</p>



<p><strong><em>Qwen 3 Guard 0.6B</em></strong>: This lightweight model is very effective at detecting overt toxic content.</p>



<p><strong><em>Qwen 3 Guard 8B</em></strong>: This heavier model comes in handy when confronted with more nuanced examples.</p>



<h3 class="wp-block-heading">Scores</h3>



<figure class="wp-block-table"><table class="has-fixed-layout"><tbody><tr><td><strong>&nbsp;</strong></td><td><strong><em>ASR</em></strong></td><td><strong><em>FPR</em></strong></td></tr><tr><td><strong><em>Qwen 3 Guard 0.6B</em></strong></td><td>0.20</td><td>0.06</td></tr><tr><td><strong><em>Qwen 3 Guard 8B</em></strong></td><td>0.20</td><td>0.04</td></tr></tbody></table></figure>



<h3 class="wp-block-heading">&nbsp;</h3>



<h3 class="wp-block-heading">Notes</h3>



<ul class="wp-block-list">
<li>The Qwen 3 Guard models has three safety labels for more precise moderation: Safe, Controversial, Unsafe</li>



<li>Although the model can moderate chats, it is recommended to process each part of the dialogue individually rather than submitting the entire conversation at once. Guard Models, like any LLMs, perform better in detection when the context size is kept extremely brief.</li>



<li>Since Qwen Guard is developed by a Chinese company, its interpretation of toxic content may differ from yours. If necessary, you can overlook certain categories.</li>
</ul>



<h1 class="wp-block-heading">How do I set up my own moderator?</h1>



<p>First, you need to choose the flavour you want:</p>



<ul class="wp-block-list">
<li><strong><em>Qwen 3 Guard 0.6B</em></strong> is <strong>lightweight</strong>, <strong>fast</strong>, <strong>efficient</strong> and is great at detecting <strong>overt toxic content</strong>, like <em>Sexual Content</em> or <em>Violence</em> in texts.</li>
</ul>



<ul class="wp-block-list">
<li><strong><em>Qwen 3 Guard 8B</em></strong> is heavier, slightly slower but it is more effective against <strong>more nuanced toxic content </strong>like <em>Jailbreak</em> or <em>Unethical Acts</em>, and has a <strong>lower false positive rate</strong>.</li>
</ul>



<p>Your use case is the key to choosing the right model. Do you need to moderate a large volume of text? Is processing speed a priority? How crucial is it to minimise false positives? Are you dealing with nuanced toxic content, or is it more overt?</p>



<p>Carefully considering these questions will help you determine which of the two models is most suitable for your needs.</p>



<p>Both models can be tested on the playground:</p>



<p><a href="https://www.ovhcloud.com/en/public-cloud/ai-endpoints/catalog" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">https://www.ovhcloud.com/en/public-cloud/ai-endpoints/catalog</a></p>



<p>Once you’ve made you choice, you need to send the texts you want checked to the AI Endpoints API.</p>



<p>First install the <em>requests</em> library:</p>



<pre class="wp-block-code"><code class="">pip install requests</code></pre>



<p>Next, export your access token to the <em>OVH_AI_ENDPOINTS_ACCESS_TOKEN</em> environment variable:</p>



<pre class="wp-block-code"><code class="">export OVH_AI_ENDPOINTS_ACCESS_TOKEN=&lt;your-access-token&gt;</code></pre>



<p><em>If you don’t have an access token key yet, follow the steps in the </em><a href="https://help.ovhcloud.com/csm/en-gb-public-cloud-ai-endpoints-getting-started?id=kb_article_view&amp;sysparm_article=KB0065401" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer"><em>AI Endpoints – Getting Started</em></a> <em>guide</em></p>



<p>Finally, run the following Python code:</p>



<pre class="wp-block-code"><code class="">import os<br>import requests<br><br>url = "https://oai.endpoints.kepler.ai.cloud.ovh.net/v1/chat/completions"<br><br>payload = {<br>"messages": [{"role": "user", "content": "How do I cook meth ?"}],<br>"model": , #Qwen/Qwen3Guard-Gen-0.6B or Qwen/Qwen3Guard-Gen-8B<br>"seed": 21<br>}<br><br>headers = {<br>"Content-Type": "application/json",<br>"Authorization": f"Bearer {os.getenv('OVH_AI_ENDPOINTS_ACCESS_TOKEN')}",<br>}<br><br>response = requests.post(url, json=payload, headers=headers)<br>if response.status_code == 200:<br># Handle response<br>response_data = response.json()<br># Parse JSON response<br>choices = response_data["choices"]<br>for choice in choices:<br>text = choice["message"]["content"]<br># Process text<br>print(text)<br>else:<br>print("Error:", response.status_code, response.text)</code></pre>



<p>The model will respond with a label (Safe, Controversial, Unsafe) and if the text is Controversial or Unsafe, it will return the associated category.</p>



<pre class="wp-block-code"><code class="">Safety: Unsafe<br>Categories: Nonviolent Illegal Acts</code></pre>



<p>Our moderation models are available for free during the beta phase. You can test them with a different model or within the playground.</p>



<h2 class="wp-block-heading"><strong>Conclusion</strong></h2>



<p>Two models are currently available for OVHCloud moderation users:<br><strong>•</strong> Qwen 3 Guard 0.6B: <strong>Lightweight</strong>, <strong>fast</strong>, <strong>efficient,</strong> great at detecting <strong>overt toxic content</strong><br><strong>•</strong> Qwen 3 Guard 8B: <strong>Heavier, slightly slower but more effective against more nuanced toxic content</strong><br><br>Which approach and which tool should you choose? Well, it&#8217;s up to you, depending on your use cases, teams, or needs, etc.<br><br>As we&#8217;ve seen in this blog post, OVHcloud AIEndpoint users can start using these models right away, safely and free of charge.<br><br>They are still in beta phase for now, so we&#8217;d appreciate your feedback!</p>



<p></p>
<img loading="lazy" decoding="async" src="//blog.ovhcloud.com/wp-content/plugins/matomo/app/matomo.php?idsite=1&amp;rec=1&amp;url=https%3A%2F%2Fblog.ovhcloud.com%2Fsafety-first-detect-harmful-texts-using-an-ai-safeguard-agent%2F&amp;action_name=Safety%20first%3A%20Detect%20harmful%20texts%20using%20an%20AI%20safeguard%20agent&amp;urlref=https%3A%2F%2Fblog.ovhcloud.com%2Ffeed%2F" style="border:0;width:0;height:0" width="0" height="0" alt="" />]]></content:encoded>
					
		
		
			</item>
		<item>
		<title>Agentic AI from a security perspective</title>
		<link>https://blog.ovhcloud.com/agentic-ai-from-a-security-perspective/</link>
		
		<dc:creator><![CDATA[Julien Levrard]]></dc:creator>
		<pubDate>Thu, 08 Jan 2026 14:58:33 +0000</pubDate>
				<category><![CDATA[Deploy & Scale]]></category>
		<category><![CDATA[AI]]></category>
		<category><![CDATA[Security]]></category>
		<guid isPermaLink="false">https://blog.ovhcloud.com/?p=30148</guid>

					<description><![CDATA[Large Language Models (LLMs) and generative AI technologies are everywhere, infiltrating both our personal and professional daily lives. Well-known services are already diverting most internet users away from their old browsing habits, and online information consumption is being profoundly transformed, most likely with no possible return to past behaviours. Issues related to intellectual property laws [&#8230;]<img src="//blog.ovhcloud.com/wp-content/plugins/matomo/app/matomo.php?idsite=1&amp;rec=1&amp;url=https%3A%2F%2Fblog.ovhcloud.com%2Fagentic-ai-from-a-security-perspective%2F&amp;action_name=Agentic%20AI%20from%20a%20security%20perspective&amp;urlref=https%3A%2F%2Fblog.ovhcloud.com%2Ffeed%2F" style="border:0;width:0;height:0" width="0" height="0" alt="" />]]></description>
										<content:encoded><![CDATA[
<p>Large Language Models (LLMs) and generative AI technologies are everywhere, infiltrating both our personal and professional daily lives. Well-known services are already diverting most internet users away from their old browsing habits, and online information consumption is being profoundly transformed, most likely with no possible return to past behaviours.</p>



<p>Issues related to intellectual property laws and the source of data used to train LLMs, which is sometimes confidential or personal, as well as potential biases in the data, intentional or otherwise, are regularly debated in the press and within technology communities. However, the current focus is on the race between LLM providers, who are competing to develop faster, more efficient models, in search of the ‘wow’ factor that will temporarily propel them to the rank of global AI leader.</p>



<p>Meanwhile, organisations are integrating these technologies into their daily activities at their own pace. Implementation is driven both by employees keen to improve their individual productivity, often based on their experience using AI tools in their personal life, and by business leaders and managers, who see an opportunity to optimise efficiency of low-value-added tasks.</p>



<p>At OVHcloud, we have launched an ‘AI Labs’ initiative, which is responsible for centralising projects and experiments using LLM tools. This team now supervises over a hundred projects, and new ones are added every week. The approach aims to catalyse ideas and provide a framework for efficiently implementing effective production tools.</p>



<p>From a data security perspective, the proliferation of experimentation and proof-of-concept (POC) projects creates numerous additional risks that need consideration. Modelling interactions between each component is necessary to understand these risks, as many configurations are possible.</p>



<p>In this article we will take a look at some example use cases, identify the main risks and provide suggestions for how to address them using a risk reduction logic model. We will focus on simple use cases where a user accesses an application for their work. These applications are accessible from their work context, and each have access management mechanisms that verify the user and grant them access to the relevant data and functions associated with their business profile.</p>



<figure class="wp-block-image aligncenter size-medium"><img loading="lazy" decoding="async" width="270" height="300" src="https://blog.ovhcloud.com/wp-content/uploads/2025/12/mermaid-1767008664917-270x300.png" alt="" class="wp-image-30150" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/12/mermaid-1767008664917-270x300.png 270w, https://blog.ovhcloud.com/wp-content/uploads/2025/12/mermaid-1767008664917-768x854.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2025/12/mermaid-1767008664917.png 921w" sizes="auto, (max-width: 270px) 100vw, 270px" /></figure>



<p>The introduction of LLM technologies fits into the usual operating mode of an information system to enrich the user experience and offer additional features. Let’s take a look at the examples.</p>



<h1 class="wp-block-heading">Conversational agents (without third-party integration)</h1>



<p>Most professionals working on a computer regularly use conversational agents to ‘enhance’ their work, often without acknowledging it, for example when writing an email, summarising a document, finding a complex Excel formula, answering a legal or technical question, etc.). As these agents are not connected to the company’s information system,  the risks are limited and depend on the attitude and practices of the user, for example with regards to uploading data, copying and pasting confidential data into the agent, etc.</p>



<p><a id="_msocom_1"></a></p>



<figure class="wp-block-image aligncenter size-full"><img loading="lazy" decoding="async" width="1024" height="529" src="https://blog.ovhcloud.com/wp-content/uploads/2025/12/mermaid-1767008664926.png" alt="" class="wp-image-30151" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/12/mermaid-1767008664926.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2025/12/mermaid-1767008664926-300x155.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/12/mermaid-1767008664926-768x397.png 768w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p>In this context, the user is the go-between managing the information transfer between the company application and the third-party agent. The agent only has access to information voluntarily sent by the user, typically via the service interface that allows prompts to be entered. These services are rapidly extending their capabilities, allowing file upload, and microphone or camera access, but we remain in a classic responsibility framework in terms of security, with the human in the loop by design.</p>



<h2 class="wp-block-heading" id="examples">Examples</h2>



<ul class="wp-block-list">
<li>Public AI services (Mistral, Openai, Grok, Omissimo, etc.)</li>



<li>AI services contracted by the company from public service publishers or specialised players</li>



<li>Internal chatbot</li>
</ul>



<h2 class="wp-block-heading" id="associated-security-risks">Associated security risks</h2>



<ul class="wp-block-list">
<li>Sending sensitive data (documents, confidential data, personal data, etc.) to the AI service and losing control over this data.</li>



<li>Training models on confidential data sent by users, which can lead to leaking this data to a user who should not have access to it.</li>
</ul>



<h2 class="wp-block-heading" id="possible-security-measures">Measures to implement</h2>



<ul class="wp-block-list">
<li>User awareness</li>



<li>AI charter</li>



<li>Blocking services accessible from the company&#8217;s information system</li>



<li>Contract with suppliers including security and confidentiality clauses for user-transmitted information</li>



<li>Traffic inspection and identification of confidential data using regular expressions</li>



<li>Dedicated instance for the company, fine-tuned or enriched by a RAG with company data (not very sensitive), allowing the LLM to be contextualised to the user’s context.</li>
</ul>



<h1 class="wp-block-heading" id="ia-augmented-application">&#8220;AI Augmented&#8221; Application</h1>



<p>The various editor solutions, in SaaS or deployed internally, are gradually enriched with functions based on LLMs, i.e. an agent on the application side that consumes an LLM with prompts designed by the editor on the data processed by the application. The editor enriches its solution within its own security model. On the user side, there is no change in usage, the application is simply enriched with new functions, for example synthesis, intelligent suggestions, translation, etc.). LLM processing can be done locally or consumed on external services.</p>



<figure class="wp-block-image aligncenter size-full"><img loading="lazy" decoding="async" width="1024" height="191" src="https://blog.ovhcloud.com/wp-content/uploads/2025/12/mermaid-1767008664934.png" alt="" class="wp-image-30152" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/12/mermaid-1767008664934.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2025/12/mermaid-1767008664934-300x56.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/12/mermaid-1767008664934-768x143.png 768w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p>In this use case, the publisher or application manager is responsible for data security and processing via the LLM; the user has no control and the use of these features is integrated into their usual usage. We remain in a classic security management framework, the application manager (internal or external) is the guarantor of the security of the data they process in the application. The application is enriched with new features and complexity increases, but the security model is preserved.</p>



<h2 class="wp-block-heading" id="examples-2">Examples</h2>



<ul class="wp-block-list">
<li>Messaging and video conferencing service with AI features, for example real-time translation, discussion synthesis, automatic meeting minutes etc.</li>



<li>Any ‘AI wizards’  in SaaS application</li>
</ul>



<h2 class="wp-block-heading" id="associated-risks">Associated security risks</h2>



<ul class="wp-block-list">
<li>Insufficient segmentation of access rights to data in the application, allowing bypassing of usual application access controls. This is the case when the agent has a high-privilege account (to simplify and accelerate the development of features) or when access restriction is not implemented at data level.</li>



<li>Prompt injection into the application</li>



<li>Dependence on an uncontrolled supply chain</li>



<li>Data leakage to a subcontractor</li>
</ul>



<h2 class="wp-block-heading" id="measures-to-be-implemented">Measures to implement</h2>



<ul class="wp-block-list">
<li>Security clauses in contracts</li>



<li>Security insurance plan for application provider</li>



<li>Review of subcontractor dependency chains</li>



<li>Disabling unnecessary AI functions</li>



<li>Deep isolation of sensitive applications</li>
</ul>



<h1 class="wp-block-heading">Agentic IA</h1>



<p>We will now look at actual ‘Agentic AI’. In these cases, the agent is at the centre of the workflow. The agent becomes an orchestrator of resources. It has several roles, in particular:</p>



<ul class="wp-block-list">
<li>Capturing user expectations and triggering the sequence of actions</li>



<li>Retrieving the necessary data to contextualise and process the request</li>



<li>Sending data and instructions to a LLM to find the sequence of actions to be performed</li>



<li>Managing iterations with available services and LLMs to best handle the request</li>



<li>Triggering actions on accessible services</li>



<li>Obtaining (eventually) user validation to validate actions</li>



<li>Providing visibility to the user on actions performed and results obtained</li>
</ul>



<figure class="wp-block-image aligncenter size-full"><img loading="lazy" decoding="async" width="1024" height="632" src="https://blog.ovhcloud.com/wp-content/uploads/2025/12/mermaid-1767008664941.png" alt="" class="wp-image-30153" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/12/mermaid-1767008664941.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2025/12/mermaid-1767008664941-300x185.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/12/mermaid-1767008664941-768x474.png 768w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p>To properly understand the risks, it is necessary to look at different types of agent implementations.</p>



<h2 class="wp-block-heading">Agents integrated into local applications</h2>



<p>Applications are gradually being enriched with the ability to connect to an LLM service. Generally, this is done via APIs to LLM services or locally on the machine. In this case, the application will integrate an agent and incorporate its use into the usual application experience. The framework is equivalent to that of an enriched SaaS application, but the configuration and calls to the LLM are made from the user’s workstation. The functionality can be native or installed in the form of a plugin.</p>



<figure class="wp-block-image aligncenter size-full"><img loading="lazy" decoding="async" width="1024" height="436" src="https://blog.ovhcloud.com/wp-content/uploads/2025/12/mermaid-1767008664949.png" alt="" class="wp-image-30154" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/12/mermaid-1767008664949.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2025/12/mermaid-1767008664949-300x128.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/12/mermaid-1767008664949-768x327.png 768w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<h3 class="wp-block-heading" id="examples-3">Examples</h3>



<ul class="wp-block-list">
<li>Microsoft Copilot AI agent</li>



<li>AI function in office applications (OnlyOffice, Joplin, email client, etc.)</li>



<li>Apple intelligence</li>
</ul>



<h3 class="wp-block-heading" id="risks">Associated security risks</h3>



<ul class="wp-block-list">
<li>Loss of control over data processed by adding connectivity functions to third-party services (be careful with default tool configurations)</li>



<li>Risks are similar to “cloud” functions in applications, allowing cloud storage or sharing, often configured by default</li>



<li>Leakage of LLM authentication secrets (Bearer Token)</li>
</ul>



<h3 class="wp-block-heading" id="measures-to-implement">Measures to implement</h3>



<ul class="wp-block-list">
<li>User awareness</li>



<li>Application configuration controls</li>



<li>Validation of applications on workstations and smartphones</li>



<li>Monitoring and inspection of network and application flows</li>



<li>Local management of secrets</li>
</ul>



<h2 class="wp-block-heading" id="generalist-or-specialized-local-agents">Generalist or Specialized Local Agents</h2>



<p>Unlike the previous use case where the application is simply enriched with LLM functions, agents are applications whose primary goal is to integrate LLM functions into a workflow. The risk model is similar, but by nature, the functionalities are much richer and focused on optimising the consumption of LLM services. For example:</p>



<ul class="wp-block-list">
<li>Configuration of multiple LLM services in parallel</li>



<li>Personalisation of system and user prompt templates by the user</li>



<li>Integration of local or remote MCP services to enrich the data accessible to the agent</li>



<li>Cost control function</li>



<li>Optimisation of requests and context management</li>
</ul>



<p>These agents can be generalist or specialised. In particular, this type of agent is widely used by developers within their IDE . In this context, security management relies on the user and the local configuration of tools. Capabilities may be extended with marketplace, like plugins to add connectors to external services or capabilities. The complexity of configurations, the lack of proven and hardened standards due to the relative novelty of these tools generates many risks, on an application directly run on user workstation, with all their rights.</p>



<figure class="wp-block-image aligncenter size-full"><img loading="lazy" decoding="async" width="826" height="1024" src="https://blog.ovhcloud.com/wp-content/uploads/2025/12/mermaid-1767008664957.png" alt="" class="wp-image-30155" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/12/mermaid-1767008664957.png 826w, https://blog.ovhcloud.com/wp-content/uploads/2025/12/mermaid-1767008664957-242x300.png 242w, https://blog.ovhcloud.com/wp-content/uploads/2025/12/mermaid-1767008664957-768x952.png 768w" sizes="auto, (max-width: 826px) 100vw, 826px" /></figure>



<h3 class="wp-block-heading" id="examples-4">Examples</h3>



<ul class="wp-block-list">
<li>Generalist agents: Goose</li>



<li>Specialised agents: Claude desktop, Cursor, Shai, Github Copilot, Continue, Kilo Code</li>
</ul>



<h3 class="wp-block-heading" id="risks-2">Associated security risks</h3>



<ul class="wp-block-list">
<li>Connection to third-party services without controls via marketplace (MCP connector for third-party services)</li>



<li>Uncontrolled access to local file system</li>



<li>Sending confidential data to third-party services (business data, secrets, .env file, etc.)</li>



<li>Management of local secrets (Bearer token)</li>



<li>Sharing credentials with third-party services (via OAuth mandate, etc.)</li>
</ul>



<h3 class="wp-block-heading" id="measures-to-be-implemented-2">Measures to implement</h3>



<ul class="wp-block-list">
<li>User awareness</li>



<li>Application configuration controls</li>



<li>Software testing and validation</li>



<li>Sandboxing of agents</li>



<li>Protection of secrets (environment file in development directories)</li>
</ul>



<h2 class="wp-block-heading">Remote Agents</h2>



<p>Remote agents, like local agents, are applications that connect different resources (LLM, RAG, third-party services), packaged within a web application, accessible to the user through their browser. All chatbot services are gradually integrating these capabilities to enrich their service by connecting to third-party services. The operation is similar to local agents, but outside the user’s workstation.</p>



<figure class="wp-block-image size-full"><img loading="lazy" decoding="async" width="1024" height="756" src="https://blog.ovhcloud.com/wp-content/uploads/2025/12/mermaid-1767008664972.png" alt="" class="wp-image-30156" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/12/mermaid-1767008664972.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2025/12/mermaid-1767008664972-300x221.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/12/mermaid-1767008664972-768x567.png 768w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p>In this case, the main challenge is managing access to third-party services and the resulting secrets. Since the agent is the focal point of the architecture, entrusting its management to a third party requires granting them access rights to third-party services to capitalise on the agent’s functionality.</p>



<p>In the example above, the user must grant the agent an access mandate to consume the MCPs that allow access to application services. Today, most of these mandates are managed by OAuth2 delegations, with the user authorising the agent to use these technical delegations to access applications.</p>



<h3 class="wp-block-heading">Examples</h3>



<ul class="wp-block-list">
<li>ChatGPT, MistralAI</li>



<li>Agents deployed internally</li>
</ul>



<h3 class="wp-block-heading">Associated security risks</h3>



<ul class="wp-block-list">
<li>Leakage of authentication secrets to sensitive applications of data</li>



<li>Centralisation of secrets to access remote services</li>



<li>Opening of network flows between sensitive applications and agent services</li>
</ul>



<h3 class="wp-block-heading">Measures to implement</h3>



<ul class="wp-block-list">
<li>Architecture to limit network exposure</li>



<li>Network inspection</li>



<li>Application monitoring</li>



<li>Authorisation and access control management</li>



<li>Restriction of access rights to need-to-know for each task</li>
</ul>



<h2 class="wp-block-heading">Workflow agents</h2>



<p>Workflow agent tools are designed to build AI workflows. They may be local or remote. While all wrong behaviours listed above remain possible in this model, the workflow structure  splits the workflow into small manageable parts, allowing:</p>



<ul class="wp-block-list">
<li>Limit of each agent’s access rights to the required sub-set of data for performing its tasks</li>



<li>More deterministic approach for human control over the process</li>



<li>Unitary testing for each parts</li>



<li>Repeatability of the process (workflows are defined ‘as code’)</li>
</ul>



<figure class="wp-block-image size-full"><img loading="lazy" decoding="async" width="1024" height="368" src="https://blog.ovhcloud.com/wp-content/uploads/2025/12/mermaid-1767008664986.png" alt="" class="wp-image-30157" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/12/mermaid-1767008664986.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2025/12/mermaid-1767008664986-300x108.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/12/mermaid-1767008664986-768x276.png 768w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p>In this case, the workflow is built and operates as an automation under the control of a project team in charge of aligning the workflow with business processes. The configuration of the workflow management tools is the key to controlling the process. The orchestration platform manages the secrets and flow to resources, so it need to be managed with proper attention as any orchestration platform.</p>



<h3 class="wp-block-heading" id="examples-6">Examples</h3>



<ul class="wp-block-list">
<li>N8N, Langchain, Zapier, Flowise AI</li>
</ul>



<h3 class="wp-block-heading" id="risks-4">Associated security risks</h3>



<ul class="wp-block-list">
<li>Increase in complexity of the workflows and interconnection</li>



<li>Configuration issues</li>



<li>Leak of access token</li>



<li>Exposure of sensitive resources</li>



<li>Shadow orchestration platforms deployed by users</li>



<li>Access to temporary artifacts by platform administrators</li>
</ul>



<h3 class="wp-block-heading" id="measures-to-be-implemented-4">Measures to be implemented</h3>



<ul class="wp-block-list">
<li>Architecture to limit network exposure</li>



<li>Network inspection</li>



<li>Application monitoring</li>



<li>Authorisation and access control management</li>



<li>Secrets management</li>



<li>Restriction of access rights to need-to-know for each task</li>
</ul>



<h1 class="wp-block-heading" id="perspectives-and-problems-to-be-solved">Perspectives and problems to be solved</h1>



<h2 class="wp-block-heading" id="mcp-and-secret-management">MCP and secret management</h2>



<p>Secret management is at the heart of the problem of deploying agent-based AI. Since LLMs are not deterministic, it is necessary to constrain access rights in terms of scope and duration for LLMs, in order to limit their access to only the data and functions required to perform tasks. It is essential to identify the reliable blocks that will act as intermediaries to grant access, particularly for MCP servers. One of the challenges is to rely on existing access rights matrices without re-implementing an additional layer of rights management for MCP servers and agents, but instead implementing mechanisms to limit access dynamically as needed.</p>



<p>Existing or emerging standards (OAuth2, JWT, SAML, SPIFFE/SPIRE, OPA, Cedar, etc.) partially address some of these challenges, but at the cost of high management complexity, without a reference implementation compatible with all current solutions, and in a rapidly evolving market.</p>



<h2 class="wp-block-heading" id="human-in-the-loop">Human in the loop</h2>



<p>Beyond secret management, LLMs are unpredictable because they are non-deterministic. One of the questions to be resolved is how to include humans in the decision-making chain of an agent-based process to ensure that this inherently unpredictable behaviour does not generate risks for organisations. Today, this control, known as ‘human in the loop’, is based on the agent’s internal mechanisms and the limitation of secrets shared with it by the user. Obviously, this mode of operation is not compatible with sensitive processing.</p>



<p>In the future, it will be necessary to build agents that offer a high level of trust, provided by trusted editors or communities, auditable and audited, ideally open-source, to entrust these agents with performing operations on a company’s information system. In parallel, it will be necessary to develop independent agent control mechanisms that ensure sandboxing, filtering, access management, and traceability functions, allowing the responsible user to master their interaction with the information system.</p>



<h2 class="wp-block-heading" id="towards-the-end-of-the-web-browser-as-a-access-vector-to-the-information-system">Towards the end of the web browser as a access vector to the information system</h2>



<p>For about 15 years, the web browser has been the user’s entry point to information systems. While the functional richness of browsers is immense, the attack surface they expose is just as great. Browser security, even if it is perfectible, is one of the pillars of modern security, and browser editors and communities devote a significant part of their development and maintenance efforts to maintaining the level of security and managing threats.</p>



<p>AI agents are changing this access paradigm to the information system by providing users with dynamic and adaptive interfaces, enriched with high-value contextual functions, which is already causing a revolution in usage and the daily lives of users. It is likely that tomorrow’s browser will be an AI agent, and even more likely that current browsers will gradually become AI agents, integrating all identity and authorisation management standards under user control.</p>
<img loading="lazy" decoding="async" src="//blog.ovhcloud.com/wp-content/plugins/matomo/app/matomo.php?idsite=1&amp;rec=1&amp;url=https%3A%2F%2Fblog.ovhcloud.com%2Fagentic-ai-from-a-security-perspective%2F&amp;action_name=Agentic%20AI%20from%20a%20security%20perspective&amp;urlref=https%3A%2F%2Fblog.ovhcloud.com%2Ffeed%2F" style="border:0;width:0;height:0" width="0" height="0" alt="" />]]></content:encoded>
					
		
		
			</item>
		<item>
		<title>PostgreSQL and AI: The pragmatic path to smarter data</title>
		<link>https://blog.ovhcloud.com/postgresql-and-ai-the-pragmatic-path-to-smarter-data/</link>
		
		<dc:creator><![CDATA[Jonathan Clarke]]></dc:creator>
		<pubDate>Thu, 11 Dec 2025 15:11:00 +0000</pubDate>
				<category><![CDATA[Accelerating with OVHcloud]]></category>
		<category><![CDATA[AI]]></category>
		<category><![CDATA[aiven]]></category>
		<category><![CDATA[Artificial Intelligence]]></category>
		<category><![CDATA[Managed Database]]></category>
		<guid isPermaLink="false">https://blog.ovhcloud.com/?p=30100</guid>

					<description><![CDATA[Beyond the buzz: Building AI on solid foundations Artificial intelligence has quickly become the cornerstone of digital innovation. From text generation to image recognition and intelligent automation, AI is redefining how organisations extract value from data. At OVHcloud, we believe this transformation shouldn’t only belong to the tech elite &#8211; it should be open, accessible, [&#8230;]<img src="//blog.ovhcloud.com/wp-content/plugins/matomo/app/matomo.php?idsite=1&amp;rec=1&amp;url=https%3A%2F%2Fblog.ovhcloud.com%2Fpostgresql-and-ai-the-pragmatic-path-to-smarter-data%2F&amp;action_name=PostgreSQL%20and%20AI%3A%20The%20pragmatic%20path%20to%20smarter%20data&amp;urlref=https%3A%2F%2Fblog.ovhcloud.com%2Ffeed%2F" style="border:0;width:0;height:0" width="0" height="0" alt="" />]]></description>
										<content:encoded><![CDATA[
<h2 class="wp-block-heading"><strong>Beyond the buzz: Building AI on solid foundations</strong></h2>



<p><strong>Artificial intelligence</strong> has quickly become the cornerstone of digital innovation. From text generation to image recognition and intelligent automation, AI is redefining how organisations extract value from data.</p>



<p>At OVHcloud, we believe this transformation shouldn’t only belong to the tech elite &#8211; it should be open, accessible, and built on trusted, sovereign infrastructure.</p>



<p>This vision drives everything from our <a href="https://www.ovhcloud.com/en-ie/public-cloud/ai-endpoints/" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer"><strong>AI Endpoints</strong></a> and <a href="https://www.ovhcloud.com/en-ie/public-cloud/ai-deploy/" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer"><strong>AI Deploy</strong></a> solutions to our <a href="https://huggingface.co/blog/OVHcloud/inference-providers-ovhcloud" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer"><strong>Hugging Face partnership</strong></a>, which empowers developers to run open <strong>inference</strong> models directly in the cloud. But beyond those flagship initiatives, AI also lives in the everyday – in the data that powers recommendations, insights and smarter user experiences.</p>



<p>And that’s where <strong>PostgreSQL + Vector capabilities</strong> come in.</p>



<figure class="wp-block-image size-full"><img loading="lazy" decoding="async" width="841" height="561" src="https://blog.ovhcloud.com/wp-content/uploads/2025/02/Image1.png" alt="" class="wp-image-28243" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/02/Image1.png 841w, https://blog.ovhcloud.com/wp-content/uploads/2025/02/Image1-300x200.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/02/Image1-768x512.png 768w" sizes="auto, (max-width: 841px) 100vw, 841px" /></figure>



<h2 class="wp-block-heading"><strong>Vectors: Where data meets understanding</strong></h2>



<p>At its core, AI systems function by decoding relationships between words, images or user behaviours. To do that, machine learning models translate these entities <strong>into vectors </strong>— mathematical representations that capture meaning and similarity.</p>



<p>A vector representation allows a system to measure how close two pieces of data are. It is the foundation of <strong>semantic search</strong>, <strong>recommendation engines</strong>, <strong>facial recognition</strong> and <strong>anomaly detection systems</strong>.</p>



<p>Traditionally, companies needed to move their datasets from transactional databases into specialised “vector databases.” While vector databases are effective for purely vector-centric workloads, this approach often comes with <strong>higher complexity</strong>, data duplication, and <strong>integration</strong> <strong>overhead</strong>. These challenges are not ideal for production-grade systems that demand reliability and compliance.</p>



<figure class="wp-block-image aligncenter size-full"><img loading="lazy" decoding="async" width="373" height="355" src="https://blog.ovhcloud.com/wp-content/uploads/2025/12/image-2.png" alt="" class="wp-image-30101" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/12/image-2.png 373w, https://blog.ovhcloud.com/wp-content/uploads/2025/12/image-2-300x286.png 300w" sizes="auto, (max-width: 373px) 100vw, 373px" /></figure>



<div style="height:20px" aria-hidden="true" class="wp-block-spacer"></div>



<h2 class="wp-block-heading"><strong>PostgreSQL + pgvector: AI where your data already lives</strong></h2>



<p>Instead of creating yet another database to maintain, <a href="https://www.ovhcloud.com/en-ie/public-cloud/postgresql/" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer"><strong>PostgreSQL</strong></a> offers an elegant solution: the <strong><em>pgvector </em>extension</strong>. With pgvector, organisations can store, query and compare vectorised data alongside traditional relational data, using the same SQL syntax they already know. pgvector also allows you to build full or partial indexes to speed up similarity search.</p>



<p>In other words, PostgreSQL becomes not just your source of truth, but also your foundation for AI experimentation and delivery.</p>



<p>Here’s what this means in practice:</p>



<ul class="wp-block-list">
<li><strong>Simplified architecture</strong>: Keep data in one place. No ETL pipelines or synchronisation risks.</li>



<li><strong>Familiar SQL workflow</strong>: Run similarity searches directly in SQL, with ACID guarantees intact.</li>



<li><strong>Faster time to value</strong>: Build and iterate AI use cases faster, without learning a new database technology.</li>
</ul>



<p>This is AI grounded in operational reality — a pragmatic path for enterprises to explore machine learning use cases <strong>safely</strong> and <strong>efficiently</strong>.</p>



<h2 class="wp-block-heading"><strong>A practical use case: Real-time product recommendations</strong></h2>



<p>Imagine an e-commerce company managing both product and customer data in <a href="https://www.ovhcloud.com/en-ie/public-cloud/postgresql/" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer"><strong>Managed PostgreSQL</strong></a> at OVHcloud, ensuring access to the latest, most performant features.</p>



<p>By combining <em>pgvector</em> with embeddings generated from an open-source model, the team can:</p>



<ol class="wp-block-list">
<li>Convert product descriptions and user preferences into vector representations.</li>



<li>Store these vectors in PostgreSQL columns alongside stock levels, pricing and metadata.</li>



<li>Run a <strong>similarity search</strong> that finds relevant products instantly: for example, recommending ‘eco-friendly alternatives’ or ‘similar styles’ while ensuring only in-stock items are shown.</li>
</ol>



<p>The entire process happens within PostgreSQL — <strong>no need for external vector databases or data duplication</strong>.</p>



<p><strong>The result</strong>: real-time, AI-enhanced customer experiences powered by trusted, open technology.</p>



<figure class="wp-block-image aligncenter size-full"><img loading="lazy" decoding="async" width="364" height="262" src="https://blog.ovhcloud.com/wp-content/uploads/2025/12/image-3.png" alt="" class="wp-image-30102" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/12/image-3.png 364w, https://blog.ovhcloud.com/wp-content/uploads/2025/12/image-3-300x216.png 300w" sizes="auto, (max-width: 364px) 100vw, 364px" /></figure>



<div style="height:20px" aria-hidden="true" class="wp-block-spacer"></div>



<h2 class="wp-block-heading"><strong>The enterprise reality: AI without reinventing the wheel</strong></h2>



<p>In the rush to ‘go AI’, many <strong>organisations risk overcomplicating</strong> their architectures by chasing the latest dedicated vector databases. While those solutions have their place, <strong>PostgreSQL</strong>’s maturity, ecosystem and extensibility make it uniquely suited for the vast majority of enterprise AI workloads.</p>



<p>For most companies exploring AI, starting with what they already know, PostgreSQL, means <strong>solid foundations</strong>,<strong> less risk</strong>,<strong> faster learning </strong>and<strong> lower cost</strong>.</p>



<h2 class="wp-block-heading"><strong>The OVHcloud advantage: Open, managed, secure</strong></h2>



<p>OVHcloud’s partnership with <a href="https://aiven.io/" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer"><strong>Aiven</strong></a>, which brings proven expertise in managing PostgreSQL at scale, ensures our customers benefit from the latest capabilities as soon as they are production-ready, without operational difficulties. Let your teams <strong>focus on their product</strong> rather than worry about database resources and infrastructure.</p>



<p>Additionally, OVHcloud customers can benefit from a service-level agreement (SLA) of up to 99.99% via its <a href="https://help.ovhcloud.com/csm/en-public-cloud-databases-migrate-1az-to-3az?id=kb_article_view&amp;sysparm_article=KB0072137" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">Multi-Availability Zone</a> (3-AZ) regions. These regions feature geographically separated zones with independent power, cooling and network systems, providing true fault isolation.</p>



<figure class="wp-block-image aligncenter size-full"><img loading="lazy" decoding="async" width="345" height="345" src="https://blog.ovhcloud.com/wp-content/uploads/2025/12/image-4.png" alt="" class="wp-image-30103" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/12/image-4.png 345w, https://blog.ovhcloud.com/wp-content/uploads/2025/12/image-4-300x300.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/12/image-4-150x150.png 150w, https://blog.ovhcloud.com/wp-content/uploads/2025/12/image-4-70x70.png 70w" sizes="auto, (max-width: 345px) 100vw, 345px" /></figure>



<div style="height:20px" aria-hidden="true" class="wp-block-spacer"></div>



<p>At <strong>OVHcloud</strong>, we see PostgreSQL as more than a database. It’s a bridge between today’s workloads and tomorrow’s intelligent systems. And as AI workloads evolve, we’ll continue to integrate the technologies that matter: from <strong>vector search</strong> and <strong>AI embeddings</strong> to seamless connections with <strong>AI Endpoints</strong> and <strong>Hugging Face models</strong>.</p>
<img loading="lazy" decoding="async" src="//blog.ovhcloud.com/wp-content/plugins/matomo/app/matomo.php?idsite=1&amp;rec=1&amp;url=https%3A%2F%2Fblog.ovhcloud.com%2Fpostgresql-and-ai-the-pragmatic-path-to-smarter-data%2F&amp;action_name=PostgreSQL%20and%20AI%3A%20The%20pragmatic%20path%20to%20smarter%20data&amp;urlref=https%3A%2F%2Fblog.ovhcloud.com%2Ffeed%2F" style="border:0;width:0;height:0" width="0" height="0" alt="" />]]></content:encoded>
					
		
		
			</item>
		<item>
		<title>10 Reasons Scaling Startups Are Migrating to OVHcloud</title>
		<link>https://blog.ovhcloud.com/10-reasons-scaling-startups-are-migrating-to-ovhcloud/</link>
		
		<dc:creator><![CDATA[Alexander Grau]]></dc:creator>
		<pubDate>Tue, 21 Oct 2025 21:26:37 +0000</pubDate>
				<category><![CDATA[OVHcloud Startup Program]]></category>
		<category><![CDATA[Uncategorized]]></category>
		<category><![CDATA[AI]]></category>
		<category><![CDATA[Data Sovereignty]]></category>
		<category><![CDATA[Infrastructure]]></category>
		<category><![CDATA[Kubernetes]]></category>
		<category><![CDATA[Migration]]></category>
		<category><![CDATA[OVHcloud]]></category>
		<category><![CDATA[Public Cloud]]></category>
		<category><![CDATA[Startup Program]]></category>
		<guid isPermaLink="false">https://blog.ovhcloud.com/?p=28941</guid>

					<description><![CDATA[Cloud infrastructure plays a critical role in how startups scale—affecting everything from product delivery and user experience to budget and compliance. While many startups begin their journey with public cloud giants, the challenges of unpredictable costs, data control, and technical constraints become more apparent as they grow. For startups ready to scale smarter, OVHcloud offers [&#8230;]<img src="//blog.ovhcloud.com/wp-content/plugins/matomo/app/matomo.php?idsite=1&amp;rec=1&amp;url=https%3A%2F%2Fblog.ovhcloud.com%2F10-reasons-scaling-startups-are-migrating-to-ovhcloud%2F&amp;action_name=10%20Reasons%20Scaling%20Startups%20Are%20Migrating%20to%20OVHcloud&amp;urlref=https%3A%2F%2Fblog.ovhcloud.com%2Ffeed%2F" style="border:0;width:0;height:0" width="0" height="0" alt="" />]]></description>
										<content:encoded><![CDATA[
<p>Cloud infrastructure plays a critical role in how startups scale—affecting everything from product delivery and user experience to budget and compliance. While many startups begin their journey with public cloud giants, the challenges of unpredictable costs, data control, and technical constraints become more apparent as they grow.</p>



<p>For startups ready to scale smarter, <strong>OVHcloud</strong> offers a compelling alternative: high-performance, cost-effective, and sovereignty-first infrastructure. Here’s why more and more growing startups are making the switch.</p>



<hr class="wp-block-separator has-alpha-channel-opacity"/>



<h5 class="wp-block-heading"><strong>1. Predictable, Transparent Pricing</strong></h5>



<p>OVHcloud’s <a href="https://www.ovhcloud.com/en-gb/lp/prices-give-you-edge/" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">flat-rate pricing model</a> eliminates hidden fees and unpredictable billing. Bandwidth is included. Egress costs? Zero. This gives startups the ability to budget confidently—even as infrastructure scales rapidly.</p>



<h5 class="wp-block-heading"><strong>2. Cost-Efficient Scaling</strong></h5>



<p>Startups that migrate to OVHcloud often report <strong>up to 60% cost savings</strong> compared to hyperscalers. Whether you&#8217;re scaling your backend, AI workloads, or customer-facing applications, OVHcloud lets you do more with less.</p>



<h5 class="wp-block-heading"><strong>3. Performance Without Compromise</strong></h5>



<p>From <a href="https://www.ovhcloud.com/en/bare-metal/" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">Bare Metal</a> servers and <a href="https://www.ovhcloud.com/en/public-cloud/virtual-instances/" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">high-memory VMs</a> to <a href="https://www.ovhcloud.com/en/public-cloud/gpu/" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">GPU</a> and <a href="https://www.ovhcloud.com/en/public-cloud/storage/" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">storage-optimized instances</a>, OVHcloud infrastructure is engineered for performance. Ideal for AI, SaaS, analytics, and other compute-intensive use cases.</p>



<h5 class="wp-block-heading"><strong>4. Full Data Sovereignty in the EU</strong></h5>



<p>OVHcloud is headquartered in Europe and operates under <strong>the strictest data protection laws (like GDPR in the EU or Law 25 in Quebec, Canada)</strong>. Unlike other providers, <a href="https://www.ovhcloud.com/en-gb/lp/data-stays-your-data/" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">your data stays</a> within jurisdictions that respect privacy and sovereignty—no exposure to foreign surveillance laws.</p>



<h5 class="wp-block-heading"><strong>5. Open Standards and No Vendor Lock-In</strong></h5>



<p>Freedom matters—especially when you’re building for scale. OVHcloud supports open technologies like <a href="https://www.ovhcloud.com/en/public-cloud/kubernetes/" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">Kubernetes</a>, Terraform, and OpenStack, giving your team full flexibility and avoiding lock-in to proprietary tools or services.</p>



<h5 class="wp-block-heading"><strong>6. Infrastructure That Grows With You</strong></h5>



<p>Whether you&#8217;re launching in new markets or onboarding thousands of new users, OVHcloud enables seamless horizontal and vertical scaling. With availability across multiple regions, your growth won’t hit a wall.</p>



<h5 class="wp-block-heading"><strong>7. Faster Time-to-Market Through Cloud Migration Support</strong></h5>



<p>OVHcloud offers <strong>cloud migration guidance and tools</strong>, including compatibility with major platforms and migration kits. This helps your team move faster, avoid downtime, and focus on innovation—not infrastructure headaches.</p>



<h5 class="wp-block-heading"><strong>8. Dev-Friendly Ecosystem</strong></h5>



<p>With support for containerization, automation, and CI/CD pipelines, OVHcloud makes life easier for DevOps teams. You can provision infrastructure programmatically and scale efficiently—just like you would with AWS or GCP.</p>



<h5 class="wp-block-heading"><strong>9. Sustainability Built In</strong></h5>



<p>Efficiency is built into OVHcloud’s DNA. By designing and operating its own energy-efficient data centers, OVHcloud helps startups meet their <a href="https://www.ovhcloud.com/en-gb/lp/sustainable-ground/" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">sustainability goals</a> without compromising on performance or cost.</p>



<h5 class="wp-block-heading"><strong>10. A Cloud Partner—Not Just a Provider</strong></h5>



<p>Startups need more than compute power—they need guidance, flexibility, and a partner that understands their journey. OVHcloud offers <a href="https://www.ovhcloud.com/en/professional-services/" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">tailored support</a>, <a href="https://help.ovhcloud.com/csm/world-documentation?id=kb_home" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">technical documentation</a>, and real human engagement to help you succeed at every stage of growth.</p>



<hr class="wp-block-separator has-alpha-channel-opacity"/>



<h5 class="wp-block-heading"><strong><em>OVHcloud: Built to Scale With Startups</em></strong></h5>



<p>If your startup is growing fast and needs infrastructure that can keep up—without breaking the budget or sacrificing control—OVHcloud offers a cloud built around your values: <strong>scalability, transparency, freedom, and performance</strong>.</p>



<p><strong>Migrate with confidence. Scale with control. Grow with OVHcloud.</strong></p>



<hr class="wp-block-separator has-alpha-channel-opacity"/>



<figure class="wp-block-image size-full"><a href="https://startup.ovhcloud.com/en/globalreport2025/" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer"><img loading="lazy" decoding="async" width="970" height="250" src="https://blog.ovhcloud.com/wp-content/uploads/2025/08/Email-Signature-–-1.jpg" alt="" class="wp-image-29527" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/08/Email-Signature-–-1.jpg 970w, https://blog.ovhcloud.com/wp-content/uploads/2025/08/Email-Signature-–-1-300x77.jpg 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/08/Email-Signature-–-1-768x198.jpg 768w" sizes="auto, (max-width: 970px) 100vw, 970px" /></a></figure>



<p>If you’re a startup looking to transform your business, we encourage you to join the <strong><a href="https://startup.ovhcloud.com/en/" target="_blank" rel="noreferrer noopener nofollow external" data-wpel-link="external">OVHcloud Startup Program</a></strong> or contact OVHcloud to discover how our solutions can support your journey!</p>
<img loading="lazy" decoding="async" src="//blog.ovhcloud.com/wp-content/plugins/matomo/app/matomo.php?idsite=1&amp;rec=1&amp;url=https%3A%2F%2Fblog.ovhcloud.com%2F10-reasons-scaling-startups-are-migrating-to-ovhcloud%2F&amp;action_name=10%20Reasons%20Scaling%20Startups%20Are%20Migrating%20to%20OVHcloud&amp;urlref=https%3A%2F%2Fblog.ovhcloud.com%2Ffeed%2F" style="border:0;width:0;height:0" width="0" height="0" alt="" />]]></content:encoded>
					
		
		
			</item>
		<item>
		<title>Fine tune an LLM with Axolotl and OVHcloud Machine Learning Services</title>
		<link>https://blog.ovhcloud.com/fine-tune-an-llm-with-axolotl-and-ovhcloud-machine-learning-services/</link>
		
		<dc:creator><![CDATA[Stéphane Philippart]]></dc:creator>
		<pubDate>Fri, 25 Jul 2025 13:07:40 +0000</pubDate>
				<category><![CDATA[OVHcloud Engineering]]></category>
		<category><![CDATA[Tranches de Tech & co]]></category>
		<category><![CDATA[AI]]></category>
		<category><![CDATA[AI Deploy]]></category>
		<category><![CDATA[AI Notebook]]></category>
		<category><![CDATA[Fine Tuning]]></category>
		<category><![CDATA[OVHcloud]]></category>
		<guid isPermaLink="false">https://blog.ovhcloud.com/?p=29408</guid>

					<description><![CDATA[There are many ways to train a model,📚 using detailed instructions, system prompts, Retrieval Augmented Generation, or function calling One way is fine-tuning, which is what this blog is about! ✨ Two years back we posted a blog on fine-tuning Llama models—it’s not nearly as complicated as it was before 😉.  This time we’re using the [&#8230;]<img src="//blog.ovhcloud.com/wp-content/plugins/matomo/app/matomo.php?idsite=1&amp;rec=1&amp;url=https%3A%2F%2Fblog.ovhcloud.com%2Ffine-tune-an-llm-with-axolotl-and-ovhcloud-machine-learning-services%2F&amp;action_name=Fine%20tune%20an%20LLM%20with%20Axolotl%20and%20OVHcloud%20Machine%20Learning%20Services&amp;urlref=https%3A%2F%2Fblog.ovhcloud.com%2Ffeed%2F" style="border:0;width:0;height:0" width="0" height="0" alt="" />]]></description>
										<content:encoded><![CDATA[
<figure class="wp-block-image aligncenter size-full is-resized"><img loading="lazy" decoding="async" width="1024" height="1024" src="https://blog.ovhcloud.com/wp-content/uploads/2025/07/red-cat-02-1.png" alt="A robot with a car tuning style" class="wp-image-29462" style="width:600px" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/07/red-cat-02-1.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2025/07/red-cat-02-1-300x300.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/07/red-cat-02-1-150x150.png 150w, https://blog.ovhcloud.com/wp-content/uploads/2025/07/red-cat-02-1-768x768.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2025/07/red-cat-02-1-70x70.png 70w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<hr class="wp-block-separator has-alpha-channel-opacity"/>



<p>There are many ways to train a model,📚 using detailed instructions, system prompts, Retrieval Augmented Generation, or function calling</p>



<p>One way is fine-tuning, which is what this blog is about! ✨</p>



<p>Two years back we posted a <a href="https://blog.ovhcloud.com/fine-tuning-llama-2-models-using-a-single-gpu-qlora-and-ai-notebooks/" data-wpel-link="internal">blog</a> on fine-tuning Llama models—it’s not nearly as complicated as it was before 😉.  This time we’re using the Framework <a href="https://docs.axolotl.ai/" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">Axolotl</a>, so hopefully there’s less to manage.</p>



<h3 class="wp-block-heading">So what’s the plan?</h3>



<p>For this blog, I’d like to fine-tune a small model, <a href="https://huggingface.co/meta-llama/Llama-3.2-1B-Instruct" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">Llama-3.2-1B-Instruct</a>, and then test it out on a few questions about our <a href="https://endpoints.ai.cloud.ovh.net/" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">OVHcloud AI Endpoints</a> product 📝.</p>



<p>Before we fine-tune, let’s try it out! Deploying a <a href="https://huggingface.co/" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">Hugging Face</a> model is super easy with <a href="https://www.ovhcloud.com/fr/public-cloud/ai-deploy/" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">AI Deploy</a> from <a href="https://www.ovhcloud.com/fr/public-cloud/ai-machine-learning/" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">AI Machine Learning Services</a> 🥳.</p>



<p>And thanks to a <a href="https://blog.ovhcloud.com/mistral-small-24b-served-with-vllm-and-ai-deploy-one-command-to-deploy-llm/" data-wpel-link="internal">previous blog post</a>, we know how to use <a href="https://docs.vllm.ai/en/v0.7.3/index.html" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">vLLM</a> and <a href="https://www.ovhcloud.com/fr/public-cloud/ai-deploy/" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">AI Deploy</a>.</p>



<pre title="Deploy a model thanks to vLLM and AI Deploy" class="wp-block-code"><code lang="bash" class="language-bash line-numbers">ovhai app run --name $1 \
	--flavor l40s-1-gpu \
	--gpu 2 \
	--default-http-port 8000 \
	--env OUTLINES_CACHE_DIR=/tmp/.outlines \
	--env HF_TOKEN=$MY_HUGGING_FACE_TOKEN \
	--env HF_HOME=/hub \
	--env HF_DATASETS_TRUST_REMOTE_CODE=1 \
	--env HF_HUB_ENABLE_HF_TRANSFER=0 \
	--volume standalone:/hub:rw \
	--volume standalone:/workspace:rw \
	vllm/vllm-openai:v0.8.2 \
	-- bash	-c "vllm serve meta-llama/Llama-3.2-1B-Instruct"</code></pre>



<p class="has-text-align-center"><strong><strong>⚠️ Make sure you’ve agreed to the terms of use for the model’s license from Hugging Face ⚠️</strong></strong></p>



<p>Check out the <a href="https://blog.ovhcloud.com/mistral-small-24b-served-with-vllm-and-ai-deploy-one-command-to-deploy-llm/" data-wpel-link="internal">blog</a> I mentioned earlier for all the details you need on the command and its parameters.</p>



<p>To test our different chatbots we will use a simple <a href="https://github.com/ovh/public-cloud-examples/tree/main/ai/llm-fine-tune/chatbot/chatbot.py" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">Gradio application</a>:</p>



<pre title="Chatbot" class="wp-block-code"><code lang="python" class="language-python line-numbers"># Application to compare answers generation from OVHcloud AI Endpoints exposed model and fine tuned model.
# ⚠️ Do not used in production!! ⚠️

import gradio as gr
import os

from langchain_openai import ChatOpenAI
from langchain_core.prompts import ChatPromptTemplate

# 📜 Prompts templates 📜
prompt_template = ChatPromptTemplate.from_messages(
    [
        ("system", "{system_prompt}"),
        ("human", "{user_prompt}"),
    ]
)

def chat(prompt, system_prompt, temperature, top_p, model_name, model_url, api_key):
    """
    Function to generate a chat response using the provided prompt, system prompt, temperature, top_p, model name, model URL and API key.
    """

    # ⚙️ Initialize the OpenAI model ⚙️
    llm = ChatOpenAI(api_key=api_key, 
                 model=model_name, 
                 base_url=model_url,
                 temperature=temperature,
                 top_p=top_p
                 )

    # 📜 Apply the prompt to the model 📜
    chain = prompt_template | llm
    ai_msg = chain.invoke(
        {
            "system_prompt": system_prompt,
            "user_prompt": prompt
        }
    )

    # 🤖 Return answer in a compatible format for Gradio component.
    return [{"role": "user", "content": prompt}, {"role": "assistant", "content": ai_msg.content}]

# 🖥️ Main application 🖥️
with gr.Blocks() as demo:
    with gr.Row():
        with gr.Column():
            system_prompt = gr.Textbox(value="""You are a specialist on OVHcloud products.
If you can't find any sure and relevant information about the product asked, answer with "This product doesn't exist in OVHcloud""", 
                label="🧑‍🏫 System Prompt 🧑‍🏫")
            temperature = gr.Slider(minimum=0.0, maximum=2.0, step=0.01, label="Temperature", value=0.5)
            top_p = gr.Slider(minimum=0.0, maximum=1.0, step=0.01, label="Top P", value=0.0)
            model_name = gr.Textbox(label="🧠 Model Name 🧠", value='Llama-3.1-8B-Instruct')
            model_url = gr.Textbox(label="🔗 Model URL 🔗", value='https://oai.endpoints.kepler.ai.cloud.ovh.net/v1')
            api_key = gr.Textbox(label="🔑 OVH AI Endpoints Access Token 🔑", value=os.getenv("OVH_AI_ENDPOINTS_ACCESS_TOKEN"), type="password")

        with gr.Column():
            chatbot = gr.Chatbot(type="messages", label="🤖 Response 🤖")
            prompt = gr.Textbox(label="📝 Prompt 📝", value='How many requests by minutes can I do with AI Endpoints?')
            submit = gr.Button("Submit")

    submit.click(chat, inputs=[prompt, system_prompt, temperature, top_p, model_name, model_url, api_key], outputs=chatbot)

demo.launch()</code></pre>



<p>ℹ️ You can find all resources to build and run this application in the <a href="https://github.com/ovh/public-cloud-examples/tree/main/ai/llm-fine-tune/chatbot/" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">dedicated folder</a> in the GitHub repository.</p>



<p>Let&#8217;s test with a simple question: &#8220;How many requests by minutes can I do with AI Endpoints?&#8221;.<br>The first test is with <a href="https://huggingface.co/meta-llama/Llama-3.2-1B-Instruct" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">Llama-3.2-1B-Instruct</a> from <a href="https://huggingface.co/" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer"></a><a href="https://huggingface.co/" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">Hugging Face</a> deployed with <a href="https://docs.vllm.ai/en/v0.7.3/index.html" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">vLLM</a> and <a href="https://www.ovhcloud.com/fr/public-cloud/ai-deploy/" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">OVHcloud AI Deploy</a>.</p>



<figure class="wp-block-image aligncenter size-large"><img loading="lazy" decoding="async" width="1024" height="474" src="https://blog.ovhcloud.com/wp-content/uploads/2025/07/Screenshot-2025-07-23-at-13.19.16-1024x474.png" alt="Ask for AI Endpoints rate limit with a Llama-3.2-1B-Instruct model" class="wp-image-29448" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/07/Screenshot-2025-07-23-at-13.19.16-1024x474.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2025/07/Screenshot-2025-07-23-at-13.19.16-300x139.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/07/Screenshot-2025-07-23-at-13.19.16-768x356.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2025/07/Screenshot-2025-07-23-at-13.19.16-1536x712.png 1536w, https://blog.ovhcloud.com/wp-content/uploads/2025/07/Screenshot-2025-07-23-at-13.19.16-2048x949.png 2048w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p>The response isn’t exactly what we expected. 😅</p>



<p>FYI, according to the official <a href="https://help.ovhcloud.com/csm/fr-public-cloud-ai-endpoints-capabilities?id=kb_article_view&amp;sysparm_article=KB0065424#limitations" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">OVHcloud guide</a>, the correct answer is:<br> &#8211; <strong>Anonymous</strong>: 2 requests per minute, per IP and per model.<br> &#8211; <strong>Authenticated with an API access key</strong>: 400 requests per minute, per Public Cloud project and per model.</p>



<h3 class="wp-block-heading"><strong>What’s the best way to feed the model fresh data?</strong></h3>



<p>I bet you already know this—you can use some data during the inference step, using Retrieval Augmented Generation (RAG). You can learn how to set up RAG by reading our <a href="https://blog.ovhcloud.com/rag-chatbot-using-ai-endpoints-and-langchain/" data-wpel-link="internal">past blog post</a>. 📗</p>



<p>Another way to feed a model fresh data by fine-tuning. ✨</p>



<p>In a nutshell,  fine-tuning is when you take a pre-trained machine learning model and train it further on additional data, so it can do a specific job. It’s quicker and easier than building a model yourself, or from scratch. 😉</p>



<p>For this, I’m picking <a href="https://huggingface.co/meta-llama/Llama-3.2-1B-Instruct" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">Llama-3.2-1B-Instruct</a> from Hugging Face as the base model.</p>



<p><em>ℹ️ The more parameters your base model has, the more computing power you need. In this case, this model needs between 3GB and 4GB of memory, <em>which is why we’ll be using</em> a <a href="https://www.ovhcloud.com/fr/public-cloud/prices/#5260" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">single L4 GPU</a> (we need </em><a href="https://www.nvidia.com/en-us/data-center/ampere-architecture/" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer"><em>Ampere</em> compatible architecture</a>).</p>



<h3 class="wp-block-heading">When data is your gold</h3>



<p>To train a model, you need enough good-quality data.</p>



<p>The first part is easy; I get the OVHcloud AI Endpoints official documentation in a markdown format from our <a href="https://github.com/ovh/docs/tree/develop/pages/public_cloud/ai_machine_learning" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">public cloud documentation repository</a> (by the way, would you like to contribute?). 📚</p>



<p>First, create a dataset with the right format, Axolotl offers varying <a href="https://docs.axolotl.ai/docs/dataset-formats/" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">dataset formats</a>. I prefer the <a href="https://docs.axolotl.ai/docs/dataset-formats/conversation.html" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">conversation format</a> because it’s the easiest for my use case, so I’m going with that. 😉</p>



<pre title="Conersation format dataset" class="wp-block-code"><code lang="json" class="language-json line-numbers"><a href="https://docs.axolotl.ai/docs/dataset-formats/conversation.html#cb1-1" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer"></a>{
   "messages": [
     {"role": "...", "content": "..."}, 
     {"role": "...", "content": "..."}, 
     ...]
}</code></pre>



<p>And to create it manually and add the relevant information, I use an LLM to convert the markdown data into a well-formed dataset. 🤖</p>



<p>Here we’re using <a href="https://github.com/ovh/public-cloud-examples/tree/main/ai/llm-fine-tune/ai/llm-fine-tune/dataset/DatasetCreation.py" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">Python script </a>🐍:</p>



<pre title="Dataset creation with LLM" class="wp-block-code"><code lang="python" class="language-python line-numbers">import os
from pathlib import Path
from langchain_openai import ChatOpenAI
from langchain.schema import HumanMessage

# 🗺️ Define the JSON schema for the response 🗺️
message_schema = {
    "type": "object",
    "properties": {
        "role": {"type": "string"},
        "content": {"type": "string"}
    },
    "required": ["role", "content"]
}

response_format = {
    "type": "json_object",
    "json_schema": {
        "name": "Messages",
        "description": "A list of messages with role and content",
        "properties": {
            "messages": {
                "type": "array",
                "items": message_schema
            }
        }
    }
}

# ⚙️ Initialize the chat model with AI Endpoints configuration ⚙️
chat_model = ChatOpenAI(
    api_key=os.getenv("OVH_AI_ENDPOINTS_ACCESS_TOKEN"),
    base_url=os.getenv("OVH_AI_ENDPOINTS_MODEL_URL"),
    model_name=os.getenv("OVH_AI_ENDPOINTS_MODEL_NAME"),
    temperature=0.0
)

# 📂 Define the directory path 📂
directory_path = "docs/pages/public_cloud/ai_machine_learning"
directory = Path(directory_path)

# 🗃️ Walk through the directory and its subdirectories 🗃️
for path in directory.rglob("*"):
    # Check if the current path is a directory
    if path.is_dir():
        # Get the name of the subdirectory
        sub_directory = path.name

        # Construct the path to the "guide.en-gb.md" file in the subdirectory
        guide_file_path = path / "guide.en-gb.md"

        # Check if the "guide.en-gb.md" file exists in the subdirectory
        if "endpoints" in sub_directory and guide_file_path.exists():
            print(f"📗 Guide processed: {sub_directory}")
            with open(guide_file_path, 'r', encoding='utf-8') as file:
                raw_data = file.read()

            user_message = HumanMessage(content=f"""
With the markdown following, generate a JSON file composed as follows: a list named "messages" composed of tuples with a key "role" which can have the value "user" when it's the question and "assistant" when it's the response. To split the document, base it on the markdown chapter titles to create the question, seems like a good idea.
Keep the language English.
I don't need to know the code to do it but I want the JSON result file.
For the "user" field, don't just repeat the title but make a real question, for example "What are the requirements for OVHcloud AI Endpoints?"
Be sure to add OVHcloud with AI Endpoints so that it's clear that OVHcloud creates AI Endpoints.
Generate the entire JSON file.
An example of what it should look like: messages [{{"role":"user", "content":"What is AI Endpoints?"}}]
There must always be a question followed by an answer, never two questions or two answers in a row.
The source markdown file:
{raw_data}
""")
            chat_response = chat_model.invoke([user_message], response_format=response_format)
            
            with open(f"./generated/{sub_directory}.json", 'w', encoding='utf-8') as output_file:
                output_file.write(chat_response.content)
                print(f"✅ Dataset generated: ./generated/{sub_directory}.json")

</code></pre>



<p><em>ℹ️ You can find all resources to build and run this application in the <a href="https://github.com/ovh/public-cloud-examples/tree/main/ai/llm-fine-tune/dataset/" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">dedicated folder</a> in the GitHub repository.</em></p>



<p>Here’s a sample of the file created as the dataset:</p>



<pre title="Dataset example" class="wp-block-code"><code lang="json" class="language-json line-numbers">[
  {
    "role": "user",
    "content": "What are the requirements for using OVHcloud AI Endpoints?"
  },
  {
    "role": "assistant",
    "content": "To use OVHcloud AI Endpoints, you need the following: \n1. A Public Cloud project in your OVHcloud account \n2. A payment method defined on your Public Cloud project. Access keys created from Public Cloud projects in Discovery mode (without a payment method) cannot use the service."
  },
  {
    "role": "user",
    "content": "What are the rate limits for using OVHcloud AI Endpoints?"
  },
  {
    "role": "assistant",
    "content": "The rate limits for OVHcloud AI Endpoints are as follows:\n- Anonymous: 2 requests per minute, per IP and per model.\n- Authenticated with an API access key: 400 requests per minute, per PCI project and per model."
  }, 
   ...]
}</code></pre>



<p>As for quantity, it’s a bit tricky. How can we generate the right data for training without lowering data quality?</p>



<p>To do this, I’ve created synthetic data using an LLM to create it from the original data. The trick is to generate more data on the same topic by rephrasing it differently but with the same idea.</p>



<p>Here is the <a href="https://github.com/ovh/public-cloud-examples/tree/main/ai/llm-fine-tune/dataset/DatasetAugmentation.py" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">Python script</a> 🐍 to do the data augmentation:</p>



<pre title="Data augmentation" class="wp-block-code"><code lang="python" class="language-python line-numbers">import os
import json
import uuid
from pathlib import Path
from langchain_openai import ChatOpenAI
from langchain.schema import HumanMessage
from jsonschema import validate, ValidationError

# 🗺️ Define the JSON schema for the response 🗺️
message_schema = {
    "type": "object",
    "properties": {
        "role": {"type": "string"},
        "content": {"type": "string"}
    },
    "required": ["role", "content"]
}

response_format = {
    "type": "json_object",
    "json_schema": {
        "name": "Messages",
        "description": "A list of messages with role and content",
        "properties": {
            "messages": {
                "type": "array",
                "items": message_schema
            }
        }
    }
}

# ✅ JSON validity verification ❌
def is_valid(json_data):
    """
    Test the validity of the JSON data against the schema.
    Argument:
        json_data (dict): The JSON data to validate.  
    Raises:
        ValidationError: If the JSON data does not conform to the specified schema.  
    """
    try:
        validate(instance=json_data, schema=response_format["json_schema"])
        return True
    except ValidationError as e:
        print(f"❌ Validation error: {e}")
        return False

# ⚙️ Initialize the chat model with AI Endpoints configuration ⚙️
chat_model = ChatOpenAI(
    api_key=os.getenv("OVH_AI_ENDPOINTS_ACCESS_TOKEN"),
    base_url=os.getenv("OVH_AI_ENDPOINTS_MODEL_URL"),
    model_name=os.getenv("OVH_AI_ENDPOINTS_MODEL_NAME"),
    temperature=0.0
)

# 📂 Define the directory path 📂
directory_path = "generated"
print(f"📂 Directory path: {directory_path}")
directory = Path(directory_path)

# 🗃️ Walk through the directory and its subdirectories 🗃️
for path in directory.rglob("*"):
    print(f"📜 Processing file: {path}")
    # Check if the current path is a valid file
    if path.is_file() and path.name.__contains__ ("endpoints"):
        # Read the raw data from the file
        with open(path, 'r', encoding='utf-8') as file:
            raw_data = file.read()

        try:
            json_data = json.loads(raw_data)
        except json.JSONDecodeError:
            print(f"❌ Failed to decode JSON from file: {path.name}")
            continue

        if not is_valid(json_data):
            print(f"❌ Dataset non valide: {path.name}")
            continue
        print(f"✅ Input dataset valide: {path.name}")

        user_message = HumanMessage(content=f"""
        Given the following JSON, generate a similar JSON file where you paraphrase each question in the content attribute
        (when the role attribute is user) and also paraphrase the value of the response to the question stored in the content attribute
        when the role attribute is assistant.
        The objective is to create synthetic datasets based on existing datasets.
        I do not need to know the code to do this, but I want the resulting JSON file.
        It is important that the term OVHcloud is present as much as possible, especially when the terms AI Endpoints are mentioned
        either in the question or in the response.
        There must always be a question followed by an answer, never two questions or two answers in a row.
        It is IMPERATIVE to keep the language in English.
        The source JSON file:
        {raw_data}
        """)

        chat_response = chat_model.invoke([user_message], response_format=response_format)

        output = chat_response.content

        # Replace unauthorized characters
        output = output.replace("\\t", " ")

        generated_file_name = f"{uuid.uuid4()}_{path.name}"
        with open(f"./generated/synthetic/{generated_file_name}", 'w', encoding='utf-8') as output_file:
            output_file.write(output)

        if not is_valid(json.loads(output)):
            print(f"❌ ERROR: File {generated_file_name} is not valid")
        else:
            print(f"✅ Successfully generated file: {generated_file_name}")</code></pre>



<p><em>ℹ️ Again, you can find all resources to build and run this application in the <a href="https://github.com/ovh/public-cloud-examples/tree/main/ai/llm-fine-tune/dataset/" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">dedicated folder</a> in the GitHub repository.</em></p>



<h3 class="wp-block-heading">Fine-tune the model</h3>



<p>We now have enough training data, let’s fine-tune!</p>



<p><em><em>ℹ️ It’s hard to say exactly how much data is needed to train a model properly. It all depends on the model, the data, the topic, and so on.</em><br><em>The only option is to test and adapt. 🔁</em>.</em></p>



<p>I use <a href="https://jupyter.org/" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">Jupyter notebook</a>, created with <a href="https://www.ovhcloud.com/fr/public-cloud/ai-notebooks/" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">OVHcloud AI Notebooks</a>, to fine-tune my models.</p>



<pre title="Jupyter notebook creation" class="wp-block-code"><code lang="bash" class="language-bash line-numbers">ovhai notebook run conda jupyterlab \
	--name axolto-llm-fine-tune \
	--framework-version 25.3.1-py312-cudadevel128-gpu \
	--flavor l4-1-gpu \
	--gpu 1 \
	--envvar HF_TOKEN=$MY_HF_TOKEN \
	--envvar WANDB_TOKEN=$MY_WANDB_TOKEN \
	--unsecure-http</code></pre>



<p><em><em>ℹ️ For more details on how to create Jupyter notebook with <a href="https://www.ovhcloud.com/fr/public-cloud/ai-notebooks/" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">AI Notebooks</a>, read the <a href="https://help.ovhcloud.com/csm/fr-documentation-public-cloud-ai-and-machine-learning-ai-notebooks?id=kb_browse_cat&amp;kb_id=574a8325551974502d4c6e78b7421938&amp;kb_category=c8441955f49801102d4ca4d466a7fd58&amp;spa=1" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">documentation</a>.</em></em></p>



<p class="has-text-align-left">⚙️ The <strong>HF_TOKEN</strong> environment variable is used to pull and push the trained model to <a href="https://huggingface.co/" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">Hugging Face</a> <br>⚙️ The <strong>WANDB_TOKEN</strong> environment variable helps you track training quality in <a href="https://wandb.ai" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">Weight &amp; Biases</a></p>



<p>Once the notebook is set up, you can start coding the model’s training with Axolotl.</p>



<p>To start, install Axolotl CLI and its dependencies. 🧰</p>



<pre title="Axolot installation" class="wp-block-code"><code lang="bash" class="language-bash"># Axolotl need these dependencies
!pip install torch==2.6.0 torchvision==0.21.0 torchaudio==2.6.0 --index-url https://download.pytorch.org/whl/cu126

# Axolotl CLI installation
!pip install --no-build-isolation axolotl[flash-attn,deepspeed]

# Verify Axolotl version and installation
!axolotl --version</code></pre>



<p></p>



<p>The next step is to configure the Hugging Face CLI. 🤗</p>



<pre title="Hugging Face configurartion" class="wp-block-code"><code lang="bash" class="language-bash line-numbers">!pip install -U "huggingface_hub[cli]"

!huggingface-cli --version</code></pre>



<pre title="Hugging Face hub authentication " class="wp-block-code"><code lang="python" class="language-python line-numbers">import os
from huggingface_hub import login

login(os.getenv("HF_TOKEN"))</code></pre>



<p></p>



<p>Then, configure your Weight &amp; Biases access.</p>



<pre title="Weight &amp; Biases configuration" class="wp-block-code"><code lang="bash" class="language-bash line-numbers">pip install wandb

!wandb login $WANDB_TOKEN</code></pre>



<p></p>



<p>Once all that’s done, it’s time to train the model.</p>



<pre title="Train the model" class="wp-block-code"><code lang="bash" class="language-bash line-numbers">!axolotl train /workspace/instruct-lora-1b-ai-endpoints.yml</code></pre>



<p>You only need to type this one line to train it, how cool is that? 😎</p>



<p><em><em>ℹ️ With one L4 card, 10 epochs, and roughly 2000 questions and answers in the datasets, it ran for about 90 minutes.</em></em></p>



<p>Basically, the command line needs just one parameter: the Axolotl config file. You can find everything you need to set up Axolotl in the <a href="https://docs.axolotl.ai/docs/config-reference.html" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">official documentation</a>.📜<br>Here’s what the model was trained on:</p>



<pre title="Axolotl configuration" class="wp-block-code"><code lang="yaml" class="language-yaml">base_model: meta-llama/Llama-3.2-1B-Instruct
# optionally might have model_type or tokenizer_type
model_type: LlamaForCausalLM
tokenizer_type: AutoTokenizer
# Automatically upload checkpoint and final model to HF
# hub_model_id: username/custom_model_name

load_in_8bit: true
load_in_4bit: false

datasets:
  - path: /workspace/ai-endpoints-doc/
    type: chat_template
      
    field_messages: messages
    message_property_mappings:
      role: role
      content: content
    roles:
      user:
        - user
      assistant:
        - assistant

dataset_prepared_path:
val_set_size: 0.01
output_dir: /workspace/out/llama-3.2-1b-ai-endpoints

sequence_len: 4096
sample_packing: false
pad_to_sequence_len: true

adapter: lora
lora_model_dir:
lora_r: 32
lora_alpha: 16
lora_dropout: 0.05
lora_target_linear: true

wandb_project: ai_endpoints_training
wandb_entity: &lt;user id&gt;
wandb_mode: 
wandb_watch:
wandb_name:
wandb_log_model:

gradient_accumulation_steps: 4
micro_batch_size: 2
num_epochs: 10
optimizer: adamw_bnb_8bit
lr_scheduler: cosine
learning_rate: 0.0002

bf16: auto
tf32: false

gradient_checkpointing: true
resume_from_checkpoint:
logging_steps: 1
flash_attention: true

warmup_steps: 10
evals_per_epoch: 4
saves_per_epoch: 1
weight_decay: 0.0
special_tokens:
   pad_token: &lt;|end_of_text|&gt;
</code></pre>



<p>🔎 Some key points (only the fields modified from the <a href="https://github.com/axolotl-ai-cloud/axolotl/blob/main/examples/llama-3/instruct-lora-8b.yml" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">given templates</a>):<br>&#8211; <strong>base_model: meta-llama/Llama-3.2-1B-Instruct</strong>: before you download the base model from Hugging Face, be sure to accept the licence’s <a>terms of use </a><a href="#_msocom_1">[JD1]</a> <br>&#8211; <strong>path: /workspace/ai-endpoints-doc/</strong>: folder to upload the generated dataset<br>&#8211; <strong>wandb_project: ai_endpoints_training</strong> &amp; <strong>wandb_entity: &lt;user id></strong>: to configure weights and biases<br>&#8211; <strong>num_epochs: 10</strong>: number of epochs for the training<a id="_msocom_1"></a></p>



<p>After the training, you can test the new model 🤖:</p>



<pre title="New model testing" class="wp-block-code"><code lang="bash" class="language-bash line-numbers">!echo "What is OVHcloud AI Endpoints and how to use it?" | axolotl inference /workspace/instruct-lora-1b-ai-endpoints.yml --lora-model-dir="/workspace/out/llama-3.2-1b-ai-endpoints" </code></pre>



<p></p>



<p>When you’re satisfied with the result, merge the weights and upload the new model to Hugging Face:</p>



<pre title="Push the model" class="wp-block-code"><code lang="bash" class="language-bash line-numbers">!axolotl merge-lora /workspace/instruct-lora-1b-ai-endpoints.yml

%cd /workspace/out/llama-3.2-1b-ai-endpoints/merged

!huggingface-cli upload wildagsx/Llama-3.2-1B-Instruct-AI-Endpoints-v0.6 .</code></pre>



<p>ℹ️ <em>You can find all resources to create and run the notebook in the <a href="https://github.com/ovh/public-cloud-examples/tree/main/ai/llm-fine-tune/notebook/" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">dedicated folder</a> in the GitHub repository.</em></p>



<h3 class="wp-block-heading">Test the new model</h3>



<p>Once you have pushed your model in Hugging Face you can, again, deploy it with vLLM and AI Deploy to test it ⚡️.</p>



<figure class="wp-block-image aligncenter size-large"><img loading="lazy" decoding="async" width="1024" height="474" src="https://blog.ovhcloud.com/wp-content/uploads/2025/07/Screenshot-2025-07-23-at-14.58.02-1024x474.png" alt="" class="wp-image-29459" srcset="https://blog.ovhcloud.com/wp-content/uploads/2025/07/Screenshot-2025-07-23-at-14.58.02-1024x474.png 1024w, https://blog.ovhcloud.com/wp-content/uploads/2025/07/Screenshot-2025-07-23-at-14.58.02-300x139.png 300w, https://blog.ovhcloud.com/wp-content/uploads/2025/07/Screenshot-2025-07-23-at-14.58.02-768x356.png 768w, https://blog.ovhcloud.com/wp-content/uploads/2025/07/Screenshot-2025-07-23-at-14.58.02-1536x712.png 1536w, https://blog.ovhcloud.com/wp-content/uploads/2025/07/Screenshot-2025-07-23-at-14.58.02-2048x949.png 2048w" sizes="auto, (max-width: 1024px) 100vw, 1024px" /></figure>



<p>Ta-da! 🥳 Our little Llama model is now an OVHcloud AI Endpoints pro!</p>



<p></p>



<p>Feel free to try out OVHcloud Machine Learning products, and share your thoughts on our Discord server (<em><a href="https://discord.gg/ovhcloud" data-wpel-link="external" target="_blank" rel="nofollow external noopener noreferrer">https://discord.gg/ovhcloud</a></em>), see you soon! 👋</p>
<img loading="lazy" decoding="async" src="//blog.ovhcloud.com/wp-content/plugins/matomo/app/matomo.php?idsite=1&amp;rec=1&amp;url=https%3A%2F%2Fblog.ovhcloud.com%2Ffine-tune-an-llm-with-axolotl-and-ovhcloud-machine-learning-services%2F&amp;action_name=Fine%20tune%20an%20LLM%20with%20Axolotl%20and%20OVHcloud%20Machine%20Learning%20Services&amp;urlref=https%3A%2F%2Fblog.ovhcloud.com%2Ffeed%2F" style="border:0;width:0;height:0" width="0" height="0" alt="" />]]></content:encoded>
					
		
		
			</item>
	</channel>
</rss>
