<div id="readability-page-1"><span></span><div><p>French startup<strong> </strong>Mistral AI is exploring designing its own chips and may eventually develop them, CEO Arthur Mensch told CNBC.</p><p>It is the first comment made by Mensch about Mistral's semiconductor ambitions, underscoring how the company is looking to control more of its infrastructure as it competes with U.S. heavyweights OpenAI and Anthropic.</p><p>&#34;Of course, it is interesting,&#34; Mensch said about the prospect of Mistral developing its own chips, adding that the company is not ruling it out<strong>.</strong></p><p>Mensch said that custom chips allow a company to &#34;lower the cost of deploying tokens to meaningful extents.&#34; Tokens are units of data processed by AI models. </p><p>&#34;Owning the chips may come, I think it should come at some point, but for now we are relying on <span id="RegularArticle-QuoteInBody-2">Nvidia<span><span id="-WatchlistDropdown"></span></span></span>, which is a great partner to us, and we're testing a few things here and there,&#34; Mensch told CNBC. </p><p>Mistral, which is valued at nearly 12 billion euros, develops AI models but is also investing in building data centers with Nvidia chips. The Paris-headquartered firm is often seen as Europe's answer to OpenAI and Anthropic. </p><p>The company is focused on enterprise and counts companies such as chip equipment giant <span id="RegularArticle-QuoteInBody-4">ASML<span><span id="-WatchlistDropdown"></span></span></span> among its top customers. </p><p>If Mistral were to develop its own chips, it would follow in the footsteps of the big American hyperscalers such as <span id="RegularArticle-QuoteInBody-5">Amazon<span><span id="-WatchlistDropdown"></span></span></span> and <span id="RegularArticle-QuoteInBody-6">Google<span><span id="-WatchlistDropdown"></span></span></span>, which have designed and deployed their own semiconductors in their data centers. </p><p>Custom chips, also known as an application-specific integrated circuit, are seen as a way for hyperscalers to have more control over their hardware and software integrations, potentially offering a differentiated product from competitors.</p></div><h2>Data center expansion </h2><div><p>Mistral announced on Thursday a new data center in France designed specifically for inferencing, which is the actual process of running AI models. </p><p>Mistral has invested 4 billion euros in data centres in France and Sweden as it looks to ramp up its compute capacity. </p><p>&#34;Europe is lagging behind when it comes to [the] buildout of infrastructure, and so we are investing to close that gap,&#34; Mensch told CNBC.</p><p>The bloc is facing not just a technological problem, but a macroeconomic problem, he said, adding that Europe is now starting to view AI as a strategic asset, the same way it has looked at gas.</p></div><div><p>&#34;You can't afford to have a commercial deficit of a trillion if you actually want to stay competitive in the race, and so that's something I think that people are realizing that we're talking about something that should be concerning for any one of us.&#34;</p><p>The additional compute capacity in France will be used to service Mistral's customers as well as other AI labs, Mensch said, without mentioning specific companies.</p><p>&#34;AI labs are in sore need of compute, and we have some of it, and some of them are actually asking us for a lot of compute today,&#34; Mensch said.</p><p>The Mistral CEO said the company needs to &#34;prioritize the access&#34; for its compute, with some going to AI labs and &#34;more importantly&#34; to its customers.</p></div><h2>Agentic focus</h2><div><p>Mistral on Thursday also unveiled a new agentic platform for enterprises, as it looks to compete with U.S. rivals including Anthropic and OpenAI, which have stepped up their own offerings in the space in recent times. </p><p>The new enterprise agent platform, which Mistral is calling &#34;Vibe,&#34; will be able to carry out tasks like drafting work and coding, the company said. </p><p>AI companies are increasingly focused on agentic AI, systems that can autonomously carry out tasks on behalf of users. </p></div><div id="RegularArticle-RelatedLinks-1"><p></p><h2>Read more</h2><p></p><div><div><div><p>Mistral secures $830 million in debt financing to fund AI data center</p></div></div><div><div><p>Amazon custom chips get a boost from Meta, giving the cloud giant another path to win in AI</p></div></div><div><div><p>Cheap AI could derail OpenAI and Anthropic’s IPOs</p></div></div></div></div><div><p>&#34;Vibe is the agent platform for the tasks at hand, putting frontier AI to work,&#34; Timothée Lacroix, Mistral's chief technology officer, said in a statement. </p><p>&#34;Users can set the brief and move on, as Vibe thinks, drafts, and delivers finished work from a single conversation. Vibe Code writes, tests, and deploys code across codebases.&#34;</p><p>The latest releases from Mistral underscore its efforts to accelerate revenue growth. The company is targeting 1 billion euros in revenue in 2026. While that figure is a step up from the 200 million euros it made the previous year, it's dwarfed by the balance sheets at OpenAI and Anthropic. </p><p>OpenAI's annualized recurring revenue stood at $20 billion in 2025, with Anthropic set to hit $10.9 billion in revenue in the second quarter of 2026.</p></div><div><p>Choose CNBC as your preferred source on Google and never miss a moment from the most trusted name in business news.</p></div></div>

NVDA

<p>Mistral AI, a French startup, is exploring the design of its own chips to enhance its infrastructure and compete with OpenAI and Anthropic. CEO Arthur Mensch highlighted the potential cost benefits of custom chips while currently relying on Nvidia. Mistral is also expanding its data centers in France and Sweden, investing 4 billion euros to increase compute capacity. Additionally, the company launched an enterprise agent platform called 'Vibe' aimed at automating tasks, targeting 1 billion euros in revenue by 2026.</p>

- French startup Mistral AI is considering developing its own chips to enhance infrastructure and compete with U.S. companies like OpenAI and Anthropic.  
- CEO Arthur Mensch highlighted the potential for custom chips to lower costs for deploying AI tokens while acknowledging reliance on Nvidia for now.  
- Mistral is also investing 4 billion euros in new data centers in France and Sweden to address compute capacity needs, as it aims for 1 billion euros in revenue by 2026.

CNBC

NVIDIA

OpenAI

Mistral to explore designing own chips, CEO says, as it ramps up infrastructure build