Mistral AI expanded into cloud infrastructure in mid-2025, teaming with Nvidia to build a “sovereign” European AI supercomputer. At the VivaTech conference in Paris, Nvidia CEO Jensen Huang announced a partnership with Mistral to create a dedicated AI computing platform in Europe running on 18,000 of Nvidia’s latest chips.
Dubbed “Mistral Compute,” the new platform offers European enterprises and governments a home-grown alternative to relying on US cloud providers like AWS, Azure, or Google Cloud.
“Sovereign AI is an imperative – no company, industry or nation can outsource its intelligence,” Huang declared, underscoring the project’s significance for European tech autonomy.
The first phase will be housed in a data center in France’s Essonne region, with plans to scale across multiple sites by 2026. This marks a strategic shift for Mistral from purely developing AI models to building full-stack AI infrastructure, aligning with Europe’s push to boost its computing capacity tenfold in two years.
Alongside the cloud platform, Mistral simultaneously unveiled new “Magistral” AI models that demonstrate its technical prowess.
In an exclusive briefing, chief scientist Guillaume Lample said these reasoning models can perform step-by-step logical thinking and even handle images, while displaying their chain-of-thought to users in multiple languages.
The company released Magistral Small (open-source) and Magistral Medium (enterprise API), which showed surprising emergent abilities during training – such as conducting web searches or coding tasks autonomously.
The Nvidia collaboration provides Mistral access to cutting-edge hardware (including Grace Hopper “superchips”) and expertise, enabling faster reinforcement-learning training methods that Lample described as a breakthrough in efficiency.
Collectively, these moves form a bold market strategy for Mistral: by controlling its own cloud infrastructure and tailoring models to European needs (including data sovereignty and sustainability concerns), the startup is positioning itself as the backbone of Europe’s AI ecosystem.
The launch of Mistral Compute not only challenges American cloud giants on their home turf, but also reinforces Europe’s capacity to develop and deploy AI on its own terms – a milestone hailed by officials and industry leaders alike.
