Mistral AI explores custom chip designs, announces new data center in France

Mistral AI, the French generative AI company, has secured $830 million in debt financing to build its first dedicated data center near Paris, announced plans to explore custom chip designs, and unveiled an enterprise AI platform called Vibe.

The data center, located in Bruyères-le-Châtel, will house 13,800 Nvidia GB300 GPUs and deliver 44 megawatts of compute capacity. Operations are expected to begin in the second quarter of 2026.

The infrastructure play

Mistral’s $830 million debt raise, funded by a consortium of banks, is aimed at constructing its own facility, stocked with Nvidia’s latest GB300 chips, to serve enterprise clients directly. The company has set a target of 200 megawatts of total AI compute capacity spread across European sites by the end of 2027.

This isn’t Mistral’s first infrastructure bet on the continent. The company has previously invested in a $1.4 billion-equivalent venture in Sweden.

The custom chip exploration adds another layer. Designing proprietary silicon is the kind of move that separates companies content to rent compute from those trying to own the full stack. It’s the same playbook Google followed with its TPUs and Amazon pursued with Graviton and Trainium.

Vibe and the enterprise push

On the software side, Mistral announced Vibe, an enterprise coding agent platform designed to slot into existing enterprise workflows. Vibe integrates with Mistral Medium 3.5, a 128-billion-parameter model that the company unveiled in late April 2026.

European enterprises, particularly those in regulated industries like finance, healthcare, and defense, increasingly want AI tools that don’t route their proprietary code through American cloud infrastructure. Mistral is positioning Vibe to be the answer to that requirement.

Europe’s sovereign AI ambitions

France has been vocal about building an independent AI ecosystem, and Mistral has become the flagship company for that vision. The involvement of Bpifrance, France’s public investment bank, and strategic partnerships with Nvidia underscore the degree to which this is a public-private effort.

The $830 million debt financing is notable for its structure. Debt, not equity, means Mistral’s existing shareholders aren’t getting diluted and signals that the company is confident enough in its revenue trajectory to take on debt obligations.

The 200-megawatt target for end of 2027 will be the number to watch. If Mistral hits it, the company will have built one of Europe’s largest dedicated AI compute networks in under two years.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.