Mistral AI, a French artificial intelligence company, announced a new 10 MW inference data center called Les Ulis facility located in the Essonne region, south of Paris. The company has expanded its industrial AI strategy into aerospace, automotive, and semiconductor engineering through a combined large language model and physics simulation stack, as part of a broader push to control compute infrastructure and serve enterprise deployments that require on-premises or sovereign data handling.
The facility will support inference workloads and is scheduled for full operation in Q3 2026. It follows an existing 40 MW training site at Bruyères-le-Châtel and is part of a wider €4 billion (US$ 4.6 billion) compute infrastructure program spanning France and Sweden, including planned capacity expansion toward 200 MW by 2027 and 1 GW by 2030. Mistral also referenced additional infrastructure development in Sweden aimed at hosting next-generation GPU systems for model training and deployment as reported by Venture Beat.
During the AI NOW Summit, Arthur Mensch, co-founder and CEO, Minstrel AI, told the audience. “We have two convictions at Mistral, the first is that in order to deploy AI in the enterprise, you actually need, as an AI provider, to own the full stack.”
Mensch also described Mistral’s business as fundamentally about “transforming electrons into tokens and intelligence,” arguing that physical infrastructure control matters as much as model quality.
The infrastructure buildout is tied to a financing structure that includes an US$ 830 million debt round raised in March 2026 to support data center construction and GPU procurement. The company framed the expansion as necessary to secure compute capacity amid supply constraints and to support enterprise customers that require dedicated or private infrastructure for AI inference and training workloads.
In parallel, Mistral expanded into industrial AI via “Mistral for Industrial Engineering,” a platform that integrates its models with physics simulation systems acquired through its purchase of Emmi AI. The system targets use cases in aircraft, automotive, and semiconductor design by replacing parts of traditional simulation pipelines with faster surrogate models trained on solver outputs.
The industrial push includes partnerships with Airbus across aircraft and defence programs, BMW Group on multimodal simulation systems for vehicle design, and ASML on engineering workflows and service diagnostics. In one example cited by ASML, the company said AI-assisted tooling delivered “120 times faster” performance at similar accuracy levels compared with existing approaches, alongside use of agents for software review and defect detection.