Mistral shifts to full-stack AI: on-prem models, data centers, and enterprise partnerships

At the AI Now Summit in Paris, Mistral AI laid out its transformation from a pure model provider to a full-stack AI company. CEO Arthur Mensch and co-founders emphasized that Mistral now owns compute infrastructure—starting with a 40MW data center in Paris, with additional facilities planned in Sweden. They launched Vibe for Work, a product similar to Claude for Work, and doubled down on open, bespoke models that enterprises can run on-premises. The message was clear: for European companies in regulated industries, sovereignty and control matter more than raw AGI capabilities.

Mistral also showcased its strategy of specialized small models outperforming general-purpose ones in speed and efficiency. Document AI powers OCR for the EU Patent Office, Voxtral handles multilingual voice for Amazon Alexa+ in Europe, and Robostral drives industrial robotics with ASML. BNP Paribas runs Mistral models on-prem for KYC in Belgium, keeping sensitive data in-house. A standout talk revealed that researchers finetuned Mistral’s Codestral to read ancient papyrus documents, unlocking 180,000 texts that would have taken 2,000 years to process manually. Mistral is positioning itself as the European AI partner delivering immediate business value, not just future promises.