Arthur Mensch, CEO of Mistral AI, spoke publicly for the first time on Thursday about the company’s plans to develop its own AI chips, stating that Mistral is exploring custom chip designs and ‘has not ruled out this possibility.’ He noted that tailored chips could ‘significantly reduce the cost of token deployment,’ adding that ‘having proprietary chips will eventually become a trend,’ though currently Mistral still relies on NVIDIA. On the same day, Mensch announced the construction of a new data center in France dedicated to AI inference; to date, Mistral has invested a total of €4 billion in data center projects in France and Sweden. The additional computing power will serve both Mistral’s corporate clients and other AI labs; Mensch revealed that ‘multiple labs have already expressed substantial demand for such resources.’ Mistral primarily targets enterprise customers, including major European firms like ASML, Airbus, and BMW. Currently valued at nearly €12 billion, the company aims to achieve €1 billion in revenue by 2026 (up from roughly €200 million in 2025).
Also on Thursday, Mistral launched Vibe, an enterprise agent platform capable of drafting documents, writing code, reasoning autonomously, and delivering results. Its subsidiary tool Vibe Code supports coding, testing, and deployment across multiple code repositories, positioning Mistral as a direct competitor to similar offerings from OpenAI and Anthropic. Describing current AI-related challenges in Europe as ‘macroeconomic issues rather than mere technical problems,’ Mensch emphasized that Europe now views AI as a strategic resource comparable to natural gas. Should Mistral’s chip development efforts succeed, they would mirror Amazon and Google’s strategies of designing cloud-specific AI chips, marking the first instance of a European AI firm adopting an infrastructure strategy akin to that of top U.S. cloud providers.