Key Points To compete with U.S. firms like OpenAI and Anthropic, Mistral AI plans to enhance its proprietary infrastructure, with in-house chip development being a crucial component. The French startup announced the construction of a new dedicated inference data center in France. CEO Arthur Mensch, in an interview with CNBC, stated that Europe lags in computational infrastructure development, and the company is increasing investments to bridge this gap.
Arthur Mensch, CEO of French AI startup Mistral AI, confirmed to CNBC that the company is exploring the development of its own chips and may officially launch related products in the future. This marks Mensch's first public comments on the chip development plan, reflecting the company's desire to gain more control over underlying infrastructure as it competes with U.S. industry giants like OpenAI and Anthropic. Discussing the concept of in-house chips, Mensch said, "This direction holds value, and the company does not rule out this possibility." He explained that custom chips could significantly reduce the deployment costs associated with AI token calls, with tokens being the fundamental units of data processed by AI models. "Developing our own chips is an inevitable future choice; it's just a matter of timing. Currently, we still rely on collaboration with NVIDIA, a very reliable partner, while we are also conducting several related tests." Mistral is currently valued at nearly 12 billion euros. It primarily focuses on AI model development and is also building data centers based on NVIDIA chips. Headquartered in Paris, the company is viewed as Europe's local AI contender against U.S. players like OpenAI and Anthropic. The company's business is centered on the enterprise market, with chip equipment giant ASML Holding NV being one of its key clients. If Mistral officially launches its own chips, it would follow the path of large U.S. tech companies like Amazon.com and Alphabet, which have long developed their own chips for use in their data centers. Custom chips, or application-specific integrated circuits, enable companies to deeply control the integration of software and hardware, creating differentiated products and competitive advantages.
Data Center Expansion Mistral announced on Thursday that it will build a new dedicated AI inference data center in France. Inference is the core process where AI models run and deliver services. To date, the company has invested a total of 4 billion euros in data centers in France and Sweden, aggressively expanding its computational capacity. Mensch stated, "Europe's progress in computational infrastructure construction is lagging, and our continued investment aims to narrow this gap." He believes that Europe currently faces not only technological shortcomings but also macroeconomic challenges. Europe now views artificial intelligence as a strategic resource as crucial as energy. "If Europe wants to remain competitive in the global AI race, it cannot afford a trade deficit on the scale of trillions. This is a consensus gradually being reached across various sectors and warrants everyone's attention." The new computational resources in France will serve both Mistral's own clients and other AI research institutions. Mensch did not disclose specific partner companies. "Many AI labs currently face significant computational shortages, and numerous institutions have approached us with requests for computational resources," he said, adding that the company's computational resources would be prioritized for partner labs and, more importantly, for enterprise clients.
Launch of Agent Platform To further compete with U.S. rivals, Mistral also launched a new enterprise agent platform on Thursday. Recently, both OpenAI and Anthropic have been intensifying their efforts in this area. Named "Vibe," this enterprise agent platform can autonomously handle tasks such as document writing and code generation. The global AI industry is currently focused on developing agent technology, systems that can autonomously perform various tasks for users. Mistral's Chief Technology Officer, Timothée Lacroix, said, "Vibe is an intelligent platform designed for practical work scenarios, applying cutting-edge AI technology. Users simply state their requirements, and the platform can independently conceive, write, and deliver complete results. Its code module can also handle code writing, testing, and full project deployment." These series of new initiatives reflect Mistral's goal of accelerating revenue growth. The company plans to achieve 1 billion euros in revenue by 2026. While this figure represents significant growth from the previous year's 200 million euros, it still lags far behind U.S. counterparts: OpenAI's annualized recurring revenue for 2025 is projected to reach 20 billion U.S. dollars, and Anthropic expects revenue to hit 10.9 billion U.S. dollars in the second quarter of 2026.
Comments