Mistral AI

Date: Share: Mistral AI is negotiating a 3 billion euro fundraising round at a valuation of 20 billion euros, according to Bloomberg. A doubling in less than a year. Discussions have just begun. Mistral AI, the French laboratory specializing in artificial intelligence, is reportedly in talks with several investors to complete a new 3 billion euro funding round, Bloomberg reports. 7 billion euros. The discussions are only at a preliminary stage, and conditions could still change, according to sources close to the matter who requested anonymity. This new milestone comes less than a year after the record fundraising in September, which propelled Mistral AI to the rank of decacorn, companies valued at over 10 billion euros.

7 billion euro funding round had already shattered all French Tech records. If negotiations succeed, Arthur Mensch's startup would go even further, with one of the most massive fundraising rounds ever carried out in Europe for an AI company. AUKUS: Australia chooses Leidos Gibbs & Cox to design its nuclear submarine dry dock Mistral AI currently counts among its shareholders the Dutch group ASML, a world leader in semiconductor equipment, which holds 11% of the capital [1]. The French public bank Bpifrance is also among the initial investors, alongside several American venture capital funds such as Lightspeed Venture Partners and General Catalyst.

Founded in 2023 by former researchers from Google DeepMind and Meta Platforms, the company immediately set itself a goal: to establish itself as the European champion against American and Chinese artificial intelligence giants. The founders (Arthur Mensch, Guillaume Lample, and Timothée Lacroix) built Mistral AI on the bet of the open-weight model, an approach that makes model weights accessible while allowing the company to commercialize optimized and hosted versions. This strategy quickly proved attractive: by January 2026, annualized recurring revenue (ARR) reportedly exceeded 400 million dollars, compared to approximately 16 million at the end of 2024 [3].

The stated goal is to cross the billion mark by the end of 2026. Mistral AI has multiplied operations to strengthen its position. The company first acquired Koyeb, a serverless cloud provider, then Emmi AI, specializing in AI models applied to physical simulation and industrial engineering. These two acquisitions allow it to expand its offering beyond just large language models (LLMs). 2 billion euro investment to build data centers in Sweden [5]. In parallel, it raised 830 million dollars in debt to acquire 13,800 Nvidia chips, intended to equip a new data center near Paris.

This is the first time Mistral AI has raised debt, a signal that investors interpret as a mark of confidence in the company's financial solidity. Source: Capital, Maddyness During its "AI Now Summit" event held on May 28, Mistral AI unveiled its ambitions regarding computing power. The company aims for a capacity of 200 megawatts by 2027, then one gigawatt by 2030. These figures place the company in the same category as major American labs, which deploy colossal infrastructures to train and serve their generative AI models. On this occasion, Mistral AI also announced a five-year partnership with Airbus, as well as a strategic alliance with BMW.

These agreements allow it to position itself as an AI provider for European industry, a segment that the company considers strategic to ensure its medium-term profitability. Mistral AI now describes itself as a "full-stack AI company," a phrase that marks its evolution since its beginnings. The company no longer just designs language models: it also builds the infrastructure to host, train, and deploy them at scale. This rise in power as an infrastructure player places it in direct competition with giants like OpenAI, Google DeepMind, or Anthropic, all of whom have their own data centers and computing chains.