Mistral expands into cloud infrastructure
French AI startup Mistral has made its first acquisition, expanding into AI-focused cloud infrastructure to build an end-to-end deployment platform. The move signals an intent to compete with hyperscalers for AI workloads. The company's models continue to show strong performance, with its Magistral models benchmarking at over 70% on AIME2024 mathematical reasoning tasks.
- The company acquired by Mistral AI is Koyeb, a Paris-based serverless platform founded by former employees of the cloud provider Scaleway. The deal brings Koyeb's 13-person engineering team in-house to accelerate the development of "Mistral Compute," the company's AI cloud offering. - Mistral AI was founded in April 2023 by Arthur Mensch, formerly of Google's DeepMind, and Guillaume Lample and Timothée Lacroix, both from Meta. Lample was a creator of Meta's LLaMA language model. - The company's valuation reached €11.7 billion ($13.8 billion) following a €1.7 billion Series C funding round led by semiconductor equipment manufacturer ASML in September 2025. - In February 2024, Microsoft announced a multi-year partnership that included a $16 million investment in Mistral and made Mistral's premium models, including Mistral Large, available on the Azure AI platform. - The acquisition is part of a broader strategy to build a sovereign European AI infrastructure, which includes a reported $1.4 billion investment in new data centers in Sweden. - Beyond its Magistral series, the company has released newer models like Codestral 25.01, a 22B parameter model for coding that achieves an 86.6% score on the HumanEval benchmark. - In December 2025, the company launched its "Mistral 3" generation of models, featuring small, efficient "Ministral" models and Mistral Large 3, a 675-billion parameter Mixture-of-Experts (MoE) model released under the Apache 2.0 license.
