Europe’s AI goals focus on two main areas: creating advanced models in labs like Mistral and building the infrastructure to run them at scale without relying on Big Tech. Buyers struggle with vendor lock-in, expensive GPUs, and compliance issues since AWS, Azure, and Google Cloud hold over 70% of the market.
A few European challengers, like OVHcloud, Scaleway, and other smaller providers, exist, but they are much smaller and less specialised in AI than US giants. This leaves room for companies like Mistral to try a fresh approach.
Earlier this week, Paris-based Mistral AI announced its first acquisition: French serverless cloud startup Koyeb. While the terms were not shared, the deal brings Koyeb’s team and technology into Mistral Compute, helping Mistral gain full control over AI deployment.
The acquisition amount wasn’t disclosed.
Behind Koyeb: A “deploy anywhere” serverless platform
Koyeb was founded in 2020 by former Scaleway engineers Yann Léger, Édouard Bonlieu, and Bastien Chatelard. They noticed developers having a hard time dealing with multi-cloud complexity, juggling different providers for apps, data, and AI without a unified way to scale.
Their mission was to create a “deploy anywhere” serverless platform on bare metal, allowing code to run worldwide without servers or ops teams. They aimed to give developers freedom during a fragmented cloud era, supporting AI inference on CPUs and GPUs with autoscaling.
Now, their 13-16 person team, with strong skills in cloud infrastructure, developer tools, distributed systems, and AI operations, joins Mistral to build this into a complete AI stack, driven by frustration with hyperscaler silos.
Koyeb’s serverless features, such as global edge deployments, GPU-optimised AI sandboxes, and zero-configuration autoscaling, give Mistral a platform to offer on-demand inference without requiring DevOps teams. Its main benefits are multi-cloud neutrality, bare-metal efficiency for better cost and speed, and AI-native agent isolation.
What’s next?
Koyeb’s technology improves Mistral Compute, supporting everything from training to inference on any hardware. Backed by €1.2 billion in Swedish data centres, it helps ensure independence from hyperscalers.
The platform will stay live for users for now while shifting focus to enterprise AI.
