Mistral AI makes its first acquisition to build an AI cloud

Mistral AI just made its first acquisition, and it signals a major shift in how the European heavyweight plans to make money.

The French AI unicorn, last valued at $13.8 billion, bought Paris-based serverless startup Koyeb to bolster its cloud infrastructure. Financial terms were not disclosed, but the deal effectively folds Koyeb's 13-person engineering team and three co-founders into Mistral under CTO Timothée Lacroix.

From models to the full stack

Koyeb was founded in 2020 by three alumni of French cloud provider Scaleway to eliminate the headache of managing server infrastructure. As the AI boom hit, that serverless angle became critical for inference, and Koyeb even recently launched Sandboxes for isolated AI agent deployment.

Mistral will now use Koyeb's tech to accelerate Mistral Compute, the proprietary cloud offering it rolled out in June 2025. Specifically, Koyeb's software will help Mistral deploy models directly on enterprise clients' own hardware (on-premises), optimize expensive GPU usage, and scale inference workloads.

Koyeb's standalone platform will stay online for now, but its priorities have instantly flipped. New signups for Koyeb's Starter tier have been cut off as the team redirects entirely toward enterprise workloads.

The European cloud alternative

The move comes amid an aggressive infrastructure push from Mistral. Just days ago, the company announced a $1.4 billion investment in Swedish data centers to capitalize on rising demand for European alternatives to U.S. cloud giants. Last week at the Techarena conference in Stockholm, CEO Arthur Mensch was actively pitching infrastructure engineers on Mistral's identity as a homegrown European frontier research lab.

That enterprise focus and geopolitical positioning seem to be paying off: Mistral recently crossed the milestone of $400 million in annual recurring revenue (ARR).

Why it matters

We think this acquisition proves that selling frontier API access alone is a brutal, low-margin trap. The real money in enterprise AI lives in deployment, governance, and private infrastructure.

By swallowing Koyeb, Mistral turns itself into an end-to-end platform capable of running models inside a client's own data center while squeezing maximum efficiency out of scarce GPUs. That makes it an attractive, sovereign alternative to the OpenAI-Microsoft bundle for privacy-minded European enterprises. The downside is that hobbyists get left behind, as Koyeb's closed Starter tier proves Mistral's ambitions are now strictly corporate. For competitors, the bar just got raised: shipping great weights is no longer enough if you can't run the pipes underneath them.