Mistral

Mistral Buys Koyeb: Serverless Joins the Sovereign AI Cloud

On February 17, 2026, Mistral AI agreed to buy Paris serverless startup Koyeb — its first acquisition ever. The team folds into Mistral Compute, alongside 40 MW of capacity and 18,000 Blackwell GPUs.

Mistral Buys Koyeb: Serverless Joins the Sovereign AI Cloud — article cover

On February 17, 2026, Mistral AI announced a definitive agreement to acquire Koyeb, a Paris-based startup. It is the French AI company’s first acquisition ever; financial terms were not disclosed, and completion remains subject to closing conditions. TechCrunch and Reuters reported the story the same day, noting Mistral’s last known valuation of roughly $13.8 billion and framing the deal as backing for its homegrown AI cloud, Mistral Compute.

The announcement is simple; the positioning is what matters. While US labs were busy announcing giant rounds — Anthropic closed a $30 billion Series G at a $380 billion valuation the very same week — Mistral spent its money on a 16-person infrastructure team. Those are two different expansion logics: one buys compute, the other buys the people who can turn compute into a product.

The Deal: Mistral’s First Acquisition

Koyeb’s platform keeps running without disruption. Existing customers see no plan changes, billing stays put, and user data will not move to Mistral “immediately.” The change hits new signups: the free Starter tier is closed, new accounts start at the Pro plan and above, and the focus shifts squarely to enterprise customers.

Koyeb was founded by three former Scaleway engineers — Yann Léger, Edouard Bonlieu, and Bastien Chatelard — and had raised a modest $8.6 million in total: a $1.6 million pre-seed in 2020 and a $7 million seed led by Paris VC Serena in 2023. Serena principal Floriane de Maupeou’s verdict on the exit: the deal helps build “the foundations of sovereign AI infrastructure in Europe.” Mistral did not say whether more acquisitions are coming; TechCrunch could learn neither the price nor the structure, only that closing conditions remain.

For a company last valued at $13.8 billion, buying a seed-stage infrastructure team instead of renting another data center is a deliberate statement about where the scarcity lies.

What Koyeb Built

Koyeb operated a serverless compute platform: serverless GPUs, specialized accelerators such as Tenstorrent parts, and CPU workloads, with sub-second scale-to-zero, autoscaling, and isolated sandboxes purpose-built for AI agents. The platform hosted tens of thousands of applications on bare-metal servers across ten global locations.

For Mistral, the valuable asset is the operational layer — scheduling, isolation, on-demand scaling — which is precisely the hardest software to write when you want to package a GPU cluster as a cloud service.

Fitting Into Mistral Compute

Mistral Compute launched in June 2025, positioned as infrastructure similar to what Mistral runs internally. Koyeb’s technology is set to become a core component of it, with concrete uses: deploying models onto clients’ on-premises hardware, optimizing GPU utilization, and scaling AI inference. Koyeb’s blog also pledges to double down on inference, sandboxes, and serverless capabilities for MCP servers.

The hardware groundwork was already laid. Koyeb’s announcement cites Mistral’s deployment of 40 MW of data center capacity and 18,000 NVIDIA Blackwell GPUs as a starting point, and days earlier Mistral announced a $1.4 billion investment in Swedish data centers, citing customer demand for alternatives to US infrastructure. CTO Timothée Lacroix’s statement: “Koyeb’s product and expertise will accelerate our development on the Compute front.” The engineering fit is concrete: Mistral has raw capacity and needs the software that turns capacity into a billable service with isolation guarantees.

The Sovereignty Story Meets the P&L

The geopolitical backdrop needs no embellishment. CEO Arthur Mensch’s pitch at Stockholm’s Techarena was blunt: a company “headquartered in Europe, that is doing frontier research in Europe.” For European enterprises and governments unwilling to park data and workloads on US clouds, Mistral is assembling a full vertical stack from models to machines.

The commercial numbers hold the story up too: TechCrunch cites annual recurring revenue past $400 million, and the company keeps hiring for infrastructure roles. Two takeaways for developers. First, European sovereign AI is no longer slide-ware — it has data centers, a scheduling layer, and paying enterprise customers. Second, Koyeb’s absorption means one fewer independent serverless option; teams running agents in Koyeb sandboxes should start pricing out migration and supplier-concentration risk.

Sources

AI-assisted summary compiled from the sources above, reviewed by a human before publishing.

FOUND_THIS_USEFUL?

Support more practical AI articles, tutorials, and build notes.

BUY_ME_A_COFFEE
SHAREXEMAIL