Koyeb has agreed to join Mistral AI, aiming to accelerate the buildout of next-generation AI infrastructure. The acquisition, which is subject to customary closing conditions, will bring Koyeb’s platform, technology, and team into Mistral AI’s expanding Compute offering.
The combination is designed to strengthen Mistral Compute, a new infrastructure initiative intended to provide enterprises and AI teams with the same high-performance systems Mistral uses to build, run, and scale frontier AI models and applications.
Founded in 2021, Koyeb set out to build a new generation of cloud infrastructure, centered on a seamless serverless experience backed by high-performance hardware, deployable globally in seconds. The company’s founders, who previously spent a decade building and operating large-scale cloud platforms, saw traditional CPU-based infrastructure becoming commoditized while advanced, scalable platforms remained complex and costly.
The rapid rise of generative AI and agent-based workloads reshaped cloud requirements. AI applications increasingly demand serverless GPUs, specialized accelerators, and flexible CPU workloads that can efficiently scale from zero to high throughput in sub-seconds. They also require sandboxed environments capable of supporting autonomous agents and complex inference tasks.
Koyeb responded by building a purpose-built serverless platform designed specifically for AI applications. Today, the company runs tens of thousands of applications across bare metal servers in 10 locations worldwide, enabling teams to deploy APIs, agents, and high-performance GPU inference workloads without managing servers or infrastructure operations. By staying close to hardware design and optimization, Koyeb has emphasized high performance with sustainable cost structures, a critical factor for AI teams navigating growing compute demands.
According to Koyeb, conversations with Mistral AI leadership revealed a shared vision around building world-class AI infrastructure in Europe and beyond. Mistral AI has been rapidly scaling both its AI models and its enterprise footprint, and is now extending that momentum into infrastructure.
Mistral AI is deploying 40 megawatts of data center capacity and 18,000 Nvidia Blackwell GPUs as an initial foundation. The company is also expanding with a $1.4 billion data center investment in Sweden, signaling a major push to strengthen European AI infrastructure capacity.
Koyeb’s team will join Mistral AI full-time, working on Mistral Compute as a core initiative. The goal is to create state-of-the-art infrastructure accessible to AI developers and enterprises worldwide.
Koyeb confirmed that its platform will continue operating without disruption. Over the coming months, it is expected to transition into a core component of Mistral Compute, with increased focus on inference, sandbox environments, and serverless capabilities tailored for AI workloads.
For existing customers, there are no immediate changes. Current organizations, including those on Starter plans, will continue operating under the same terms. However, new users will need to subscribe to paid tiers — Pro, Scale, or Enterprise — as the company shifts its focus toward production-grade infrastructure and away from maintaining a free tier, which it described as operationally intensive.
Billing, support channels, and account access remain unchanged. Customers do not need a Mistral AI account to use Koyeb, and no data will be transferred immediately as part of the acquisition. If future changes occur, the company said it will communicate clearly in advance.
The acquisition positions Mistral AI not only as a frontier model developer but also as a vertically integrated AI infrastructure provider. By combining large-scale GPU capacity with Koyeb’s serverless platform, Mistral aims to offer enterprises a full-stack AI environment, from model development to deployment and scaling.
