Mistral AI buys Koyeb to power EU AI sovereignty

Mistral AI, Europe's largest AI company, announced its first company acquisition. The French rival of OpenAI bought Koyeb, a serverless computing platform that enables developers to run and scale AI applications without infrastructure. 

Koyeb, also based in Paris, enables developers to deploy applications globally without managing underlying servers. This technology is critical for modern AI workloads, which often require 'bursty' computational power—scaling up instantly when a user asks a question and scaling down to zero when idle to save costs. 

Founded in 2023 by former Google DeepMind engineers, Mistral AI has quickly grown into Europe’s main AI company. In 2025, it launched Le Chat, its chatbot rivaling ChatGPT, Google’s Gemini, and Anthropic’s Claude.

Why did Mistral AI acquire Koyeb?

The Koyeb acquisition serves a specific technical purpose. It will accelerate the development of Mistral Compute. This is a European-hosted AI infrastructure platform launched by the AI firm last year, designed to provide organizations with private, integrated, and sovereign AI computing resources.

With this integration, Mistral can now offer developers a seamless environment for running and scaling AI models. This capability is essential for optimizing the use of Graphics Processing Units (GPUs), which are critical to running AI models. Koyeb’s technology allows for granular control over these resources, enabling sub-200ms startup times and efficient inference scaling. 

As part of the acquisition, Koyeb’s team of 16 engineers, including its three co-founders, will join Mistral’s engineering division in March 2026. 

Mistral as a flywheel for European tech sovereignty

The acquisition of Koyeb is another step in Mistral’s expansion, underpinned by the mission to boost European tech sovereignty. In September 2025, Europe’s most valued tech company, chip-producing machines maker ASML, invested $1.5 billion in the French AI lab. Although unusual, the move has been praised for its aim of strengthening the tech ecosystem. 

Moreover, last week Mistral AI announced a €1.2 billion investment in Swedish digital infrastructure. The plan's primary action is the construction of an AI-focused data center in Borlänge. The facility will go live in 2027 and will host NVIDIA’s latest-generation GPUs, bringing cutting-edge AI compute capacity to Sweden. 

The facility will provide advanced computing capacity and localized AI capabilities matching the demand of upcoming AI models. According to the French company, its AI-native infrastructure is designed for performance, efficiency, and full European control, supporting Europe’s technological independence. 

Mistral AI's vertical integration

The convergence of Koyeb’s acquisition and its investment in a dedicated data center demonstrates the company’s intention to build its own AI cloud. By controlling the stack from the serverless deployment layer (Koyeb) down to the physical data center (Sweden), Mistral can optimize performance and squeeze out inefficiencies that third-party clouds cannot address.

Under its Stargate project, OpenAI has invested $500 billion to construct a network of datacenters in the United States. Moreover, the AI giant also revealed plans to build its first datacenter in Europe, committing $2 billion in investment. 

To this end, Mistral is also moving toward vertical integration, building infrastructure to train tomorrow's AI models.