Serverless AI application platform Koyeb has entered into a definitive agreement to join French software company Mistral AI with the aim of advancing AI infrastructure.
Koyeb’s serverless approach runs AI apps across CPUs, GPUs and accelerators. Teams use Koyeb to run inference jobs, to execute code in isolated sandboxes (for prototyped experimental workloads) and to deploy production workloads with automatic scaling and usage-based pricing. Paris, London and Palo Alto headquartered Mistral AI develops high-performance, open-weight large language models (LLMs) focused on efficiency, accessibility and portability.
Mistral Compute in Europe
Koyeb confirms that it will bring its platform, technology and team to accelerate the Mistral Compute offering. By way of definition, Mistral Compute is a European-hosted AI cloud that brings together GPU capacity, turnkey orchestration and advanced model-building tools in one integrated platform for developing and running AI.
Being an open-weight LLM, Mistral is differentiated by its hosting flexibility i.e., it can run anywhere across local, cloud or edge. It has inspectable parameters rather than being black box as we would see in a fully closed model; and it is charged on infrastructure costs only, rather than being costed on a pay-per-use-per-token basis.
Serverless, Inference, Agents
“In 2021, we set out to build the next-generation of cloud infrastructure, with a seamless serverless experience backed by high-performance hardware, deployable worldwide in seconds. After [our total of] 10 years building and operating large-scale cloud platforms, we saw traditional CPU infrastructure getting commoditized while advanced, scalable, high-performance platforms remained complex and expensive to operate,” write Koyeb CEO Yann Léger, chief product officer Edouard Bonlieu and chief technology officer Bastien Chatelard (all company co-founders) on the organization’s blog.
So, what’s happened in the last half-decade as a result?
The team says that the rise of generative AI and agentic workloads has redefined what cloud infrastructure needs to offer, i.e., it has to deliver a dramatic increase in its ability to scale and also offer a set of new requirements.
While the French co-founding software engineers don’t say specifically, we can infer that they are referring to the evolution of AI from mere chatbots to agents have the ability to reason, plan and act; all of which means cloud infrastructure has to now support stateful (so that the system retains context between requests), long-running, secure execution environments.
No Ops Overhead
Koyeb has worked to provide serverless GPUs, specialized accelerators and CPU workloads with capabilities to increase efficiency, enable new use-cases and control costs. All of which means that, today, Koyeb operates a purpose-built serverless platform for teams to run AI applications, from APIs and agents on CPU to high-performance inference on GPUs, with no Ops overhead, servers, or infrastructure management.
So, Why Join Mistral AI?
“When we first met the Mistral AI team and heard about their vision for Mistral Compute, we saw an incredible opportunity to accelerate the buildup of AI Infrastructure in Europe. Mistral AI is not just any company; it’s an incredibly fast venture, building frontier AI, running it in enterprises and scaling it to be accessible by everyone. They are now doing the same for AI Infrastructure,” blogged the team, who also note that they saw eye to eye with Mistral AI’s co-founder and CTO, Timothée (Tim) Lacroix.
The Koyeb platform will continue operating. In the coming months, Koyeb is expecting to transition to become a core component of Mistral Compute and will double down on its inference, sandboxes and serverless capabilities for MCP servers.
Nothing is changing for users in the immediate days and months ahead. Users can continue using Koyeb exactly as they do today and Koyeb will continue to bill users directly through its existing billing system.
