Mistral AI's new ambitions

With its own cloud, the French start-up wants to offer an alternative to US giants

Mistral AI no longer sees itself merely as a European alternative to OpenAI, Anthropic, or DeepSeek. The French startup specializing in generative artificial intelligence also dreams of becoming a competitor to Amazon Web Services, Microsoft Azure, and Google Cloud. On Wednesday, it announced the launch of its own cloud computing platform.

Called Compute, this service will be paired with a software layer designed to help companies quickly develop and deploy AI services. To bolster the credibility of this ambitious project, Mistral emphasizes its ties with Nvidia, which it refers to as a partner. In practice, its role may be limited to supplying 18,000 GB200 superchips — the most powerful AI GPU in its lineup. This would represent an unprecedented level of computing power for a European player.

European governments and companies “need a European solution”

Mistral also highlights the experience it has gained since launching two years ago. “We’ve spent most of our time operating GPUs and building a platform to create applications,” said CEO Arthur Mensch at the Vivatech conference, where he shared the stage with none other than Nvidia CEO Jensen Huang.

But Mistral’s strongest selling point remains the ongoing debates around technological sovereignty — just as with its large language models. European governments and companies “need a European solution,” Arthur Mensch argues. But, so far, no European cloud provider truly rivals the US giants, either in computing power or software offerings. Orange, BNP Paribas, and the Kyutai research lab will be among Mistral’s first clients.

The start-up plans to build a data center roughly 30 kilometers south of Paris, with capacity that can later be expanded. It hopes to launch operations in early 2026. The startup notes that clients will be able to use other open-source models on its cloud, not just its own, but not proprietary paid models.

$100 million of annualized revenues

The question of financing remains unclear. The cost of 18,000 Nvidia GPUs alone exceeds $1 billion. It’s roughly the same as all the funding Mistral has raised since its founding in spring 2023. A new funding round, possibly also for $1 billion, could take place at the end of the year — potentially involving Nvidia, which has already invested in the US-based CoreWeave platform. “We would love to be able to fund a company like Mistral,” said Jensen Huang.

By entering the cloud market, the startup is not only seeking a new source of revenue growth but also a highly profitable line of business. According to Arthur Mensch, the company is nearing an annualized revenue rate of $100 million — three times more than at the end of 2024, but still a hundred times less than OpenAI. Realistically, Mistral is not playing in the same league as the maker of ChatGPT.

Its goal is not to offer models as powerful as those from the major US players, but rather to provide models that are competitive enough to serve as a credible alternative. In recent months, it has signed a growing number of commercial deals with major French companies, not only for reasons of sovereignty but also because partnering with the sector’s French champion now carries significant brand value.