Microsoft and Mistral expand partnership with multibillion-dollar European AI infrastructure build-out

Microsoft and Mistral AI expanded their partnership on July 21, 2026 with a multibillion-dollar agreement to build out the French startup’s AI infrastructure in Europe using thousands of Nvidia Vera Rubin GPUs, the companies said.

The build-out covers training, inference and large-scale deployment, and it deepens a European alternative to the US-centric AI stack at a moment when regulated industries want frontier models they can run under their own control. Microsoft declined to specify the dollar figure and said the deal includes no new equity investment in Mistral.

Two Mistral models, Medium 3.5 and OCR 4, are being added to Microsoft Foundry, with Medium 3.5 also available in Copilot Studio. Through Azure Local, customers in finance, healthcare and manufacturing will be able to run Mistral’s models in cloud-hosted, customer-controlled, or fully disconnected air-gapped configurations, the pitch aimed squarely at data-sensitive sectors.

Arthur Mensch, Mistral co-founder and chief executive, said two-thirds of the company’s existing customers already work with Microsoft, framing the expansion as meeting demand rather than creating it. Brad Smith, Microsoft’s vice chair and president, said Europe “should have access to the world’s most capable AI without compromising control over their data, operations or digital future.”

The arrangement stops short of the equity ties Microsoft holds elsewhere, keeping Mistral independent while binding its distribution to Azure. How much of Mistral’s compute ends up running on Microsoft’s cloud, versus its own European build-out, will show whether that balance survives as spending scales.