Microsoft locked in two separate deals this week, one stretching its relationship with data giant Databricks well into the 2030s, the other giving European businesses a new option for running AI workloads without leaving French soil.
The Databricks agreement covers a lot of ground. The company plans to run more of its internal operations and analytics on Azure, while shifting more workloads onto Microsoft’s own Cobalt processors, Arm-based chips designed for data-heavy tasks and agentic AI applications. Databricks is currently on Cobalt 100 and plans to move to Cobalt 200, which Microsoft says delivers up to 50% better performance depending on the workload, along with memory encryption turned on by default.
Beyond infrastructure, Microsoft is weaving Databricks’ software deeper into its own product stack. That includes Genie, a conversational analytics tool that lets people query business data using plain language, and Unity AI Gateway, which Databricks describes as a central control layer for managing AI models, agents, and access policies across an organization.
Both tools will now integrate more tightly across Microsoft services including Entra, Power BI, Teams, and Copilot. Databricks CEO Ali Ghodsi framed the integration as a way to help enterprises unify scattered data and ground their AI systems in actual business context. The company also disclosed a new funding round valuing it at $188 billion, expected to close later this summer.
Separately, Microsoft struck a multibillion-dollar infrastructure deal with French AI company Mistral, though neither side disclosed the total value. Under the agreement, Azure customers gain access to Mistral’s data centers in France, giving European and regulated organizations another way to run AI workloads on infrastructure physically located in France while still using Microsoft’s software and security tools.
Microsoft President Brad Smith described the setup as blending American and European technology in a way that gives customers continuous, assured access, regardless of where geopolitical or regulatory winds shift. Mistral is targeting one gigawatt of computing capacity by 2030, built largely around Nvidia’s newer Vera Rubin GPUs, though the companies haven’t said how much of that buildout Microsoft’s spending will actually cover.
The distribution side matters too. Mistral’s Medium 3.5 and OCR 4 models are now available through Microsoft Foundry, and Medium 3.5 is accessible via Copilot Studio, giving Mistral’s models a much wider audience inside Microsoft’s ecosystem than before. Smith confirmed the deal doesn’t include a new equity stake, since Microsoft and Nvidia had already invested in Mistral previously.
