Microsoft says Anthropic’s Claude is now generally available in Microsoft Foundry on Nvidia GPU infrastructure

Microsoft has confirmed that Anthropic’s Claude models are now generally available in Microsoft Foundry, giving enterprises a way to run Claude inside their Microsoft Azure environments rather than through a separate, standalone deployment. The update, disclosed around June 30, is aimed at organizations moving from early AI experiments to production use cases that require stronger controls over access, data handling, and operational governance.

According to third-party reporting, Claude in Microsoft Foundry is hosted on Microsoft Azure infrastructure powered by Nvidia’s GB300 Blackwell Ultra GPUs. The same reporting says the deployment follows a broader partnership dynamic among Microsoft, Nvidia and Anthropic, intended to make Claude available for use on Nvidia-accelerated computing rather than requiring customers to manage their own specialized inference stack.

The shift matters because it reduces friction for Azure-native customers. Reporting on the update says organizations can connect to Claude through their existing Microsoft Azure account, leveraging Microsoft’s authentication, billing, and governance tooling. In practical terms, that means teams can build agentic applications, where AI systems take multi-step actions, without restructuring their existing cloud identity and spend controls.

Microsoft also described the role Claude is expected to play inside Foundry’s agent capabilities. The update, as characterized in external coverage, says Microsoft acknowledges Anthropic remains the seller and operator of Claude models in Microsoft Foundry, while Microsoft acts as an independent data processor for prompts and outputs. In addition, Microsoft’s Foundry Agent Service is described as using Claude as a reasoning core for multi-step planning, tool use, and task execution across enterprise systems.

Coverage of the announcement further points to additional Nvidia platform details. One report says Claude in Foundry runs on Nvidia GB300 NVL72 systems and uses Nvidia Quantum-X800 InfiniBand networking, and it links the broader “agentic” push to the need for higher inference performance and efficiency to lower total cost of ownership for ongoing workloads.

On the product side, Microsoft Foundry is part of the company’s broader effort to standardize how enterprises build and deploy AI systems. By placing Anthropic’s models into a Microsoft-managed environment, Microsoft is trying to capture more of the enterprise workflow around AI, including where models are invoked, how agents coordinate with tools, and how enterprise administrators apply security and governance policies.

What Microsoft did not disclose in the available reporting is equally important. Specific availability details such as which Claude versions are included in the general availability rollout, the current quotas or rate limits for inference requests, and any customer-specific contract terms were not provided in the excerpts available for this story. The same limitation applies to performance benchmarks, reliability targets, and detailed data-flow guarantees beyond the high-level roles described for Anthropic and Microsoft.

Next for customers and for Microsoft will be whether this “generally available” step translates into measurable enterprise adoption. Observers will likely look for follow-on announcements that expand model coverage, clarify pricing or consumption mechanics for Foundry usage, and describe how Microsoft’s agent tooling handles enterprise integrations at scale. For Anthropic, the key question is whether broader Azure distribution improves usage while sustaining its role as the provider and operator of Claude’s core model offerings.