On November 18, 2025, Microsoft, NVIDIA and Anthropic announced a set of linked partnerships: Anthropic committed to buy $30 billion of Azure compute capacity, agreed to its first deep technology partnership with NVIDIA, and received investment commitments of up to $10 billion from NVIDIA and up to $5 billion from Microsoft. On the same day, Claude models arrived in public preview in Microsoft Foundry. For .NET teams on Azure, the deal turned Claude from a separate vendor relationship into a model they can deploy, bill and secure like any other Azure service.

Key Facts#

  • Announced: November 18, 2025, jointly by Microsoft, NVIDIA and Anthropic.
  • Azure commitment: Anthropic committed to purchase $30 billion of Azure compute capacity and to contract additional capacity of up to one gigawatt.
  • NVIDIA partnership: the two companies will collaborate on design and engineering, with an initial Anthropic compute commitment of up to one gigawatt on NVIDIA Grace Blackwell and Vera Rubin systems.
  • Investments: NVIDIA committed to invest up to $10 billion and Microsoft up to $5 billion in Anthropic.
  • Foundry models: Claude Sonnet 4.5, Claude Haiku 4.5 and Claude Opus 4.1 entered public preview in Microsoft Foundry through serverless deployment.
  • Copilot: Microsoft committed to continuing access to Claude across GitHub Copilot, Microsoft 365 Copilot and Copilot Studio.
  • Multi-cloud: Anthropic said the deal makes Claude the only frontier model available on all three of the most prominent cloud platforms. Amazon remains its primary cloud provider and training partner.

What Happened#

The announcement combined three agreements. The first is a compute purchase. Anthropic will scale Claude on Azure, powered by NVIDIA hardware, with a $30 billion commitment and an option to contract up to one gigawatt of additional capacity. The companies said this would broaden access to Claude and give Azure enterprise customers more model choice.

The second agreement is new for Anthropic: for the first time, it established a deep technology partnership with NVIDIA. The two companies said they would optimize Anthropic's models for performance, efficiency and total cost of ownership on NVIDIA hardware, and would tune future NVIDIA architectures for Anthropic's workloads. The initial capacity covers Grace Blackwell systems and the later Vera Rubin generation.

The third agreement widened Claude's distribution through Microsoft. Anthropic's companion post said Claude Sonnet 4.5, Haiku 4.5 and Opus 4.1 were available in public preview in Microsoft Foundry, where Anthropic manages the serving infrastructure. Claude usage in Foundry is eligible for the Microsoft Azure Consumption Commitment (MACC) and works with existing Azure agreements and billing. Anthropic said developers can call Claude with Python, TypeScript and C# SDKs using Microsoft Entra authentication, and can point Claude Code at Foundry. Deployment started with the Global Standard option at standard API pricing, with a US DataZone option described as coming soon. Anthropic also said Claude became an option in preview in Agent Mode in Excel, in addition to existing Microsoft 365 Copilot integrations such as the Researcher agent.

Anthropic CEO Dario Amodei, Microsoft CEO Satya Nadella and NVIDIA CEO Jensen Huang appeared together to discuss the partnerships.

Background#

Anthropic had spent 2025 building what it calls a diversified compute strategy across three chip platforms. Amazon's Trainium chips power Project Rainier, and AWS remains Anthropic's primary training partner. On October 23, 2025, Anthropic said it would expand its use of Google Cloud to as many as one million TPUs. On November 12, it announced a $50 billion investment in U.S. data centers with Fluidstack. The Microsoft and NVIDIA agreement added NVIDIA GPUs at large scale through Azure, completing the set.

For Microsoft, the deal extended a relationship that already included Claude inside Microsoft 365 Copilot and Copilot Studio, and it expanded the model choice available to Azure customers. Our coverage of the amended Microsoft and OpenAI partnership describes how Microsoft's model strategy evolved in 2026.

Why It Matters for Developers#

In our view, this is the most directly practical of Anthropic's compute deals for .NET developers, because it changes how Claude reaches Azure workloads.

  • One control plane for procurement and security. Deploying Claude through Foundry lets you use Entra ID authentication instead of long-lived API keys, apply familiar Azure governance, and count spend toward existing commitments. Our guide to Azure OpenAI and Azure AI Foundry for .NET covers deployments, identity and quotas.
  • Model choice without rewrites. With Claude joining the OpenAI models already offered on Azure, the case for abstraction is stronger. Code against IChatClient from Microsoft.Extensions.AI and choose models per feature through configuration, so an evaluation result or a price change does not force a refactor.
  • Check feature parity per platform. Anthropic listed tool use, prompt caching, citations, vision, code execution and web search among the capabilities supported in Foundry, but availability can lag or differ between platforms. Verify each feature you depend on, and each region, before you commit an architecture to it.
  • Mind data residency. The launch used a Global Standard deployment, with a US DataZone to follow. Teams with residency obligations should confirm where requests are processed before sending regulated data.

For developer tooling, Microsoft's commitment to keep Claude available in GitHub Copilot matters too: teams that prefer Claude models for coding can keep them inside the tools they already use, as our guide to AI-assisted .NET development describes.

What's Next#

Part of this capacity was always expected to arrive later: in April 2026, Google still described NVIDIA's Vera Rubin as a next-generation platform due to become available later that year. The partnership remained central in 2026. In May 2026, Anthropic listed "a strategic partnership with Microsoft and NVIDIA that includes $30 billion of Azure capacity" among the compute agreements behind higher Claude usage limits. It also signed a separate Amazon agreement for up to 5 GW. Later model launches continued to reach Azure: Anthropic said Claude Opus 5.5 became available on all platforms, including Microsoft Azure, on September 22, 2026.

Open questions include how Azure capacity is split between training and serving, and how quickly Foundry gains each new Claude feature.

Sources#