Claude in Microsoft Foundry Is Now Generally Available on Azure

Claude in Microsoft Foundry is now generally available, and it changes the calculus for enterprises that want frontier AI without leaving the Azure environment they already govern. Announced on June 29, 2026 by Azure Product Lead Steve Sweetman, the release gives developers and IT teams a production-ready path to build with Anthropic’s Claude models using the authentication, billing, and compliance controls their organizations already trust. For Microsoft 365 shops that have spent years standardizing on Azure and Entra ID, this is a meaningful expansion of model choice.

Source: Microsoft Azure Blog

What Is Claude in Microsoft Foundry?

Claude in Microsoft Foundry lets enterprises access Anthropic’s Claude models directly through their existing Azure account, hosted on Azure infrastructure rather than a separate third-party service. Instead of standing up new procurement, networking, and governance processes, teams reuse what they already have in place for other Foundry models.

According to Microsoft, most enterprise AI projects don’t stall because of model quality. They stall because of everything around the model: procurement, governance, networking, and data handling. Consequently, Claude in Microsoft Foundry is positioned less as a new model launch and more as a plumbing fix for the barriers that keep pilots from becoming production systems.

How Developers Access Claude in Microsoft Foundry

Developers reach Claude through the Messages API, with support for prompt caching, extended thinking, and tool streaming. For teams building autonomous agents rather than single-turn chat features, Foundry Agent Service uses Claude as the reasoning core to orchestrate multi-step planning, tool use, and task execution across enterprise systems.

Because Claude is available natively inside Foundry, teams authenticate with Microsoft Entra ID, apply existing Azure role-based access controls, and track usage through familiar Azure management tools. Inference runs inside Azure, and customers can choose between Global and US data zones for workloads with data residency requirements. Anthropic operates the underlying inference and remains the data processor and SLA provider.

• Access via the Messages API with prompt caching, extended thinking, and tool streaming
• Foundry Agent Service uses Claude to orchestrate multi-step, goal-driven agents
• Authentication through Microsoft Entra ID and existing Azure RBAC policies
• Choice of Global or US data zones for data residency requirements
• Billing consolidated as Claude Consumption Units (CCU) on the standard Azure bill, with MACC drawdown supported

Built on the Microsoft, NVIDIA, and Anthropic Partnership

This general availability milestone builds directly on the strategic partnership Microsoft, NVIDIA, and Anthropic announced in November 2025. Under that agreement, Anthropic committed to purchase $30 billion of Azure compute capacity, with the option to contract up to one gigawatt of additional capacity on NVIDIA Grace Blackwell and Vera Rubin systems. In turn, NVIDIA and Microsoft committed to invest up to $10 billion and $5 billion in Anthropic, respectively.

That deal made Claude the only frontier model family available on all three of the world’s largest cloud platforms. Today’s release makes good on the enterprise-facing half of that promise: Claude runs on NVIDIA Blackwell Ultra systems connected by InfiniBand networking, giving Azure customers rack-scale infrastructure built specifically for inference performance and efficiency.

Source: The Official Microsoft Blog

Governance, Security, and Cost Controls IT Teams Will Notice

For high-sensitivity workloads, zero data retention is available, meaning prompts and completions are not retained by Anthropic once the API call finishes. Additionally, Foundry Control Plane continuously runs evaluations to check that agent responses match customer expectations, and it can block responses that violate rules before they ever reach a user.

Cost management gets a boost too. Microsoft’s model router feature can automatically route queries to the most appropriate Claude model for a given task, which Microsoft says saves up to 50% on inference costs while improving user satisfaction scores. For finance and IT leaders tracking AI spend, that combination of governance and cost control may matter as much as raw model capability.

Real Customers Are Already Building With Claude in Foundry

Microsoft didn’t launch this quietly. Several enterprise customers went on record describing production use, not just pilots. NVIDIA said it uses autonomous Claude-powered agents daily to help internal teams move faster on complex technical work, now running on NVIDIA GB300 GPUs inside Foundry.

Momentic, a software testing company, said its customers describe tests in plain English while Claude’s Opus models verify releases before they ship, now serving millions of tokens per minute through Foundry. Meanwhile, Everstar, a nuclear technology company, said the combination of Anthropic’s models and Azure’s security compressed a safety analysis that would have taken 200 human days into a single day. Bolt, a fintech platform, credited the setup with the sustained throughput and reliability its Fortune 500 customers expect.

Source: Microsoft Azure Blog

Where This Fits in Microsoft’s Multi-Model Strategy

Microsoft has been steadily moving away from a single-model story. Azure OpenAI remains central to the platform, but Microsoft has also shipped its own in-house MAI models and now offers Claude natively alongside them. Rather than forcing customers to pick one vendor, Foundry increasingly functions as a neutral marketplace where teams choose the best model for a given job and manage them all under one governance layer.

That matters beyond Azure, too. Microsoft has committed to continuing Claude access across its broader Copilot family, including GitHub Copilot, Microsoft 365 Copilot, and Copilot Studio. As a result, organizations already standardized on the Microsoft 365 Copilot ecosystem may see Claude-powered capabilities surface in more places over time, not just inside the Foundry portal.

What Claude in Microsoft Foundry Means for Microsoft 365 and Power Platform Admins

For admins who already manage Copilot Studio agents or custom SPFx-based Copilot experiences, this announcement is worth watching even if it doesn’t require immediate action. Foundry Agent Service is the same orchestration layer increasingly used to power custom agents across Microsoft 365, so model choice at the Foundry level can eventually influence what’s available to line-of-business agent builders.

Practically speaking, admins evaluating Claude in Microsoft Foundry should start by reviewing existing Azure governance policies, since Entra ID and RBAC controls carry over directly. Teams with strict data residency requirements should confirm which data zone, Global or US, fits their compliance posture before piloting a workload. Finally, finance teams should watch for Claude Consumption Units on the Azure bill, since it consolidates spend differently than per-model line items customers may be used to.

Getting Started: How to Try Claude in Microsoft Foundry

Teams curious about Claude in Microsoft Foundry don’t need a lengthy procurement cycle to kick the tires. Because it ships inside the same Foundry portal used for other models, getting a first project running mostly comes down to a few configuration choices rather than a new vendor relationship.

1. Open the Anthropic model catalog in Azure AI Foundry. Existing Azure subscribers can browse available Claude models directly from the Foundry portal without a separate Anthropic account.
2. Pick a data zone. Decide between Global and US data zones based on your organization’s data residency and compliance requirements before deploying anything to production.
3. Wire up Foundry Agent Service if you’re building agents. Rather than calling the Messages API directly for a single feature, most enterprise teams will get more value orchestrating Claude through Foundry Agent Service, which handles multi-step planning and tool use automatically.
4. Set budget guardrails early. Because usage bills as Claude Consumption Units, finance and platform teams should agree on spending thresholds before a pilot scales into a department-wide rollout.

None of these steps require new Entra ID tenants or separate compliance reviews, since the underlying identity and access controls are the same ones already governing other Foundry workloads. That’s precisely the friction Microsoft says it built this release to remove.

How Claude in Microsoft Foundry Compares to Other Cloud Options

Claude has been available through Amazon Bedrock and Google Vertex AI for some time, so Azure was the notable gap until this release. With Claude in Microsoft Foundry now generally available, Anthropic’s frontier models are available on all three major hyperscalers, giving enterprises consistent model access no matter which cloud platform they’ve already standardized on.

For organizations already deep into the Microsoft ecosystem, however, the Azure-native integration is the real differentiator rather than mere availability. Unlike accessing Claude through a separate vendor console, Foundry customers keep a single pane of glass for billing, identity, and governance across every model they use, including Claude, Azure OpenAI, and Microsoft’s own MAI models. That consolidation is often what tips a proof-of-concept toward production approval.

Quick Answers: Claude in Microsoft Foundry FAQ

• When did Claude in Microsoft Foundry become generally available? June 29, 2026, following the November 2025 Microsoft-NVIDIA-Anthropic partnership announcement.
• Who processes the data? Inference runs on Azure infrastructure, but Anthropic operates the inference itself and remains the data processor and SLA provider. Zero data retention is available for sensitive workloads.
• How is it billed? Through Claude Consumption Units (CCU), consolidated as a single line on the Azure bill, with MACC drawdown supported and per-model detail still visible in Foundry.
• Does this affect Microsoft 365 Copilot? Not directly today, but Microsoft has committed to continuing Claude access across GitHub Copilot, Microsoft 365 Copilot, and Copilot Studio.
• What infrastructure does it run on? NVIDIA Blackwell Ultra systems connected by InfiniBand networking, part of the broader Microsoft-NVIDIA-Anthropic compute partnership.

Want more coverage of how Microsoft’s AI ecosystem is evolving across Azure, Microsoft 365, and Power Platform? Keep exploring SharePoint Monkey as we track what Claude in Microsoft Foundry and similar announcements mean for enterprise IT teams.

Sources

• Claude in Microsoft Foundry is now generally available (Microsoft Azure Blog)
• Claude Sonnet 5 Is Now Generally Available in Microsoft Foundry (Microsoft Community Hub)
• Microsoft, NVIDIA and Anthropic announce strategic partnerships (The Official Microsoft Blog)
• Claude and Microsoft Foundry partnership overview (Anthropic)
• Anthropic models catalog in Microsoft Foundry (Azure AI Foundry)
• What is Microsoft Foundry Agent Service? (Microsoft Learn)