Claude Language Models Power Microsoft Foundry on Azure

Anthropic’s Claude language models are now fully available in Microsoft Foundry on Azure. This launch gives businesses direct access to advanced AI through a familiar cloud platform. The service runs on powerful NVIDIA hardware and integrates tightly with Azure’s security and management tools.

Claude Opus 4.8 and Claude Haiku 4.5 models are accessible via the Messages API. They operate on Azure infrastructure that uses NVIDIA GB300 Blackwell Ultra GPUs. These GPUs run on NVIDIA GB300 NVL72 systems, each packed with 72 GPUs and 36 CPUs. These systems deliver 37 terabytes of fast memory and 130 terabytes per second NVLink bandwidth. The FP4 Tensor Core performance reaches up to 1,440 petaflops, enabling high-speed AI processing.

NVIDIA networking technology plays a big role. The deployment uses NVIDIA Quantum-X800 InfiniBand networking. This setup supports powerful agentic systems for complex tasks. NVIDIA said Claude models in Foundry let enterprises build autonomous and specialized sub-agents. These sub-agents can work across different parts of a business to perform advanced work.

Azure Integration and Security Controls

Microsoft designed Claude in Foundry to fit smoothly into existing Azure environments. Customers authenticate using Microsoft Entra ID. They control access with Azure’s role-based access controls and governance policies. This makes it easy for companies to secure their AI workloads with tools they already trust.

Billing is handled through Claude Consumption Units (CCUs). Charges appear right on the Azure bill, simplifying cost management. Customers can pick between Global and US data zones for inference processing. For sensitive workloads, Microsoft offers zero data retention. This means prompts and outputs are not saved after use, protecting privacy.

Flexible Hosting and Cost Efficiency

Microsoft offers two hosting options for Claude in Foundry. One is Azure-hosted, and the other is Anthropic-hosted, which was previously called Foundry Preview. This lets customers choose the setup that fits their needs.

Foundry includes a model router that directs queries to the most suitable Claude model. This routing can cut costs by up to 50%. Microsoft’s Foundry Agent Service also uses Claude as its reasoning core. This helps with multi-step planning, tool use, and task execution, making AI workflows more powerful.

The partnership behind this launch started in November 2025. Microsoft and NVIDIA announced they would invest up to $15 billion in Anthropic. Microsoft will invest $5 billion, and NVIDIA $10 billion. Anthropic committed to buying $30 billion in Azure compute capacity. This deal values Anthropic at $965 billion after its $65 billion Series H funding round.

Anthropic plans to expand Claude’s availability in Microsoft Foundry over time. The partnership aims to deliver advanced AI capabilities to enterprises with strong hardware, cloud integration, and security. As Steve Sweetman, Microsoft Azure product lead, said, “Enterprises can build with Claude through their existing Azure account, using the authentication, billing, networking, governance, and data controls their teams already trust.”