Anthropic deploys Claude on Nvidia GPUs in Azure for first time

Anthropic announced on Monday that its Claude family of AI models is now generally available in Microsoft Foundry on Azure ↗ 2.38%, running on Nvidia's ↘ 0.52% GB300 Blackwell Ultra GPU systems — marking the AI startup's first deployment on Nvidia hardware.streetinsider+1

Enterprise AI on Next-Gen Hardware

The deployment gives Azure customers access to Claude models for building autonomous and domain-specific AI agents, as enterprises ramp up investment in agentic AI applications. The models run on Nvidia GB300 NVL72 systems with Quantum-X800 InfiniBand networking, enabling customers to deploy advanced AI agents, including specialized sub-agents that can operate across different business functions.streetinsider

Anthropic said the deployment is designed to improve inference performance and efficiency while reducing the total cost of ownership for enterprise AI workloads. The announcement follows an earlier indication from Nvidia and Microsoft in early June that Claude models would "run natively on NVIDIA GB300 Blackwell Ultra systems on Azure, with customer availability in the weeks ahead".hpcwire+1

A Partnership Months in the Making

The launch builds on a strategic partnership announced by Microsoft, Nvidia, and Anthropic in November 2025, when the three companies revealed a sweeping deal that included Anthropic committing to purchase $30 billion of Azure compute capacity. As part of that agreement, Nvidia and Microsoft committed to investing up to $10 billion and $5 billion respectively in Anthropic.blogs.microsoft+1

At the time, Anthropic's compute commitment was set to initially reach up to one gigawatt of capacity with Nvidia Grace Blackwell and Vera Rubin systems. The partnership also made Claude available across Microsoft's Copilot family, including GitHub Copilot and Microsoft 365 Copilot.blogs.microsoft

Azure's Blackwell Ultra Infrastructure

Microsoft first deployed a large-scale Azure cluster featuring more than 4,600 Nvidia GB300 Blackwell Ultra GPUs in October 2025, initially purpose-built for OpenAI's inference workloads. That cluster, connected through Nvidia's next-generation InfiniBand network, was designed to reduce AI model training times from months to weeks. Anthropic's Claude now joins as a second major AI model family running on this infrastructure, underscoring Azure's role as a multi-model platform for enterprise customers.tweaktown+2