Anthropic’s Claude Models Run Natively on Nvidia GB300 via Microsoft Azure

Anthropic’s Claude models now run natively on Nvidia GB300 Blackwell Ultra systems through Microsoft Azure, completing a large AI infrastructure push agreed in November 2025. The deal totals about $45 billion: Microsoft plans to invest up to $5 billion in Anthropic, while Nvidia commits up to $10 billion. Combined investment lifted Anthropic’s valuation to an estimated $350 billion. Anthropic pledged $30 billion in Azure compute capacity and plans to use up to 1 gigawatt of computing power from Nvidia’s Grace Blackwell and Vera Rubin systems. In performance testing, Azure’s ND GB300 v6 virtual machines reportedly hit inference throughput above 1.1 million tokens per second per rack, positioning the setup as a measurable benchmark for large model deployment. Cloud availability is broad. Claude is now accessible across AWS, Azure, and Google Cloud, reducing platform friction for customers not primarily running on AWS. This also intensifies competition: Google Cloud hosts Claude despite offering its own Gemini models, while Microsoft’s ecosystem is closely tied to OpenAI and AWS also invests in Anthropic. Nvidia’s dual role—funding Anthropic while supplying the GPU infrastructure—creates a potential “flywheel” of demand for more compute, more hardware deployments, and continued revenue for Nvidia. For traders, this is an AI-capex and infrastructure signal rather than a direct token catalyst, but it may support broader risk sentiment around tech/AI supply chains and data-center demand.