Claude Meets Blackwell Ultra: Anthropic’s Models Now Run on NVIDIA GB300 in Azure

Anthropic's Claude models are now generally available in Microsoft Foundry, hosted on Microsoft Azure and running on NVIDIA GB300 Blackwell Ultra GPUs, per a June 29 post on the NVIDIA corporate blog.

The general-availability milestone gives Azure-native enterprises "a powerful new way to build autonomous and domain-specific AI agents," per the NVIDIA post, which framed the deployment around enterprise demand for computing capacity as agentic AI becomes more autonomous.

The announcement is single-sourced to NVIDIA's own blog, a vendor channel with a commercial interest in the deployment it describes. Pricing, regional availability, capacity commitments and the specific Claude model versions covered by the general-availability designation were not specified in the available material, and direct statements from Anthropic or Microsoft were not part of the material reviewed.

Background

The deployment sits at the junction of a three-way alignment announced in November 2025, when Anthropic, Microsoft and NVIDIA set out a partnership under which Anthropic committed to large-scale Azure compute purchases and its Claude models were slotted into Microsoft's enterprise AI stack. The arrangement made Claude available alongside other frontier models in Microsoft Foundry, the Azure platform through which Microsoft offers hosted models, agent tooling and deployment services to enterprise customers — a notable broadening for a company whose model hosting had been most closely associated with Amazon Web Services and Google Cloud.

GB300 Blackwell Ultra is NVIDIA's top-end data-centre accelerator generation, packaged in rack-scale systems aimed at training and high-throughput inference for frontier models. Cloud providers have raced to field the hardware, and vendors routinely publicise which flagship models run on it: such posts serve simultaneously as customer documentation and as marketing for the chips, the cloud and the model — the commercial triangle that makes single-sourced availability claims worth flagging.

For enterprises, general availability is a meaningful contractual threshold as it typically brings service-level commitments and production support that preview offerings lack. Which Claude versions, in which Azure regions and at what price, are the details that determine how much practical capacity the milestone represents.

What comes next

Confirmation and detail would ordinarily follow through Microsoft Foundry's model-catalogue documentation and Anthropic's own channels, including covered model versions, regional rollout and pricing. Watch for whether Microsoft and Anthropic issue their own general-availability notices, and for any capacity or adoption figures in the companies' subsequent disclosures.