NVIDIA highlights Claude on Blackwell Ultra as Anthropic expands availability in Microsoft Foundry

NVIDIA is drawing fresh attention to its Blackwell Ultra platform through an update tied to Anthropic’s Claude models. According to coverage dated June 29, Claude is available in Microsoft Foundry on Microsoft Azure, where it runs on NVIDIA GB300 Blackwell Ultra GPUs.

The announcement frames the Azure deployment as part of the push toward “agentic” AI, where models are used not only to answer questions, but to coordinate sub-tasks and take action across multiple business functions. The reporting says NVIDIA and Anthropic are working together to expand developer capabilities, with NVIDIA-validated agent skills intended to give Claude domain-specific abilities.

Technically, the deployment described in the posts is anchored on NVIDIA GB300 NVL72 systems and NVIDIA Quantum-X800 InfiniBand networking. The coverage also says the setup is designed to improve inference performance and efficiency, with the argument that better inference economics can reduce total cost of ownership when organizations run AI at scale.

For enterprises building on Azure, the coverage points to a “secure agent workspace” reference design. In plain terms, that blueprint is presented as a way to run autonomous agents in a governed environment where infrastructure-level controls manage identity, network access, credentials, and runtime policy.

The update also emphasizes integration between NVIDIA’s accelerated computing stack and Anthropic’s Claude in the Microsoft Foundry environment. That combination is described as intended to help organizations develop domain-specific and “self-operating” agent systems, including autonomous sub-agents that can work across areas of an organization.

While the market headlines focus on NVIDIA’s exposure to the broader AI spending cycle, the concrete takeaway from the available material is narrower. The posts do not provide new financial guidance, contract values, or customer names, and they do not quantify how many workloads are expected to move to GB300-based infrastructure as a result of the Foundry availability.

Company and sector context is still relevant. Nvidia’s strategy has increasingly centered on enabling inference and training for large language models across major cloud providers, and Blackwell Ultra is positioned as the compute foundation for next-generation AI applications. In that sense, product availability indicates demand momentum for enterprise-grade deployments rather than consumer features.

Still, several details are not disclosed in the materials summarized here. The postings do not specify pricing for Claude on Foundry or any throughput, latency, or performance benchmarks for GB300. They also do not clarify the exact scope of “generally available” beyond the Foundry and Azure context, nor do they outline any timeline for additional model updates or tooling changes.