Anthropic's Claude models are now generally available (GA) in Microsoft Foundry, the development platform on Microsoft Azure. They run on NVIDIA's latest GPU, the GB300 Blackwell Ultra, making it easier for enterprises on Azure to build and operate autonomous AI agents as well as agents tailored to their specific business domains[1]. The move builds on the strategic partnership between Microsoft, NVIDIA, and Anthropic announced in November 2025.
Claude Reaches General Availability in Microsoft Foundry
The key point is that Claude can now be used officially from Microsoft Foundry, the development platform running on Azure. There were already ways to access Anthropic's models in a preview-style form, but general availability makes it easier for enterprises that build their systems around Azure to incorporate Claude into production environments[1].
NVIDIA positions this release as a new option for building and running autonomous agents and specialized agents divided by business domain. As agentic AI moves deeper into enterprise operations, a computing foundation that can handle inference quickly and efficiently becomes essential, and that in turn helps hold down total cost of ownership (TCO)[1].
The Foundation: NVIDIA GB300 Blackwell Ultra
Claude's execution is supported by NVIDIA's GB300 NVL72 systems. The GB300 Blackwell Ultra is NVIDIA's latest-generation accelerator built for AI training and inference, and a defining feature is its ability to bundle multiple GPUs at high speed and treat them as a single, massive pool of compute[1].
Within the system, the GPUs are connected by a high-speed network called NVIDIA Quantum-X800 InfiniBand. When running large models, the volume of data exchanged between GPUs becomes enormous, so the speed of this internal network shapes overall processing efficiency. This run on the GB300 platform is described as Anthropic's first full-scale deployment on NVIDIA hardware[1].
Mechanisms That Support Enterprise Agent Operations
NVIDIA is also collaborating with Anthropic by integrating NVIDIA's own tools into Claude's stack. This makes it easier for enterprises to give Claude-based agents specialized capabilities for each business domain. Through NVIDIA-verified agent skills, NVIDIA describes a vision in which AI agents are embedded deeply into internal operations and used as the operating system that runs the entire organization[1].
On the operations side, a design guideline called the NVIDIA Secure Agent Workspace Reference Design has also been provided. It is a blueprint for running autonomous agents in a governed environment, allowing factors such as identity, network access rights, credentials, and runtime policy to be controlled at the infrastructure level. It can be seen as a foundation for operating agents safely so they do not act beyond their bounds[1].
The November Three-Party Partnership as the Groundwork
This general availability sits on the extension of the strategic partnership announced by Microsoft, NVIDIA, and Anthropic in November 2025. Under that partnership, Anthropic stated it would purchase 30 billion USD (about 4.8 trillion yen) worth of Azure compute resources, and signaled a plan to secure up to one gigawatt of additional compute capacity[2].
On the capital side, plans were also disclosed for NVIDIA to invest up to 10 billion USD (about 1.6 trillion yen) and Microsoft up to 5 billion USD (about 800 billion yen) in Anthropic. In addition, Claude is said to become the only frontier model available across all three of the world's most prominent cloud services, and Microsoft Foundry was described as offering access to flagship models such as Claude Sonnet 4.5, Claude Opus 4.1, and Claude Haiku 4.5[2][3].※1 USD = 161 JPY (as of June 26, 2026)
Summary
Claude's general availability in Microsoft Foundry was realized on the powerful foundation of NVIDIA's latest GB300 Blackwell Ultra. With the GB300 NVL72 and its high-speed internal network, plus verified agent skills and a design guideline for safe operation, the environment for Azure-centric enterprises to embed Claude as a business agent is steadily taking shape. It marks a milestone showing that the vast compute resources and investments outlined in the November 2025 three-party partnership are beginning to move in the form of concrete, available offerings.
Source:https://blogs.nvidia.com/blog/anthropic-nvidia-gb300-blackwell-ultra-microsoft-azure/
Source:https://www.anthropic.com/news/microsoft-nvidia-anthropic-announce-strategic-partnerships
