Related: A Chinese open model now matches Anthropic’s Mythos at finding security bugs.
For the first time, Anthropic‘s Claude models are running on NVIDIA hardware. The two companies, together with Microsoft, said Claude is now generally available in Microsoft Foundry, hosted on Azure and powered by NVIDIA’s GB300 Blackwell Ultra GPUs. The arrangement gives businesses that already live inside Microsoft’s cloud a direct path to one of the most capable AI model families, and it marks a notable crossing of lines in an industry where the big labs, chipmakers, and cloud providers are increasingly partners and rivals at the same time.
The headline is the hardware pairing. Until now Anthropic had not deployed Claude on NVIDIA chips, so running the models on GB300 Blackwell Ultra is a first for the company. NVIDIA says the deployment uses its GB300 NVL72 systems with Quantum-X800 InfiniBand networking, the rack scale setup built for the heavy, sustained compute that large models demand. For enterprises the technical detail matters less than the outcome, which is that Claude now runs on the newest generation of NVIDIA’s data center silicon inside the cloud many of them already pay for, rather than only through Anthropic’s own interfaces.
The target customer is the enterprise trying to build agents rather than chat. NVIDIA frames the launch around autonomous and domain specific AI agents, the kind that can carry out multi step business tasks instead of simply answering a prompt, and points to a Secure Agent Workspace reference design that lets companies run those agents in a governed environment where identity, network access, credentials, and runtime policy are controlled at the infrastructure level. That governance angle is the quiet center of the announcement. Large organizations have been cautious about turning capable agents loose on real systems, and the pitch here is that Foundry plus NVIDIA’s tooling gives them a way to do it with guardrails rather than hope.
It also extends a partnership that was set in motion months ago. Claude in Microsoft Foundry on NVIDIA GPUs builds on the strategic agreement Microsoft, NVIDIA, and Anthropic announced in November 2025 to expand enterprise access to Claude and put Anthropic’s models on NVIDIA accelerated computing. Bringing the models to general availability on Azure is the part of that deal customers can actually use, moving it from a press release into a product they can switch on. NVIDIA is also working with Anthropic to fold its developer tools into the Anthropic stack through verified agent skills, so Claude agents can pick up domain specific abilities tied to NVIDIA’s accelerated computing.
For readers tracking the strategy, the interesting wrinkle is how tangled these alliances have become. Microsoft is a major backer of OpenAI, Anthropic’s chief rival, yet it is now hosting Claude on its cloud and steering enterprise customers toward it. NVIDIA sells the chips that power nearly every serious AI lab, including ones building hardware meant to reduce dependence on NVIDIA. Anthropic, which has leaned on other accelerators, is now running on NVIDIA inside a Microsoft data center. The throughline is demand. Enterprises want choice among top models in the environments they already use, and the companies that supply the picks and shovels would rather sell to everyone than bet on a single winner.
What it adds up to is less a dramatic shift than a steady normalization of Claude as enterprise infrastructure. The model is becoming something a company can provision in its existing cloud, govern with familiar controls, and point at real workflows, which is exactly the unglamorous plumbing that decides whether AI agents move from pilots into daily operations. Specific performance and pricing figures for the Azure deployment have circulated in early coverage, but the numbers worth trusting are the ones Microsoft and Anthropic publish in the official documentation, so the safest read today is the strategic one. A leading model, the newest NVIDIA silicon, and the enterprise cloud are now lined up in a single offering, and the contest ahead is over which lab’s agents companies actually choose to build on.
Related on Entrelligence: Anthropic’s drug discovery program for neglected diseases, a Chinese open model matching Anthropic’s Mythos, and the US lifting export controls on Claude Fable 5.
Your First 10 AI Skills
10 practical AI skills, copy-paste prompts and a 7-day plan to start using AI with confidence.
Download the guide →