NVIDIA GB300 GPU Powers Anthropic Claude Model Now Generally Available on Azure
Miles Bennett
Nvidia, Anthropic, and Microsoft have shipped their first joint product — Claude in Microsoft Foundry, powered by GB300 Blackwell Ultra GPUs and now live on Azure, moving the three-way partnership from last November's strategic announcement into a deliverable enterprise AI-agent platform.
What exactly is this product?
It is called Claude in Microsoft Foundry. In plain terms = Anthropic's Claude model runs on Microsoft Azure, with Nvidia's GB300 Blackwell Ultra GPUs supplying the compute underneath.
Each company contributes one piece: Anthropic brings the model, Microsoft brings the cloud, Nvidia brings the chips. Together they form an enterprise AI-agent development platform.
The target customers are Azure-native enterprises. The pitch: build autonomous AI agents — AI programs that make decisions and execute tasks on their own — and vertical AI agents tailored to specific industries.
How far has this partnership actually come?
The three-way collaboration was first announced in November last year, at the strategic-statement stage.
What changed now: the product is generally available — not a slide deck, but something enterprise customers can actually use.
This means → "we plan to work together" has become "the thing is built and live." That is the critical step from narrative to delivery.
What role does Nvidia play in this stack?
Compute layer: Claude runs on Nvidia's GB300 NVL72 systems, connected by Quantum-X800 InfiniBand — an ultra-high-speed chip-to-chip networking fabric — supporting multi-agent systems that coordinate across business domains.
Tooling layer: Nvidia is integrating its own tools into Anthropic's stack, letting enterprises embed AI into workflows via Nvidia-certified agent skills.
This means → Nvidia is not just selling GPUs. It is positioning itself as the "operating system" vendor for enterprise AI — shipping chips, but also shipping the software ecosystem and the standards around them.
How are security and governance handled?
Nvidia released the NVIDIA Secure Agent Workspace reference design — a blueprint for running autonomous agents in controlled environments.
It covers four key control points: identity authentication, network access, credential management, and runtime policy — the security baselines enterprise IT departments care about most.
This reflects a practical reality: the more autonomous the AI agent, the more nervous the enterprise. Without a security framework in place, large customers simply will not deploy.
What does this mean for the market?
The partnership bundles model capability + cloud infrastructure + accelerated compute into a tightly integrated enterprise AI offering.
This means → any competitor trying to match this package needs credible answers on all three layers — model, cloud, and chip — simultaneously. The bar is high.
But the key test has not arrived yet: whether this product converts into measurable enterprise contract growth will only show up in coming quarterly earnings. Put simply = going live is step one; whether it actually makes money is a question for the next few quarters.
Content is for reference only, not financial advice.