NVIDIA Partners With Microsoft on Unified Stack for Agentic AI Deployment, From Windows Devices to Cloud to Local

Overview

At Microsoft Build, NVIDIA and Microsoft announced an expanded partnership to deliver a unified stack for deploying agentic AI across Windows devices, Azure cloud, and local deployments. NVIDIA CEO Jensen Huang joined Microsoft CEO Satya Nadella's keynote via livestream from Taipei. The announcements span new Windows AI hardware, NVIDIA open models on Microsoft Foundry, and a secure runtime for autonomous agents.

Key Takeaways

NVIDIA and Microsoft expanded their partnership to deliver a full stack for agentic AI across Windows devices, Azure cloud, and local deployments.
Jensen Huang joined Satya Nadella's Build keynote via livestream from Taipei.
RTX Spark powers the first Windows PCs purpose-built for personal agents, with 1 petaflop of AI performance and up to 128GB of unified memory.
DGX Station for Windows offers up to 748GB of coherent memory and 20 petaflops of FP4 performance, running models up to 1 trillion parameters.
NVIDIA, Anthropic, and OpenAI models are now available on hosted agents in Foundry Agent Service, and Anthropic's Claude models run natively on NVIDIA GB300 Blackwell Ultra systems on Azure.
NVIDIA Nemotron 3 Ultra, a new open frontier reasoning model, is available this month on Foundry managed compute.

Stats & Key Facts

#1 petaflop of AI performance on RTX Spark
#Up to 128GB of unified memory on RTX Spark
#Up to 748GB of coherent memory on DGX Station for Windows
#20 petaflops of FP4 performance on DGX Station for Windows
#Frontier models up to 1 trillion parameters
#Over 30 years of NVIDIA innovation

NVIDIA Partners With Microsoft on Unified Stack for Agentic AI Deployment, From Windows Devices to Cloud to Local

A unified stack for agentic AI

›Delivering on agentic AI takes fast hardware, secure runtimes, a responsive data layer, and models tuned for long-running reasoning.
›NVIDIA and Microsoft are bringing that full stack across Windows devices, Azure cloud, and local deployments.
›Jensen Huang joined Satya Nadella's keynote via livestream from Taipei.

The announced components include RTX Spark and DGX Station for Windows, GPU-accelerated Microsoft Fabric, NVIDIA open models on Microsoft Foundry, the NVIDIA OpenShell secure runtime in GitHub Copilot, and next-generation NVIDIA-powered AI factories.

RTX Spark and DGX Station for Windows

The two companies are reimagining Windows PCs for AI agents.

›RTX Spark powers the first Windows PCs purpose-built for personal agents, with 1 petaflop of AI performance and up to 128GB of unified memory.
›RTX Spark offers all-day battery life and full AI and graphics performance unplugged, arriving this fall from Microsoft Surface, ASUS, Dell, HP, Lenovo, and MSI.
›DGX Station for Windows is a deskside AI supercomputer powered by the NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip.
›DGX Station offers up to 748GB of coherent memory, 20 petaflops of FP4 performance, and runs frontier models up to 1 trillion parameters, expected in Q4 from ASUS, Dell, GIGABYTE, HP, MSI, and Supermicro.

Both products run NVIDIA OpenShell, a secure-by-design runtime for autonomous agents.

NVIDIA open models on Microsoft Foundry

›NVIDIA, Anthropic, and OpenAI models, plus Hermes special agents, are now on hosted agents in Foundry Agent Service.
›Anthropic's Claude models run natively on NVIDIA GB300 Blackwell Ultra systems on Azure, with customer availability in the weeks ahead.
›NVIDIA Nemotron 3 Ultra, a new open frontier reasoning model for long-running agents, is available this month on Foundry managed compute.
›It is joined by Nemotron 3.5 ASR for speech recognition and Nemotron 3.5 Content Safety.

The article says enterprises can bring agentic systems to life on Azure with built-in identity and governance, composing Nemotron alongside frontier and local models to optimize cost and quality.

A broader open model portfolio

›NVIDIA Cosmos 3 is described as the first fully open omnimodel for physical AI, with vision reasoning, world simulation, and action generation.
›NVIDIA Earth-2 AI weather models are available through Microsoft Planetary Computer Pro and Foundry for forecasting and risk analysis.
›NVIDIA Agent Toolkit and NemoClaw blueprints give developers an open source platform to build production agents on Foundry.
›CUDA-X libraries including cuDF, cuOpt, AI-Q, and NeMo are now accessible to agents as domain-specific skills.

Developer ecosystem

›RTX Spark and DGX Station let developers build, tune, and run agents natively on Windows.
›NVIDIA brings over 30 years of innovation including CUDA, RTX, DLSS, and TensorRT.
›The OpenShell runtime is also featured in GitHub Copilot.

Frequently Asked Questions

What is the focus of the NVIDIA and Microsoft partnership expansion?

Delivering a unified full stack for deploying agentic AI across Windows devices, Azure cloud, and local deployments, announced at Microsoft Build.

What are RTX Spark and DGX Station for Windows?

RTX Spark powers Windows PCs for personal agents with 1 petaflop of AI performance and up to 128GB of unified memory, while DGX Station is a deskside AI supercomputer with up to 748GB of coherent memory, 20 petaflops of FP4, and support for models up to 1 trillion parameters.

Where do Anthropic's Claude models run in this announcement?

Anthropic's Claude models run natively on NVIDIA GB300 Blackwell Ultra systems on Azure, with customer availability in the weeks ahead.

What is NVIDIA Nemotron 3 Ultra?

It is a new open frontier reasoning model for long-running agents across coding, research, and enterprise workflows, available this month on Foundry managed compute.

What runtime secures the new Windows agent hardware?

Both RTX Spark and DGX Station for Windows run NVIDIA OpenShell, a secure-by-design runtime for autonomous agents.

The Build announcements position NVIDIA and Microsoft to deliver agentic AI across Windows hardware, Azure, and local deployments with new chips, open models, and a secure agent runtime.