NVIDIA, Microsoft Kick Off a New Beginning for Windows PCs With RTX Spark and AI Agents
NVIDIA Blog Gerardo Delgado ● Covered by 3 sources
NVIDIA and Microsoft are putting AI agents directly on Windows PCs. New RTX Spark and DGX Station gear aims to keep those agents local, not in the cloud.
Based on reporting by NVIDIA Blog, Gerardo Delgado — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
NVIDIA and Microsoft used a San Francisco event on Wednesday to sketch a pretty specific future for Windows: AI agents living on the PC, not just talking to it from afar. Jensen Huang leaned hard into the company history, saying NVIDIA grew up with Windows and that without it there would be no GeForce. Satya Nadella framed the same moment as a long build rather than a sudden pivot. Both sounded convinced the desktop is about to become the main stage for agent software.
The software piece matters as much as the hardware. Microsoft announced general availability of Microsoft Execution Containers, or MXC, a Windows layer meant to let agents run safely and persistently in the background under operating system control. Pavan Davuluri said Microsoft is building the needed primitives directly into Windows so agents can be secured, observed and governed. That’s the pitch: make the desktop the safest place for agents to execute, then let the rest of the stack follow.
On the hardware side, NVIDIA’s RTX Spark is the company’s push to bring its full AI stack into Windows laptops and compact desktops. Laptop preorders are open now, with availability set for October 16, and compact desktops are due in November. Surface Laptop Ultra is built around RTX Spark, and Davuluri said it can ship with up to 128 gigabytes of unified memory and up to a petaflop of AI compute, enough for models that won’t fit on a traditional machine.
The system specs are hefty: a Blackwell RTX GPU with up to 6,144 cores, an up to 20-core Grace CPU, and 600 GB/s of connectivity between them. NVIDIA says RTX Spark can run models like Qwen 3.8 Flash Next locally, without sending data to the cloud. It also keeps the full CUDA stack intact, which matters for developers who want to move tools and workflows between devices without rewriting everything.
And then there’s the bigger machine. NVIDIA also previewed DGX Station for Windows, calling it the first deskside AI supercomputer to bring GB300 Grace Blackwell-class infrastructure into Windows. The company says it delivers 748GB of coherent memory and up to 20 petaFLOPS of FP4 AI compute, enough to run models up to a trillion parameters locally. For enterprises that have lived in both Windows and Linux for years, that’s the blunt promise here: stop bouncing between environments and keep the heavy AI work on the same machine as everything else.
My take — AI-written commentary, not fact-checked reporting
This is the kind of announcement that quietly says the cloud was getting too comfortable. If Microsoft can make Windows the place where agents run, and NVIDIA can make local hardware strong enough to matter, the old “everything goes to a remote cluster” story starts looking lazy. Also, Windows finally gets to be more than the office machine everybody tolerated.
Read more about this at: NVIDIA Blog