At a joint event in San Francisco, Nvidia CEO Jensen Huang and Microsoft CEO Satya Nadella announced a deep co-engineering effort to bring AI agents directly to Windows PCs. The centerpiece of this push is the new RTX Spark platform, which integrates Nvidia’s Blackwell RTX GPU with Grace CPU architecture into consumer laptops and compact desktops. Preorders for RTX Spark laptops open today, with availability set for October 16, while compact desktops are slated for November.

Local Agents, Not Cloud Dependencies

Huang framed the announcement as a return to Nvidia’s roots, stating that 'if not for Windows there would be no GeForce.' He emphasized that in the agentic era, the personal computer becomes a personal assistant because the agent runs locally on the user's machine. This shift reduces reliance on cloud APIs, allowing models like Qwen 3.8 Flash Nextβ€”a 125B parameter modelβ€”to run unmetered and without sending user data to external servers.

Microsoft Execution Containers Secure the OS

Microsoft EVP Pavan Davuluri detailed the software side of the equation, announcing the general availability of Microsoft Execution Containers (MXC). This OS-level infrastructure allows agents to run safely and persistently in the background under strict operating system control. Davuluri explained that MXC, combined with Microsoft Security and Agent 365, ensures agents can be 'secured, observed and governed,' effectively turning the desktop into a sandboxed environment for autonomous code.

Hardware Specs: Petaflops on the Desktop

The hardware delivering this vision is robust. RTX Spark systems feature up to 128GB of unified memory and 1 petaflop of FP4 AI compute. The Surface Laptop Ultra, built around this chip, allows developers to run models that previously required traditional workstations. For enterprise users, Nvidia previewed the DGX Station for Windows, a deskside supercomputer running the GB300 Grace Blackwell Ultra Desktop Superchip. It offers 748GB of coherent memory and up to 20 petaFLOPS, enabling trillion-parameter model inference locally without needing Linux.

Key Takeaways

  • RTX Spark laptops from Acer, ASUS, Dell, HP, Lenovo, and others open for preorders now, shipping October 16.
  • Microsoft Execution Containers (MXC) are now generally available, providing OS-level security for background AI agents.
  • DGX Station for Windows brings GB300-class compute to the Windows ecosystem, bridging the gap between Linux AI dev and Windows productivity tools.

The Bottom Line

Nvidia and Microsoft are finally aligning hardware and OS architecture to make local AI agents a first-class citizen on Windows. By locking down execution via MXC and providing petaflop-scale compute in laptops, they are betting that privacy and persistence will beat cloud latency.