Summary
NVIDIA and Microsoft announced Windows infrastructure for background AI agents, alongside RTX Spark laptops and compact desktops for running AI locally. NVIDIA also previewed a Windows version of its DGX Station system for larger AI workloads.
At a Microsoft event in San Francisco on Oct. 7, 2026, NVIDIA and Microsoft outlined hardware and Windows software intended to support AI agents and local AI workloads. Microsoft announced the general availability of Microsoft Execution Containers (MXC), while NVIDIA opened preorders for RTX Spark laptops and previewed NVIDIA DGX Station for Windows.
Windows adds agent controls as RTX Spark arrives
Microsoft describes MXC as operating-system-level infrastructure for agents to run persistently in the background under Windows control. The company says the containers are designed to let agents be secured, observed and governed, alongside Microsoft Security and Agent 365. The announcement positions those controls as part of making Windows a platform for agents that operate on a user’s computer.
NVIDIA’s RTX Spark is a hardware platform for Windows laptops and compact desktops aimed at running AI models locally. Its announced specifications include a Blackwell RTX GPU with up to 6,144 cores and an NVIDIA Grace CPU with up to 20 cores, connected at 600 GB/s. NVIDIA specifies one petaflop of FP4 AI performance and up to 128GB of unified memory. FP4 refers to a four-bit number format used in AI computing; unified memory provides a common memory pool for the system’s processors.
Preorders for RTX Spark laptops opened on Oct. 7, with availability scheduled for Oct. 16. Compact desktop systems are expected to go on sale in November. Systems are planned from Acer, ASUS, Dell, HP, Lenovo, Microsoft, MSI and Gigabyte. Microsoft’s Surface Laptop Ultra is among the announced designs; Microsoft executive Pavan Davuluri said it will offer up to 128GB of unified memory and up to a petaflop of AI compute.
NVIDIA says RTX Spark can run models locally, including Qwen 3.8 Flash Next, and that its systems use the NVIDIA CUDA platform. Local execution can keep model workloads on the PC rather than sending their data to a cloud service, depending on the application and how it is configured.
DGX Station targets larger workloads on Windows
NVIDIA also previewed DGX Station for Windows, a deskside system based on the GB300 Grace Blackwell Ultra Desktop Superchip. NVIDIA specifies 748GB of coherent memory and up to 20 petaFLOPS of FP4 AI compute, and says the system can run models at up to trillion-parameter scale locally.
The Windows version is intended to bring heavier model development and inference into the same environment as Windows applications and workflows. NVIDIA says teams will be able to build and run agents that connect to those applications and infrastructure, while Linux AI toolchains remain accessible through Windows Subsystem for Linux. DGX Station previously ran on Linux, according to NVIDIA. The company invited prospective users to sign up for availability updates; it did not give a release date in the announcement.