NVIDIA and Microsoft Push Windows PCs as Local AI-Agent Platforms; RTX Spark Laptops Due Oct. 16

NVIDIA and Microsoft used a joint event in San Francisco on Wednesday to make a specific pitch about the future of Windows PCs: not just as AI-capable computers, but as machines designed to run AI agents locally, with Windows providing the security and runtime features and NVIDIA supplying the hardware.

The clearest near-term product news was on NVIDIA’s RTX Spark platform. NVIDIA said laptop preorders opened Wednesday, the first laptops are due Oct. 16 and compact desktop systems are planned for November.

That timing matters because the companies framed the announcement as more than a hardware refresh. In NVIDIA’s recap of the event, the company said it and Microsoft are “co-engineering hardware and software for AI agents to run on Windows PCs,” tying Windows’ agent-focused features directly to RTX Spark systems.

NVIDIA founder and CEO Jensen Huang appeared alongside Microsoft CEO Satya Nadella, while Microsoft Executive Vice President Pavan Davuluri outlined Windows changes aimed at making those agents manageable and secure. According to NVIDIA’s event write-up, Microsoft announced the general availability of Microsoft Execution Containers, or MXC, which NVIDIA described as operating system-level infrastructure that lets agents run safely and persistently in the background under OS control.

Microsoft’s security message was central to the event. “We needed to make the desktop the most secure place for agents to execute,” Nadella said, according to NVIDIA’s Oct. 7 event blog. Davuluri added: “We are building these primitives directly into Windows, so agents can be secured, observed and governed. With MXC, Microsoft Security, and Agent 365, we’re unlocking the full potential of agents on Windows.”

For consumers, RTX Spark was the main focus. NVIDIA said the platform can be configured with up to a 6,144-core Blackwell RTX GPU, up to a 20-core Grace CPU, up to 128 GB of unified memory and up to 1 petaflop of FP4 AI performance. NVIDIA also said the Grace CPU and Blackwell GPU are linked at 600 GB per second, which the company says is intended to support large local models without cloud round trips.

Systems based on RTX Spark are set to come from Microsoft Surface and other manufacturers including Dell, HP, Lenovo and ASUS, according to NVIDIA. The company did not post broad retail pricing in the materials reviewed, and costs appear likely to vary by manufacturer and configuration.

Huang used the appearance to underscore the companies’ long relationship with the PC market. “If you look at the entire journey of our company, Windows was at the core of it. If not for Windows there would be no GeForce,” he said, according to NVIDIA’s event blog.

The event also included a preview of DGX Station for Windows, the enterprise counterpart to the consumer-focused Spark push. NVIDIA said the deskside system will offer up to 748 GB of coherent memory and up to 20 petaFLOPS of FP4 AI compute, aimed at running models of roughly 1 trillion parameters locally on Windows.

None of this was presented for the first time Wednesday. RTX Spark, DGX Station for Windows and Microsoft’s agent-related Windows features had already been introduced or previewed earlier in 2026 around events including Computex, GTC Taipei and Microsoft Build.

What was new in San Francisco was the joint public push from both CEOs, the tighter link between Windows’ agent infrastructure and NVIDIA’s client hardware, and the first concrete consumer rollout dates. For Microsoft, the goal is to make Windows an operating-system platform for AI agents. For NVIDIA, it is to sell PCs built to run more of that work on-device instead of sending it to the cloud. No pricing was disclosed.