Microsoft is preparing Windows for Nvidia’s Arm-based RTX Spark laptops with agent setup, memory management, gaming compatibility, and emulator tuning.
Microsoft has made changes to Windows for Nvidia’s RTX Spark laptops, which pair an Arm-based processor with a Blackwell GPU and unified memory. The work spans local AI agents, app compatibility, gaming and system memory, according to Microsoft representatives and The Verge’s hands-on report published October 15, 2026.
A new Windows “Get Started” experience is intended to help owners find and install an AI agent in two clicks. Hermes, Perplexity and OpenClaw are the first agents officially supported at launch, and Microsoft says they can be installed and updated through the Windows Store. The reporter found Hermes setup largely straightforward but needed to uninstall and reinstall OpenClaw after an issue with its local gateway.
For applications without native Arm versions, the laptops use Microsoft’s Prism emulator to translate x64 software. Microsoft says it has tuned the emulator for RTX Spark’s chip design and ships compatibility improvements monthly. The platform also has a Windows app slider for adjusting how much unified memory is reserved for the GPU. Microsoft says its memory work aims to balance model capacity with system responsiveness, battery life and other applications running alongside demanding tasks.
Gaming is another focus. The report says Fortnite, Marvel Rivals, Arc Raiders, Counter-Strike 2 and Dota 2 are expected to work at launch, with additional anti-cheat support in development. Activision’s Ricochet is planned for 2027 support, EA’s Javelin for early 2027, and PUBG for the first half of 2027; Riot has committed to bringing League of Legends and Valorant at an unspecified future time. In the reporter’s testing, many games ran through emulation, with performance varying by title.
Some Windows features will arrive later: Copilot Plus PC experiences, including Recall, are not available on RTX Spark laptops at launch. Microsoft says it expects to bring those experiences to Windows Insiders later in 2026. The article also notes that large local models responded more slowly than cloud models in the reporter’s experience, making the laptops’ local AI capabilities a different tradeoff rather than a direct replacement for cloud inference.