On Wednesday, Oct. 7, Microsoft opened preorders for new Windows laptops built around a chip NVIDIA calls RTX Spark. Machines from ASUS, Dell, HP, Lenovo, MSI and Microsoft's own Surface line start shipping Oct. 16. The aim is to run AI models on the laptop itself instead of on a company's servers over the internet, which NVIDIA says means no metered bills and no data sent to the cloud.
The companies say what sets these machines apart is memory. Large AI models need more memory than most laptops have, and in a typical laptop the graphics chip, which does most of the AI work, has its own separate memory. RTX Spark lets the graphics chip and the main processor share one pool of up to 128 gigabytes, which Dell says lets its laptop handle AI models that would not fit on a traditional one. Microsoft says that is enough to run models with more than 120 billion parameters, the numbers a model tunes as it learns; a bigger count means a bigger model, not necessarily a better one. Microsoft's fine print says the graphics chip can't use all 128 gigabytes.
Microsoft's Surface Laptop Ultra starts at $2,599 and arrives Oct. 16, and Microsoft is offering up to $1,000 back for trading in an eligible MacBook Pro, online in the U.S. and Canada through Nov. 23. A small desktop, the Surface RTX Spark Dev Box, costs $5,999, is sold only on Microsoft's U.S. store and ships in November. NVIDIA says compact RTX Spark desktops go on sale in November, and a more powerful deskside machine, DGX Station for Windows, is due later this year.
Microsoft also released a tool for Windows 11 that lets organizations set which files and networks an AI agent can reach, with those limits enforced while it runs. An AI agent is AI that can carry out tasks on its own, such as using apps and websites, instead of only answering questions. OpenAI's Codex and GitHub Copilot, a Microsoft coding assistant, already work with the tool, and Anthropic's Claude Code will. Microsoft says its separate Copilot assistant will be able to use files on your PC, take actions for you and run AI models locally, with your permission, on some Windows PCs in the coming months.
Microsoft says the new machines started replying up to 2.1 times sooner than a 16-inch MacBook Pro with Apple's M5 Pro chip and the same 64 gigabytes of memory, and made AI images 4.3 times faster and AI video 6.2 times faster. Microsoft and NVIDIA ran or paid for those tests in September on early, preproduction machines. Microsoft has not said how much memory the $2,599 model comes with, and it is not yet clear how the laptops perform on everyday work, what the other makers will charge, or how much of the AI people use today will move off the cloud.