Surface Laptop Ultra Puts Nvidia RTX Spark and a Petaflop in a Windows Laptop
Microsoft announced the Surface Laptop Ultra, a Windows laptop powered by Nvidia's RTX Spark Arm-based chip, at a Windows and Surface-focused keynote in San Francisco on October 7. The base configuration starts at $2,599 and will be released on October 16. The Verge
Microsoft CEO and Nvidia CEO Jensen Huang were scheduled to appear together to unveil the system at the event. Reuters The scheduling put both companies on stage for a single client device, with compute supplied by Nvidia and the OS and industrial design supplied by Microsoft.
That $2,599 entry system pairs an 8-core CPU with 24GB of RAM and 512GB of storage. The Verge Microsoft lists up to one petaflop of AI compute for the Laptop Ultra and describes the Nvidia silicon as combining an ultra-efficient CPU with a powerful RTX GPU. Microsoft The company also lists configurations with up to 128GB of unified memory. Microsoft
Pre-event reporting pointed to higher-tier RTX Spark options in the same chassis, including an 18-core part with 5,120 GPU cores and a 20-core part with 6,144 GPU cores. Windows Central Nvidia has previously said RTX Spark laptops are engineered to be as slim as 14 millimeters and as light as three pounds, in 14- to 16-inch sizes. Nvidia
The port is gone. The Laptop Ultra drops the proprietary Surface Connect magnetic charging port in favor of built-in magnetic USB-C charging. The Verge For fleet managers and accessory vendors, that removes a long-standing Surface-specific connector from the power path and aligns charging with the USB-C ecosystem.
Microsoft also announced the Surface RTX Spark Dev Box, a mini PC for developers. It is available for preorder at $5,999 and will ship in November. The Verge Microsoft lists up to one petaflop of AI compute for the Dev Box, with 128GB of unified memory. Microsoft Nvidia separately offers RTX Spark in a small, ultra-efficient desktop form factor. Nvidia
On software, Microsoft announced Windows Hybrid Intelligence features that let Copilot use files on a Windows PC to take actions, alongside a new Windows search experience that lets users invoke quick actions from the search bar. The Verge Nvidia describes RTX Spark as powering Windows PCs purpose-built for personal agents. Nvidia
Looking at what this means for working technologists, the interesting coupling is large unified memory with local, high-throughput GPU compute in a portable power envelope. Memory is unified. That matters for local inference on large-parameter models, agentic toolchains that keep substantial context resident, and build-test loops where moving weights and datasets between discrete pools adds latency. A petaflop-class laptop and a matching deskside box give developers a way to prototype agent behavior locally, then target the same Windows file, search, and Copilot surfaces those agents will use in deployment. In this author's view, the long-term optimism here is straightforward: if the Arm driver stack, CUDA toolchain, and Windows on Arm compatibility hold up under daily use, this lowers the friction for building and running capable assistants without a round trip to the cloud for every token.
