Nvidia unveiled RTX Spark, its first Arm-based PC superchip, at Computex 2026 on May 31, per Tom's Hardware. The chip combines a 20-core Arm CPU with a Blackwell GPU rated at 6,144 CUDA cores, 128GB of unified LPDDR5X memory, and 300 GB/s of bandwidth. First laptops ship this fall from Microsoft, Dell, HP, Lenovo, ASUS, and MSI, with Acer and GIGABYTE following. Microsoft's launch device is a new Surface Ultra.
The number that frames the move is the $200 billion annual laptop CPU market TechCrunch reports Nvidia is targeting. That's roughly the combined Intel client business, AMD Ryzen, and the Qualcomm Snapdragon X book that has been the Arm-on-Windows entrant for the last two years. Nvidia is contesting all three at once, and the OEM coalition tells you how the major Windows brands are placing the bet. Microsoft launching a Surface Ultra on Spark is the loudest single signal in that list.
The 128GB unified-memory architecture is the technical play, and it lands as an unmistakable shot at Apple Silicon. Apple's M-series tops out at 128GB unified memory on the Mac Studio, which has been the configuration that lets local frontier-model inference run without a discrete GPU. Spark puts the same architecture in a 14mm Windows chassis with 6,144 CUDA cores attached. For the workloads buyers were quietly considering Macs for (local model serving, agentic workflows that hammer memory bandwidth), Windows now has a credible answer that runs the existing CUDA software stack.
Bottom Line
Intel and AMD share a laptop CPU socket Nvidia just attacked, and Microsoft brought the launch hardware. The fall 2026 Spark laptops will test how fast PC OEMs and procurement actually switch architectures when the silicon ships with CUDA on board.