On June 1, 2026, NVIDIA and Microsoft announced NVIDIA RTX Spark, a superchip designed to power Windows PCs for personal AI agents. The June 1, 2026 date is stated on the NVIDIA investor press release. NVIDIA CEO Jensen Huang framed it as bringing CUDA, RTX, and the company’s AI platform into a single superchip (quote traces to the primary).
The press release lists RTX Spark as delivering 1 petaflop of AI compute, a Blackwell RTX GPU with 6,144 CUDA cores, a 20-core Grace CPU, and up to 128GB of unified memory, in laptop designs as thin as 14mm and as light as 3 pounds (all specs trace to the primary). NVIDIA states the chip lets users run 120-billion-parameter LLMs locally, edit 12K video, and play AAA games at 1440p with 100-plus fps (claims trace to the primary). Devices are slated for fall 2026 from manufacturers including ASUS, Dell, HP, Lenovo, Microsoft Surface, and MSI, including a Surface Laptop Ultra powered by RTX Spark (availability traces to the primary).
For business and technical leaders, RTX Spark signals a push to move frontier-scale model inference onto local hardware. Running a 120B-parameter model on a laptop reshapes assumptions about cost, latency, and data privacy for AI workloads that would otherwise depend on cloud APIs, and it positions the personal computer as a first-class place to run AI agents.