Jensen Huang took the stage at COMPUTEX 2026 in Taipei to announce two flagship products: the NVIDIA RTX Spark superchip and Cosmos 3, an open foundation model for physical AI. Together, they represent Nvidia’s push to bring AI compute to every desktop and accelerate the development of robots and autonomous systems.
The RTX Spark integrates a Blackwell RTX GPU with 6,144 CUDA cores and a 20-core NVIDIA Grace CPU, connected via NVLink-C2C. It delivers up to 1 petaflop of AI compute and supports up to 128GB of unified memory. Laptops and desktops from ASUS, Dell, HP, Lenovo, Microsoft Surface, and MSI are expected in fall 2026. Cosmos 3, meanwhile, is billed as the world’s first fully open “omnimodel” — built on a Mixture-of-Transformers (MoT) architecture, it can natively understand and generate text, images, video, sound, and actions with physics accuracy. Available in Super (32B/64B) and Nano (8B/16B) variants, it’s aimed at training robots and autonomous vehicles to navigate the real world.
Source
Announcements via NVIDIA, NVIDIA Investor Relations, and coverage from CBS News, PCMag, and Business Insider.
Commentary
Nvidia’s COMPUTEX keynote was classic Jensen: ambitious, wide-ranging, and designed to make you feel like the future is already here. RTX Spark is essentially Nvidia’s play to own the “AI PC” category — 1 petaflop of AI compute in a laptop is genuinely impressive and positions the chip as a serious platform for local model inference, not just a marketing gimmick.
Cosmos 3 is the more technically interesting announcement. An open multimodal model that handles text, vision, audio, and physical actions in a single architecture is exactly what the robotics and autonomous vehicle industries need. The fact that it’s open-source (and available in smaller variants) means the research community can actually build on it. Nvidia isn’t just selling chips anymore — they’re building the entire AI stack from silicon to foundation models, and COMPUTEX 2026 made that ambition unmistakable.