Overview
- Nvidia revealed RTX Spark on Monday, June 1, as an integrated Arm-based SoC that pairs a 20‑core CPU with a Blackwell GPU of 6,144 CUDA cores, claiming up to 1 petaflop of AI compute and support for as much as 128 GB of unified memory.
- The chip was developed with Microsoft and MediaTek, will be built by TSMC on its 3N process, and Nvidia named multiple OEMs including Microsoft Surface, Dell, HP, Lenovo, ASUS and MSI that plan to ship RTX Spark systems this fall.
- Nvidia and Microsoft say the platform is meant to run local AI agents and larger on-device models securely using new Windows security primitives and Nvidia OpenShell, with company claims that the hardware can host models of roughly 120 billion parameters and a million‑token context.
- Markets reacted immediately with Nvidia shares rising and some rival chipmakers’ shares falling, while analysts flagged open questions about final pricing, legacy Windows-on-Arm app compatibility, and whether supply and software support can scale for fall deliveries.
- The next milestones are device pricing, independent performance and compatibility tests, and real-world user reporting once laptops and compact desktops ship, with early models likely to target premium creators, developers and power users because of expected cost and thermal tradeoffs.