Nvidia just did something it has never done before: build a complete chip for personal computers. The company that became the most valuable in the world by selling AI hardware to data centers is now coming for your laptop.
CEO Jensen Huang unveiled the RTX Spark on June 1 at the GTC event held during Computex in Taipei. The Arm-based system-on-chip combines a custom Grace CPU with a Blackwell-architecture GPU, and it’s designed specifically for Windows laptops and compact desktops. Units are expected to ship in fall 2026.
What’s actually inside the RTX Spark
The RTX Spark packs a 20-core Arm CPU, a Blackwell GPU with 6,144 CUDA cores, and 128GB of unified LPDDR5X memory. Nvidia claims the chip delivers approximately 1 petaflop of AI compute. In English: that’s enough horsepower to run large AI models with up to 120 billion parameters and a context window of one million tokens, all without pinging a cloud server.
Nvidia developed the chip in partnership with Microsoft, with a focus on enabling AI agents to run securely and directly on the device. The pitch is straightforward: keep your data local, skip the latency of cloud round-trips, and still get performance that rivals what you’d find in a server rack.
