NVIDIA's RTX Spark superchip and N1X SoC bring 1-petaflop AI to Windows PCs with Microsoft

NVIDIA today announced the RTX Spark superchip and N1X system-on-chip (SoC), a major push into the AI PC market developed in collaboration with Microsoft. The RTX Spark is a 1-petaflop superchip that integrates a Blackwell RTX GPU with a 20-core Grace CPU, designed to deliver desktop-class AI performance directly on Windows laptops and desktops. This architecture enables local AI agents and advanced generative AI workloads without relying on cloud connectivity, addressing latency and privacy concerns for enterprise users. The N1X SoC, meanwhile, targets lighter AI tasks and efficient power management, complementing the RTX Spark in a hybrid computing model.

Initial devices featuring these chips are expected later this year from partners including Dell and Microsoft’s own Surface line. The collaboration aims to make AI assistants, real-time content generation, and complex data analysis seamless on personal computers. By offloading AI processing to dedicated hardware, NVIDIA and Microsoft are positioning the RTX Spark as a foundational component for the next generation of Windows PCs, where AI capabilities become as integral as traditional CPU and GPU functions. For developers and power users, this means faster local inference for models like Llama 3 and Stable Diffusion, and the ability to run multi-agent frameworks directly on-device.