Nvidia unveils RTX Spark AI superchip for Windows PCs

Nvidia announced RTX Spark, a new AI‑focused “superchip” that combines a Blackwell‑generation RTX GPU with a 20‑core Grace CPU, linked by NVLink‑C2C. The chip is designed for local AI inference on Windows computers, delivering up to 1 petaflop of AI performance and up to 128 GB of unified memory.

The company said the first device to ship with RTX Spark will be Microsoft’s Surface Laptop Ultra, which will feature the chip, a 15‑inch mini‑LED display and an upgraded cooling system. Nvidia also confirmed that the platform will appear later this year in select models from ASUS (ProArt P14/P16), Dell, HP, Lenovo and MSI.

RTX Spark is part of a broader shift from cloud‑only AI training toward edge‑side inference, with Nvidia promoting an ecosystem that includes CPU, GPU, memory, networking and power‑delivery components. Nvidia, Microsoft and MediaTek are collaborating on the system software, including the NVIDIA OpenShell runtime and new Windows security controls. Analysts see the move as a potential catalyst for companies that supply the supporting infrastructure, while noting that the success of the platform will depend on the ability to manage new bottlenecks in memory bandwidth, networking and power consumption.