NVIDIA has just introduced two groundbreaking AI supercomputers designed for desktop use: the DGX Spark and the DGX Station. Announced during the company’s GTC event, CEO Jensen Huang unveiled these cutting-edge machines aimed at accelerating AI research and development.
The DGX Spark, initially teased earlier this year under the name Project Digits at CES, is now officially recognized as the world’s smallest AI supercomputer. Priced at $3,000, this compact powerhouse is built around the GB10 Grace Blackwell Superchip, engineered to deliver an astounding 1,000 trillion AI operations per second. Remarkably, NVIDIA has managed to pack this performance into a form factor comparable to the Mac mini. The device is designed to efficiently fine-tune next-generation AI models, including the recently announced GR00T N1 robotic system. Preorders for the DGX Spark are now live.
For users requiring even greater computational capabilities, NVIDIA is introducing the DGX Station. This AI workstation is powered by the GB300 Grace Blackwell Ultra Desktop Superchip, which boasts an astonishing 20 petaflops of AI processing performance and comes equipped with 784GB of unified system memory. While NVIDIA has not yet disclosed pricing details, the DGX Station is expected to launch later this year. Leading hardware manufacturers—including ASUS, BOXX, Dell, HP, Lambda, and Supermicro—are preparing their own versions of this AI workstation to meet the growing demands of researchers and data scientists worldwide.
With these innovations, NVIDIA continues to push the boundaries of AI computing, delivering unprecedented power in increasingly compact and accessible formats.
