Nvidia Brings Superintelligence to Your Desk with DGX Spark and DGX Station

Nvidia took the stage at the GTC keynote to unveil two groundbreaking personal supercomputers — the DGX Spark and DGX Station — purpose-built for developers and data scientists, and designed to bring AI computing power to an entirely new level.

During the GTC keynote, CEO Jensen Huang revealed two innovative PC architectures: the DGX Spark and the DGX Station, both powered by the Grace Blackwell platform. These are not conventional computers — they are genuine AI intelligence centers, engineered to run complex neural networks and optimize local AI models at scale.

The DGX Spark, the more compact of the two, is anything but limited in capability. Equipped with the GB10 Grace Blackwell Superchip, it features a Blackwell GPU and fifth-generation Tensor Cores, capable of delivering up to 1,000 trillion operations per second. The DGX Station, meanwhile, stands as the more formidable of the pair, incorporating the GB300 Grace Blackwell Ultra Desktop Superchip. It boasts 784 GB of coherent memory and supports network speeds reaching 800 Gb/s via the ConnectX-8 SuperNIC.

The DGX systems are designed not only to function as standalone AI laboratories, but also as bridge systems — seamlessly facilitating the transition of models from a local environment to the DGX Cloud or any other cloud AI infrastructure, with minimal code changes.

The DGX ecosystem will be further expanded through partnerships with Asus, Dell, HP, and Lenovo, who will be responsible for developing and commercializing both systems. Reservations for the DGX Spark are already open, while the DGX Station is expected to reach the market later in 2025. Though Nvidia has yet to confirm specific pricing, earlier estimates placed the entry-level DGX Spark at approximately $3,000.

With the arrival of these personal supercomputers, Nvidia is not merely launching new products — it is opening the door to a future in which AI integrates more naturally and efficiently into the daily work of developers and data scientists everywhere.