Nvidia has introduced a groundbreaking lineup of personal AI supercomputers — the DGX Spark and DGX Station — designed to meet the growing demand for local AI model development. Announced during Nvidia’s GTX keynote, CEO Jensen Huang described these systems as a new class of AI PCs built for developers, researchers, and data scientists.
Bridging AI Innovation
The DGX systems are powered by Nvidia’s Grace Blackwell platform. Offering substantial computational power, they enable users to train, fine-tune, and run large AI models locally. These machines can also act as “bridge systems,” allowing developers to seamlessly transition their AI workloads from desktop environments to Nvidia’s DGX Cloud or other AI cloud platforms with minimal code adjustments.
“AI has transformed every layer of the computing stack. It stands to reason a new class of computers would emerge — designed for AI-native developers and to run AI-native applications,” Huang stated.
System Highlights
- DGX Spark: Equipped with the GB10 Grace Blackwell Superchip and fifth-generation Tensor Cores, DGX Spark delivers up to 1,000 trillion operations per second (TOPS) for AI workloads.
- DGX Station: Designed for more demanding tasks, the DGX Station features the GB300 Grace Blackwell Ultra Desktop Superchip with 784GB of coherent memory. It also includes a ConnectX-8 SuperNIC supporting network speeds up to 800Gb/s.
Manufacturer Partnerships
Asus, Dell, HP, and Lenovo will produce and distribute both DGX Spark and DGX Station. Additional manufacturers, including BOXX, Lambda, and Supermicro, are set to release DGX Station units later in 2025. Reservations for DGX Spark are available now, while DGX Station will launch by the end of the year.
Pricing details remain unconfirmed, but Nvidia previously suggested that entry-level DGX Spark configurations may start around $3,000.
