NVIDIA Unveils Next-Generation Desktop Supercomputers
At the recent GTC conference, NVIDIA made headlines with its announcement of revolutionary desktop supercomputers, DGX Spark and DGX Station. CEO Jensen Huang first gave attendees a glimpse of what would eventually be dubbed DGX Spark during CES earlier this year, where he introduced Project Digits. This newly branded $3,000 device is being marketed as the smallest AI supercomputer in existence.
The DGX Spark is equipped with an advanced GB10 Grace Blackwell Superchip that has been strategically miniaturized to fit within a form factor comparable to the previous Mac mini models. According to NVIDIA, this powerful chip can execute up to 1,000 trillion operations every second for AI computations. This capability makes it particularly suitable for optimizing cutting-edge AI reasoning models—like the GR00T N1 robotic system highlighted by Huang in his concluding remarks at the GTC keynote. Enthusiastic users can now preorder their DGX Sparks starting today.
NVIDIA
Enhanced Processing Power with DGX Station
For those researchers and data scientists requiring even greater computational capabilities, NVIDIA also introduced the more powerful DGX Station featuring the highly capable GB300 Grace Blackwell Ultra Desktop Superchip. With an impressive performance capacity of 20 petaflops and a robust unified system memory of 784GB, this model promises top-tier efficiency for complex AI workloads.
Although NVIDIA hasn’t disclosed pricing details for the DGX Station yet, it is expected to launch later in the year with notable tech companies such as ASUS, BOXX Technologies, Dell Technologies, HP Inc., Lambda Labs, and Supermicro set to produce their customized iterations of this innovative computing solution.
