Tech Desk 
Mar 20, 2025
NVIDIA has launched DGX Spark and DGX Station, two powerful AI supercomputers designed for developers, researchers, and data scientists. Based on the Grace Blackwell architecture, these desktop systems deliver data centre-class AI performance to individual workstations.
NVIDIA, with these new systems, aims to simplify AI development by providing high-performance desktop solutions to handle complex AI tasks. NVIDIA’s CEO Jensen Huang stated: “AI has transformed computing. These new DGX personal AI computers enable AI to span from cloud services to desktops and edge applications.”
NVIDIA DGX Spark
DGX Spark, the smallest AI supercomputer in the world, is designed for AI model fine-tuning, robotics development, and research. It has the GB10 Grace Blackwell Superchip with 1,000 trillion operations per second via fifth-generation Tensor Cores.
This enables users to train and deploy locally or transfer them effortlessly to NVIDIA DGX Cloud for wider-scale computing.
NVIDIA DGX Station
The DGX Station is an even more capable AI development workstation with the GB300 Grace Blackwell Ultra Superchip and 784GB of memory. It is designed for big AI training and inference, with high-speed NVLink connectivity between the CPU and GPU.
Moreover, it has the ConnectX-8 SuperNIC, which can support up to 800Gb/s networking, making it well-suited for managing huge AI workloads.
NVIDIA DGX Spark and DGX Station availability
Reservations for the DGX Spark are currently open, with the DGX Station being made available later this year by leading manufacturers like ASUS, Dell, HP, and Lenovo.
