NVIDIA expands local AI hardware with DGX Station and DGX Spark

NVIDIA is expanding its hardware offerings for local AI development with the DGX Station GB300 and the DGX Spark.

The DGX Station GB300 is designed for office environments, offering a relatively quiet operation with a 1600W power supply. It is positioned as an alternative to multi-GPU servers, such as those utilizing 4x or 8x RTX PRO 6000 Blackwell GPUs. While the DGX Station provides higher memory bandwidth and lower power consumption suitable for standard office sockets, high-density RTX PRO 6000 servers offer higher precision and greater aggregate GPU memory bandwidth for specific workloads.

For more compact needs, the NVIDIA DGX Spark, formerly known as Project DIGITS, provides a desktop-sized AI platform. Featuring a Grace Blackwell superchip and 128GB of coherent unified memory, the device is capable of running models with up to 200 billion parameters locally. The DGX Spark aims to bring data-center-class AI capabilities to a personal development setup, though it faces competition from high-end consumer hardware like the RTX 5090 and Mac Studio.