What Is NVIDIA DGX Spark: The World’s Smallest AI Supercomputer Delivered To Elon Musk

NVIDIA has started rolling out the DGX Spark, described as the world’s smallest AI supercomputer. The company’s founder and CEO Jensen Huang delivered the DGX Spark to Elon Musk at SpaceX on Oct. 13, marking a major step in bringing advanced AI computing out of the data centre and into the hands of individuals. 

The DGX Spark combines NVIDIA GPUs, CPUs, networking, CUDA libraries and AI software, all packed into a compact desktop device.

This week, NVIDIA and its partners, including Acer, ASUS, Dell Technologies, GIGABYTE, HPI, Lenovo and MSI, will start shipping DGX Spark systems.

Despite its small size, DGX Spark delivers immense computing muscle. With 128GB of unified CPU-GPU memory, it can run AI models with up to 200 billion parameters locally, providing around a petaflop of AI performance. It’s powered by the NVIDIA GB10 Grace Blackwell Superchip, designed to offer up to one petaflop of performance at FP4 precision.

The system also includes NVIDIA ConnectX networking for clustering, NVLink-C2C offering up to five times the bandwidth of PCIe, and NVMe storage for rapid data access.

The DGX Spark is roughly the size of a folded sheet of paper and about as thick as a hardcover book. Weighing just 1.2 kg, it is described by NVIDIA as a “rocket engine for AI.” Compact yet capable, it’s designed to empower creators and researchers to work directly from their desks without needing to depend on large data-centre infrastructure or cloud instances.

DGX Spark isn’t just hardware, it comes with the full NVIDIA AI software stack, including frameworks, libraries, pretrained models and NVIDIA NIM microservices. These tools support many types of AI workflows. You can fine-tune image models like FLUX.1, build vision search and summarisation agents with NVIDIA Cosmos, or create optimised chatbots using Qwen3.

Nine years after launching the first DGX system, NVIDIA’s vision for AI computing has moved beyond centralised facilities. The DGX Spark places supercomputing power literally within arm’s reach of developers, researchers and creators, enabling faster experimentation, iteration and innovation in AI.