NVIDIA has officially begun shipping DGX Spark™, the world’s smallest AI supercomputer, designed to bring petascale AI performance directly to desktops and labs worldwide. Built on the NVIDIA Grace Blackwell architecture, DGX Spark integrates GPUs, CPUs, high-speed networking, CUDA libraries, and the full NVIDIA AI software stack, delivering 1 petaflop of performance and 128GB of unified CPU-GPU memory. This compact system enables developers to run inference on models with up to 200 billion parameters and fine-tune models up to 70 billion parameters locally, supporting agentic and physical AI development without reliance on cloud or large-scale data centers.
Also Read: Implicit Unveils KnowledgeOS with Industry-Agnostic AI and Permanent Freemium Experience
“In 2016, we built DGX-1 to give AI researchers their own supercomputer… With DGX Spark, we return to that mission — placing an AI computer in the hands of every developer to ignite the next wave of breakthroughs,” said Jensen Huang, NVIDIA founder and CEO, who personally delivered one of the first units to Elon Musk at SpaceX. Early adopters including Anaconda, Hugging Face, Meta, Microsoft, and research labs like NYU Global Frontier Lab are already leveraging DGX Spark to accelerate AI model prototyping, advanced algorithm experimentation, and secure, high-performance AI workflows across industries.