Nvidia Launches DGX Spark: The World’s Smallest AI Supercomputer for Developers

Nvidia has introduced the DGX Spark, calling it the world’s smallest AI supercomputer. The compact device is powered by the company’s GB10 Grace Blackwell superchip and includes ConnectX-7 networking and Nvidia’s full AI software suite. It’s aimed at startups, research teams, and individual developers looking to access cutting-edge AI computing without paying for large-scale data center resources.The DGX Spark can be configured with up to 128GB of memory and supports complex AI models with hundreds of billions of parameters. Nvidia says two Spark units can be linked to handle models with up to 405 billion parameters, giving developers a powerful local alternative to cloud-based AI training.

The new system continues Nvidia’s tradition of democratizing access to AI computing. “In 2016, we built DGX-1 to give AI researchers their own supercomputer. I hand-delivered the first system to Elon Musk at OpenAI — and from it came ChatGPT,” said Nvidia CEO Jensen Huang. “With DGX Spark, we’re once again putting advanced AI tools directly in the hands of developers.”

Like its predecessor, Huang personally delivered one of the first DGX Spark systems to Musk. The Spark complements Nvidia’s larger DGX Station, which uses the GB300 Grace Blackwell Ultra chip for even greater performance.

Unlike standard PCs, the DGX Spark runs Nvidia’s Linux-based DGX OS and AI software stack, purpose-built for developing and deploying AI models rather than gaming or general computing.With DGX Spark, Nvidia aims to make high-performance AI computing more accessible, empowering smaller companies and individual innovators to build their own models locally. The launch strengthens Nvidia’s dominance in the AI chip market, as the company continues to secure major deals — including a $100 billion GPU investment from OpenAI and multi-billion-dollar partnerships with CoreWeave, Amazon, Google, and other tech leaders.