NVIDIA has announced the global rollout of DGX Spark, the world’s smallest AI supercomputer, designed to bring powerful AI development capabilities directly to developers’ desktops. Built on the NVIDIA Grace Blackwell architecture, DGX Spark delivers up to one petaflop of AI performance and 128GB of unified CPU-GPU memory in a compact desktop form factor, enabling developers to run and fine-tune large AI models locally. The system integrates NVIDIA GPUs, CPUs, high-speed networking via NVIDIA ConnectX-7 200 Gb/s, NVLink-C2C technology for five times the bandwidth of PCIe Gen 5, and the full NVIDIA AI software stack. This combination allows developers to perform inference on models with up to 200 billion parameters and fine-tune models of up to 70 billion parameters without relying on the cloud. Jensen Huang, NVIDIA’s founder and CEO, personally delivered the first DGX Spark unit to Elon Musk at SpaceX’s Starbase facility in Texas. The gesture symbolized a full-circle moment, recalling Huang’s 2016 delivery of the first DGX-1 supercomputer to Musk’s then-small startup OpenAI — a system that helped lay the foundation for the AI revolution, including the development of ChatGPT. DGX Spark is now available for order on NVIDIA.com starting October 15, with systems offered by major partners including Acer, ASUS, Dell Technologies, GIGABYTE, HP, Lenovo, MSI, and Micro Center in the U.S. It will also be available through NVIDIA’s global channel partners. The preinstalled NVIDIA AI software stack gives developers immediate access to tools like CUDA libraries, models, and NVIDIA NIM microservices. This enables local workflows such as customizing FLUX.1 models from Black Forest Labs for image generation, building vision agents using the NVIDIA Cosmos Reason model, or creating optimized AI chatbots with Qwen3. Early adopters including Anaconda, Hugging Face, Meta, Microsoft, Google, Roboflow, Ollama, and JetBrains are already testing and optimizing their software for the platform. Research institutions like NYU’s Global Frontier Lab have also begun using DGX Spark to accelerate AI experimentation, particularly in sensitive domains like healthcare where data privacy is critical. “DGX Spark allows us to access peta-scale computing on our desktop,” said Kyunghyun Cho, professor at NYU’s Global Frontier Lab. “It enables rapid prototyping and experimentation with advanced AI models, even in privacy-sensitive environments.” By placing a full AI supercomputer on the developer’s desk, NVIDIA aims to democratize access to high-performance AI development, fueling innovation across industries and empowering the next generation of AI breakthroughs.