NVIDIA Unveils DGX Spark: The World’s Smallest AI Supercomputer
News Summary: NVIDIA founder and CEO Jensen Huang recently delivered the DGX Spark to Elon Musk at SpaceX. This week, NVIDIA and its partners began shipping DGX Spark, touted as the world’s smallest AI supercomputer, making advanced AI computing more accessible in a compact desktop form factor. Major companies like Acer, ASUS, Dell Technologies, GIGABYTE, HPI, Lenovo, and MSI have launched DGX Spark systems to expand powerful AI capabilities.
NVIDIA has officially announced that it will begin shipping the highly anticipated NVIDIA DGX Spark™, recognized as the smallest AI supercomputer on the market. As AI workloads rapidly evolve, many developers find their existing PCs, workstations, and laptops insufficient, compelling them to transition their work to cloud services or local data centers.
DGX Spark introduces a new category of computing, providing a petaflop of AI performance alongside 128GB of unified memory—all nestled within a sleek desktop design. This innovative setup empowers developers to perform inference on AI models with up to 200 billion parameters and fine-tune local models of up to 70 billion parameters, marking a significant step for desktop AI capabilities.
“Back in 2016, we launched DGX-1 to empower AI researchers with their own supercomputer. I personally delivered the first unit to Elon Musk at OpenAI, which eventually led to the development of ChatGPT and sparked the AI revolution,” stated Jensen Huang, NVIDIA’s founder and CEO. “With DGX Spark, we reaffirm our mission—putting an AI supercomputer into the hands of every developer to stimulate the next wave of breakthroughs.”
DGX Spark amalgamates the complete NVIDIA AI platform, integrating cutting-edge GPUs, CPUs, networking, CUDA® libraries, and NVIDIA’s AI software stack into a compact design suitable for labs and offices alike. This integration offers the performance needed to accelerate both agentic and physical AI developments, converting your desktop into a comprehensive AI development environment.
Featuring up to 1 petaflop of AI performance, DGX Spark is powered by the NVIDIA GB10 Grace Blackwell Superchip, coupled with NVIDIA ConnectX®-7 200 Gb/s networking and NVIDIA NVLink™-C2C technology that offers five times the bandwidth of fifth-generation PCIe, and 128GB of coherent CPU-GPU memory. This configuration positions DGX Spark as a versatile tool for developers pushing the envelope of AI.
Developers will find the NVIDIA AI software stack preinstalled, facilitating immediate engagement with AI projects. With access to essential NVIDIA AI ecosystem tools—including models, libraries, and NVIDIA NIM™ microservices—DGX Spark supports local workflows for customizing advanced models, building vision search agents, and creating AI chatbots optimized specifically for this platform.
To commemorate the global launch of DGX Spark, Jensen Huang personally delivered one of the initial units to Elon Musk, who has been integral to DGX’s legacy since the shipment of the first NVIDIA DGX™-1 supercomputer in 2016. Other notable early adopters like Anaconda, Google, Meta, and Microsoft are already validating and optimizing their tools for DGX Spark, enhancing its capability and reach in the AI landscape.
Leading research organizations, including NYU’s Global Frontier Lab, have previewed DGX Spark to enhance their AI development processes. “DGX Spark enables us to access peta-scale computing directly from our desktop,” remarked Kyunghyun Cho, professor of computer and data science at NYU. “This game-changing approach to AI research allows rapid prototyping and experimentation with complex AI algorithms, even in sensitive fields like healthcare.”
Order DGX Spark Today
Starting Wednesday, October 15, the DGX Spark can be ordered directly from NVIDIA’s website. Additionally, partner systems will also be available through top retailers such as Acer, ASUS, Dell Technologies, GIGABYTE, HP, Lenovo, MSI, and Micro Center locations across the U.S., alongside NVIDIA’s global channel partners.
