Nvidia CEO Jensen Huang introduces DGX Spark and DGX Station—personal AI supercomputers powered by the Grace Blackwell platform. Discover how these cutting-edge systems enable local AI model development and seamless cloud integration
Nvidia CEO Jensen Huang unveiled two "personal AI supercomputers" called DGX Spark and DGX Station, both powered by the company’s Grace Blackwell platform and designed for users to work on large AI models with or without a connection to a data centre.
Huang announced on Tuesday during his Nvidia GTX keynote that the DGX Spark is now available for preorder, while the DGX Station is expected to be available later this year through manufacturing partners such as ASUS, BOXX, Dell, HP, Lambda and Supermicro.
Initially introduced as "Project DIGITS" in January, these desktop systems enable developers, researchers and data scientists who need to prototype, optimize and run large AI models locally to have access to advanced AI capabilities.
AI developers can use DGX systems as "bridge systems" to seamlessly transfer their models from desktops to DGX Cloud or other AI cloud infrastructures with minimal code changes, or they can function as independent desktop AI labs.
DGX Spark, claimed as the world's smallest AI supercomputer, offers exceptional performance and capabilities, empowering millions of researchers, data scientists, robotics developers and students to advance generative and physical AI.
At the core of DGX Spark is the Nvidia GB10 Grace Blackwell Superchip, designed specifically for a desktop form factor.
Equipped with a powerful Nvidia Blackwell GPU featuring fifth-generation Tensor Cores and FP4 support, the GB10 can deliver up to 1,000 trillion operations per second of AI compute, making it ideal for fine-tuning and inference tasks using advanced AI reasoning models such as the Nvidia Cosmos Reason world foundation model and the Nvidia GR00T N1 robot foundation model.
Nvidia's full‑stack AI platform simplifies prototyping, optimizing and iterating workflows, enabling DGX Spark users to seamlessly transfer their models from desktops to DGX Cloud or any accelerated cloud or data centre infrastructure with minimal code changes.
Nvidia DGX Station brings data centre‑level performance to desktop AI development. Featuring an expansive 784 GB of coherent memory, it is the first desktop workstation built with the Nvidia GB300 Grace Blackwell Ultra Desktop Superchip, designed to accelerate intensive training and inference tasks.
The GB300 Desktop Superchip combines a powerful Nvidia Blackwell Ultra GPU, equipped with the latest Tensor Cores and FP4 precision, with an Nvidia Grace CPU via NVLink‑C2C for optimal system communication and performance.
DGX Station also includes the Nvidia ConnectX‑8 SuperNIC, engineered to enhance hyperscale AI computing workloads. With support for networking speeds of up to 800Gb/s, the ConnectX‑8 enables fast, efficient data transfers and seamless connectivity between multiple DGX Stations for larger‑scale tasks.