Nvidia has officially unveiled a groundbreaking line of artificial intelligence (AI) supercomputers that promise to deliver unprecedented processing power within a compact, portable, and desktop-friendly chassis. Initially referred to as Project Digits, these cutting-edge machines were first introduced at CES 2025 in January and have now been rebranded as the DGX Spark and DGX Station. These powerful computers are powered by Nvidia’s innovative Blackwell Ultra platform, boasting up to a remarkable petaFLOP in processing power—over 1,000 times faster than the most advanced laptops or high-end desktop PCs.
The Blackwell Ultra platform is specifically designed for massive-scale AI training and testing. The introduction of the DGX machines aims to empower data scientists, AI researchers, and students by making this level of performance accessible at a relatively affordable price. Essentially, these supercomputers put the power of a data center into a device small enough to fit comfortably on your desk.
The DGX Spark is a compact machine, slightly smaller than a typical laptop, making it easy to place in a corner of your desk or even carry in your bag. With a height of just under 2 inches (5 centimeters) and a width of slightly under 6 inches (15 cm), its core features the GB10 Grace Blackwell Superchip, which can deliver a staggering 1,000 AI TOPS (trillions of operations per second). This powerful configuration also includes 128 gigabytes of unified system memory and comes bundled with Nvidia’s complete suite of AI software, which features a variety of tools, libraries, and pretrained models.
Among the software offerings is the CUDA Deep Neural Network (cuDNN) library, which enhances neural network layers during both training and inference, as well as a pre-trained SegFormer model. For those interested in acquiring this powerful machine, the Nvidia version of the DGX Spark is available for reservation online, starting at a competitive price of $3,999. Additionally, other models will soon be available from major manufacturers like ASUS, Dell, and Lenovo.
The DGX Station serves as the larger and more powerful sibling to the DGX Spark, resembling a professional workstation in size. It is built around the GB300 Grace Blackwell Ultra Desktop Superchip and features an astonishing 748 GB of large coherent memory, which allows multiple processors to access memory simultaneously. This model also includes Nvidia’s ConnectX-8 SuperNIC, facilitating network connectivity at an incredible speed of up to 800 gigabits per second—fast enough to download approximately five 4K movies within a single second.
Additionally, the DGX Station utilizes the NVLink-C2C Interconnect to connect internal components at a blistering 900 gigabytes per second. This supercomputer is designed to handle large-scale AI training and inferencing workloads directly from your desktop, eliminating the need for additional resources through the cloud. While the DGX Station is not yet available for reservation, Nvidia has announced that it will be accessible later in 2025 through partners like Asus, Dell, HP, Lambda, and Supermicro.
Nvidia’s founder and CEO, Jensen Huang, emphasized that the new DGX machines represent a significant advancement in AI development. By combining immense processing power with a compact design, Nvidia is paving the way for a new era of AI research and development, making high-performance computing more accessible than ever before.