Nvidia introduced "Project Digits", a revolutionary desktop-sized AI supercomputer powered by the new GB10 Grace Blackwell Superchip at the CES 2025 tech event. Designed to deliver unparalleled processing power, Project Digits can manage sophisticated AI models with up to 200 billion parameters while operating efficiently from a standard power outlet. This system starts at just $3,000, offering cutting-edge capabilities in a compact form factor reminiscent of a Mac Mini.

“The Grace Blackwell Superchip brings AI supercomputing to millions of developers,” said Jensen Huang, CEO of Nvidia, in the announcement. He emphasised that Project Digits enables data scientists, AI researchers, and students to actively shape the age of AI by placing groundbreaking technology directly on their desks.

Hardware and expandable capacity

Each Project Digits unit boasts an impressive 128GB of unified memory, dwarfing the typical 16GB or 32GB RAM found in laptops, along with up to 4TB of NVMe storage. For even greater demands, two units can be linked to handle models with up to 405 billion parameters, matching the scale of Meta’s Llama 3.1 model. Powered by the GB10 Grace Blackwell Superchip, the system delivers up to 1 petaflop of AI performance at FP4 precision. It features Nvidia’s latest CUDA cores, fifth-generation Tensor Cores, and a 20-core Grace CPU developed in collaboration with MediaTek for optimised power efficiency.

AI-ready software and tools

Project Digits comes preloaded with Nvidia’s extensive AI software library, including development kits, orchestration tools, and pre-trained models available through the Nvidia NGC catalogue. It operates on the Linux-based Nvidia DGX OS, supporting popular frameworks such as PyTorch, Python, and Jupyter Notebooks. Developers can fine-tune models with Nvidia’s NeMo framework or accelerate workflows using Nvidia RAPIDS libraries. The system allows users to develop and test AI locally and seamlessly deploy their models to cloud services or data centres using Nvidia AI Enterprise software, maintaining consistency across platforms.

Compact AI power for all applications

This innovation marks Nvidia’s latest step in making AI accessible. In December, Nvidia launched the Jetson Orin Nano Super, a $249 AI computer for hobbyists and startups, capable of handling models with up to 8 billion parameters. Project Digits takes this mission further, targeting developers, researchers, and students with unparalleled AI capabilities.