If you’ve ever dreamed of having your own personal AI supercomputer, NVIDIA has just turned that dream into reality. At the 2025 Consumer Electronics Show (CES) in Las Vegas, NVIDIA unveiled their latest innovation, Project Digits, which promises to bring high-end AI computing power directly to your desk.

At the core of Project Digits is the new GB10 Grace Blackwell Superchip. This isn’t your average chip; it’s designed to run complex AI models that previously would have demanded the resources of a much larger, energy-intensive setup. This superchip offers the capability to handle AI models with up to 200 billion parameters — a figure that places it in the league of some of the most advanced AI systems out there. The system itself is surprisingly compact, resembling the size and form of a Mac mini, yet it’s priced at a starting point of $3,000.

NVIDIA’s CEO, Jensen Huang, shared his vision during the announcement, stating, “AI will be mainstream in every application for every industry. With Project Digits, the Grace Blackwell Superchip comes to millions of developers. Placing an AI supercomputer on the desks of every data scientist, AI researcher, and student empowers them to engage and shape the age of AI.”

Specs and performance

Each Project Digits unit is equipped with an impressive 128GB of unified, coherent memory. Storage isn’t an issue either, with up to 4TB of NVMe storage space. For those who need even more computing muscle, two Project Digits units can be linked to tackle AI models up to 405 billion parameters, which is the scale of Meta‘s Llama 3.1 model.

The GB10 chip within Digits achieves a stunning performance of up to 1 petaflop at FP4 precision, translating to a capability of performing 1 quadrillion AI calculations per second. This is bolstered by NVIDIA’s latest CUDA cores and fifth-generation Tensor Cores, all connected through NVLink-C2C to a 20-core Grace CPU based on Arm technology. The collaboration with MediaTek, a leader in Arm-based SoC designs, has been pivotal in enhancing the chip’s efficiency and performance.

Software and development

Project Digits comes preloaded with NVIDIA’s comprehensive AI software stack, granting users access to tools like development kits, orchestration tools, and a catalog of pre-trained models through the NVIDIA NGC catalog. The system operates on NVIDIA DGX OS, a Linux-based platform tailored for AI development, supporting frameworks like PyTorch, Python, and Jupyter notebooks. Developers can leverage NVIDIA NeMo for fine-tuning AI models and utilize RAPIDS libraries for accelerated data science tasks.

This setup allows for local development and testing of AI models, which can then be seamlessly deployed to cloud environments or data centers without losing the benefits of the Grace Blackwell architecture or the NVIDIA AI Enterprise software platform.

NVIDIA isn’t new to making AI accessible to a broader audience. Just before the CES announcement, in December, they unveiled the Jetson Orin Nano Super, a $249 AI computer aimed at hobbyists and startups, capable of handling AI models with up to 8 billion parameters. This move indicates NVIDIA’s strategy to democratize AI technology, making it available not just to large corporations but also to individual developers, researchers, and students.

Project Digits, with its launch scheduled for May 2025, marks a significant step in this direction, potentially transforming how AI is developed and utilized across various sectors. Whether you’re a seasoned data scientist or a student dipping your toes into AI, NVIDIA’s offering could very well be the tool that shapes your future projects or research endeavors.