If there has been a technology company with a 2024 like no other, it was Nvidia. The American manufacturer skyrocketed its market value thanks to providing the necessary hardware for data centers used in the training and execution of artificial intelligence models andat times, became the most valuable company in the world. It is currently in second positiononly behind Apple and surpassing Microsoft and Google. This Monday, the company presented within the framework of the CES 2025 a new hardware for AI, a mini supercomputer, similar in size to a MacMiniaimed at researchers, data scientists, and students who want to experiment with AI models—such as ChatGPT-style chatbots and image generators—from home. It is about Project Digits.
Project Digits is a desktop computer that will cost 3,000 dollars and includes the new superchip GB10 Grace Blackwell from Nvidia. It will be released in May, can function as a stand-alone PC or connect to a Windows or Mac computer and handle AI models with up to 200 billion parameters.
In the presentation prior to the opening of the Consumer Electronics Show to the public this Tuesday, Jensen Huangfounder and CEO of Nvidia, described the new system as ‘a cloud computing platform that sits on your desk’. The company has designed Project DIGITS as a bridge between what it means to develop locally on a computer and deployment in the cloud. Developers can build and test AI applications with Project DIGITS, then move them to cloud services or data centers that use hardware with Grace Blackwell architecture and Nvidia AI Enterprise software platform.
‘AI will be mainstream in every application and for every industry. With Project Digits, Grace Blackwell superchip reaches millions of developers‘ says Huang. ‘Placing an AI supercomputer on the desk of every data scientist, AI researcher and student allows them to participate in and shape the AI era.’
Each Project Digits comes equipped with 128GB unified memory, up to 4TB storage NVMe and Two Project Digits can be linked to handle models with up to 400 billion parameters. As a reference, the best model of Goal, Call 3.1uses that many parameters.
The GB10 chip offers up to 1 petaflop of AI performance1 quadrillion AI calculations per second, with FP4 precision and the system has the cores CUDA latest generation and cores Tensor fifth generation, both from Nvidia, connected via NVLink-C2C to one Grace CPU with 20 Arm cores high energy efficiency. MediaTekknown for its Arm-based chip designs, collaborated on the development of the GB10 to optimize its power efficiency and performance.
Users also will gain access to Nvidia’s AI software librarywhich includes development kits, orchestration tools, and pre-trained models available through the Nvidia NGC catalog. The system works with Nvidia DGX OSbased on Linuxand is compatible with popular frameworks such as PyTorch, Python and Jupyter. Developers can tune models using the framework Nvidia NeMo and accelerate data science workflows with libraries Nvidia RAPIDS.