It's small enough to fit in a bag. However, this miniature laptop would be able to run AI models that usually require processing by huge data centers via the Cloud, according to Live Science.
This innovation was presented by the Nvidia company during the 2025 edition of the CES show, one of the most important technological events of the year, which is held from January 7 to 11 in Las Vegas. Called “Project Digits”, this mini-PC was designed by a team of scientists, developers and researchers, to be as powerful as a supercomputer, machines capable of achieving the highest possible performance, particularly in terms of calculation speed. .
1,000 times more efficient than a traditional computer
“Project Digits” contains the GB10 Grace Backwell, an ultra-powerful core. This is powered by the most efficient processor developed by the company Nvidia, all with 128 GB of memory and 4 TB of SSD storage.
China, India, Nigeria: our computers sow death
The machine presented by the artificial intelligence giant could thus offer a power of 1 petaflop – the flop is a unit of measurement of the speed of calculation of computer systems. 1 petaflop corresponds to 10 to the power of 15 flops.
This new computer is therefore 1,000 times more powerful than an average personal computer. However, it is still far from the capacity of the most capable supercomputers, which can achieve up to more than 1,000 petaflops.
A computer capable of running complex AI models
But then in what context can this computer be used? This “desktop” computer should allow researchers to run generative AI systems that use up to 200 billion parameters.
The more parameters an AI model has, the more capable it is of understanding complex nuances and generating precise responses. Two Project Digits devices connected together could even reach 405 billion parameters. For comparison, the first version of ChatGPT released in 2022 had around 175 billion parameters.
“Nvidia's DIGITS project enables robotics researchers…to experiment, refine and scale solutions faster than ever, all while fitting on a desktop”said Raquel Urtasun, professor of computer science at the University of Toronto in a press release from her automobile company Waabi.
For the moment, no image of the prototype has been revealed by Nvidia, the design of the device not being completely finalized. It should be marketed from May at a price of around 3,000 dollars, a very reasonable sum, with the possibility for those interested to register to reserve one.
