eWEEK content and product recommendations are editorially independent. We may make money when you click on links to our partners. Learn More.
Chipmaker NVIDIA is launching a new mini PC called “Project Digits” that will provide a petaFLOP of power for local AI processing and data science. NVIDIA’s CEO Jensen Huang announced Project Digits at the CES 2025 tech expo taking place in Las Vegas this week.
“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,” Huang said. Designed for developers, researchers, students, and data scientists who need to run artificial intelligence models, the personal supercomputer’s final design is still forthcoming, but it will be small enough to fit on a desk—and possibly even in a bag.
Grace Blackwell Superchip to Power AI Personal Supercomputer
NVIDIA will be able to pack so much artificial intelligence into such a small package thanks to its GB10 Grace Blackwell Superchip, which combines a Blackwell graphics card and an NVIDIA Grace processor. Putting the separate but linked components on a single chip decreases the amount of time it takes to move data between them.
The machine will also offer 128 gigabytes of memory and four terabytes of SSD storage. NVIDIA says the device will be approximately 1,000 more powerful than the best PCs and laptops currently available, even though it will still be powered by a standard electrical outlet.
Project Digits to Offer Up To 1 PetaFLOP of AI Performance
Supercomputer performance capabilities are measured by the number of floating-point arithmetic calculations they can perform on a per-second basis, known as floating-point operations per second, or FLOPS. NVIDIA’s Project Digits computer petaFLOP of performance means it will be capable of performing one quadrillion operations per second.
While this rate is below that of the very best supercomputers, it’s a lot more powerful than a regular PC or laptop. The Project Digits device also promises to be far more portable than a typical supercomputer, balancing size with performance. NVIDIA said that an individual Project Digits computer can run up to 200-billion-parameter large language models. Two Project Digits devices can also be linked using NVIDIA ConnectX Networking, which will support up to 405-billion-parameter models.
Project Digits will be available for purchasing beginning in May, with an anticipated starting cost of $3,000. Interested parties can sign up for notifications on the NVIDIA website.
Learn more about the businesses defining the technology landscape on our list of the top AI companies.