During a 90 - odd - minute keynote computer address atCES 2025 in Las Vegas on Monday , Nvidia CEO Jensen Huang showed off a brawny desktop computer for home AI enthusiasts . Currently endure by Project Digits , this $ 3,000 gimmick take up about as much space as aMac miniand propose 1 PFLOPS of FP4 swim tip functioning .

Nvidia reportedly used its DGX 100 server design as breathing in for the ego - contained screen background AI , with Projects Digits being powered by a 20 - nub GB10 Grace Blackwell Superchip on 128 GB of LPDDR5X memory with a 4 TB NVMe solid - country effort ( SSD ) .

“ AI will be mainstream in every program for every industry , ” Huang say during the keynote . “ With Project Digits , the Grace Blackwell Superchip come to million of developers . point an AI supercomputer on the desk of every data scientist , AI researcher , and student authorize them to employ and shape the age of AI . ”

Nvidia collaborate extensively with Arm to create the Grace Blackwell chip . “ Our collaborationism with weapon system on the GB10 Superchip will fuel the next coevals of design in AI , ” Ashish Karandikar , Nvidia ’s frailty chairwoman of SoC Products , tell inthe company ’s press release .

That much compute powerfulness imply a Project Digits machine will be able to locally work a 200 - billion - loop AI big language model ( LLM ) when it is schedule for discharge in May . ChatGPTusing theGPT-4o model , for example , is only 12 - billion iterations . Digits devices can also be tether together , increasing their theoretical max fashion model size up to 405 billion iterations .

The ship’s company is targeting users who need full - baron Master of Laws lean locally as their basal customers for Project Digits . As such , each gadget comes preloaded with Nvidia ’s entire AI Enterprise software plenty — everything an initiative and enquiry kit would ask to get their AI up and running — with the mind being that users “ epitome , fine - tune , and tryout framework on local Project Digits systems ” before deploy them to the swarm .

Artificial intelligence has long been criticize as a top - down technology in that its development has largely been aim by a prize numeral of company and introduction ( specifically , those wealthy enough to afford the monumental education and operating costs AI requires ) . This is the precise opposite of how the first estimator rotation came about in the seventies to 1990s , where uncountable host of individual hobbyist , tinkerers , and experimentalists drove the development of personal computer engineering .

Project Digits could provide the same opportunity for the AI diligence . Granted , $ 3,000 will surely bound its initial adoption to a point , but to be fair , citizenry did n’t wink at the $ 3,500 price ticket on Apple ’s failedVision Pro headset , and that thing was just a fancy showing strapped to their cheek . If you ’re concerned in trying Digits for yourself , Nvidia has a presentment house - upon its internet site .