at CES 2025 in Las VegasNvidia Revealed. Project Digits, a “personal AI supercomputer” that uses the company's Grace Blackwell hardware platform in a compact form factor.
AI researchers; Project Digits, designed for data scientists and students, is prototyping Nvidia's new GB10 Grace Blackwell Superchip; Provides up to a petflop of computing performance for AI models for tuning and processing.
Nvidia claims that a Project Digits unit can process models up to 200 billionths of a scale. Parameters are roughly related to a model's problem-solving ability, and models with more parameters generally perform better than those with fewer parameters.
Co-developed with MediaTek, the GB10 features an Nvidia Blackwell GPU paired with a 20-core Nvidia Grace CPU. Inside the Project Digits box, the chips are paired with 128GB of memory and up to 4TB of flash storage.
If a job demands it, two Project Digits machines can be linked together to run up to 405 billion-parameter models, Nvidia says.
But they are not cheap. Project Digits devices running Nvidia's Linux-based DGX OS will be available starting in May for $3,000 from “top partners,” the company said.
So it won't be just anyone who can afford their own Project Digits unit. Nvidia CEO Jensen Huang thinks there's a market.
“With Project Digits, the Grace Blackwell Superchip is coming to millions of developers,” he said in a statement. “Putting an AI supercomputer on every data scientist's desk empowers AI researchers and students to engage and shape the era of AI.”