Nvidia Project Digits: A Linux-powered desktop for AI developers

The buzz at CES 2025 has been all about AI-powered TVs, innovative smart home gadgets, and the latest laptop releases. But for me, the real game-changer is Nvidia's Project Digits. This desktop AI supercomputer is set to revolutionize the world of artificial intelligence by delivering unmatched computing power directly to developers, researchers, and students. And guess what? It's powered by Nvidia's DGX OS, a tailored version of Ubuntu Linux 22.04.

DGX OS isn't just any Linux distro; it's fine-tuned with specific optimizations, configurations, drivers, and tools for monitoring and diagnostics. It's designed to run AI, machine learning, and analytics applications seamlessly on Nvidia DGX Supercomputers.

More Than Just Hardware

Nvidia isn't stopping at hardware. They're also rolling out AI software development kits, orchestration tools, and the Nvidia NGC catalog with frameworks and models. Plus, there's the Nvidia NeMo framework for fine-tuning models and Nvidia Rapids libraries to speed up data science tasks. And since it's built on Ubuntu Linux, you can run all your favorite Ubuntu and Linux software. I mean, who wouldn't want to play Doom on this beast?

Compact Powerhouse

Imagine a box the size of a Mac mini packing a petaflop of AI performance. That's what Nvidia's Project Digits promises. A petaflop, for those not in the know, is a quadrillion (1,000,000,000,000,000) floating-point operations per second. Before IBM's Roadrunner supercomputer in 2008, no one had ever reached that speed. Now, you could have it right on your desk. Even the latest Top 500 SuperComputer List features machines that barely surpass two petaflops, and they need over a thousand cores to do it. And yes, they all run Linux.

Project Digits, powered by Nvidia's new GB10 Grace Blackwell Superchip, is nearly as fast as those top-tier supercomputers. With 128GB of memory and Nvidia's AI software stack preinstalled, developers can work on large AI models up to 200 billion parameters locally, then easily deploy them to data centers or the cloud.

The Grace Blackwell Superchip

Named after Grace Hopper, the legendary computer scientist and "Mother of Cobol," and David Harold Blackwell, a pioneering mathematician, the Grace Blackwell Superchip is a marvel. It features an ARM-based Grace CPU with 10 Cortex-X925 and 10 Cortex-A725 cores, totaling 20 cores. This is complemented by a Blackwell GPU with the latest CUDA and RT cores. With 128GB of RAM and up to 4TB of flash storage, this chip is ready to tackle any task.

Nvidia promises that AI developers can handle models up to 200 billion parameters on a single unit. For even more demanding tasks, you can connect two Project Digits machines to manage models with up to 405 billion parameters.

A New Era for AI Development

Nvidia's CEO, Jensen Huang, highlighted the transformative potential of Project Digits, saying, "Placing an AI supercomputer on the desks of every data scientist, AI researcher, and student empowers them to engage with and shape the age of AI."

Priced at a starting point of $3,000, Project Digits aims to democratize high-performance AI development, making it accessible to small businesses and educational institutions. At this price, I wouldn't be surprised to see these desktops popping up in homes.

Set to launch in May 2025, this Linux-powered system is poised to redefine AI development by bringing data center-level performance to a compact, desk-friendly package. I'm already planning how to get my hands on one.