Nvidia’s $3,000 ‘personal AI supercomputer’ will let you ditch the data center

Nvidia already sells boatloads of computer chips to every major company that builds its own products artificial intelligence models. But now, at a time when public interest in open source and do-it-yourself AI is growing, the company announced that it will also begin offering a “personal AI supercomputer” later this year, starting at $3,000, that anyone can use in their own home or office.

Nvidia’s new desktop machine, called Digits, will go on sale in May and is about the size of a small book. It contains an Nvidia “superchip” called the GB10 Grace Blackwell, optimized to accelerate the calculations needed to train and run AI models, and is equipped with 128GB of unified memory and up to 4TB of NVMe storage to handle particularly large AI programs.

Jensen Huang, founder and CEO of Nvidia, announced the new system, along with several other AI offerings, during a keynote today at CESan annual computer industry confab held in Las Vegas (you can check out all the biggest announcements at WIRED CES Live Blog).

“Putting an AI supercomputer on the desks of every data scientist, AI researcher and student empowers them to engage and shape the AI ​​era,” Huang said in a statement released ahead of his keynote.

Nvidia says the Digits machine, which stands for “GPU Intelligence Learning System for Deep Learning,” will be able to handle a large language model with up to 200 billion parameters, a rough measure of the complexity and size of the model. To do that today, you’d need to rent space from a cloud provider like AWS or Microsoft, or build a custom system with a handful of chips designed to run AI. If double-digit machines are connected using their own high-speed connection link, Nvidia says they’ll be able to run most capable an open source version of Meta’s Llama model available, which has 405 billion parameters.

The figures will make it easier for hobbyists and researchers to experiment with models that approach the basic capabilities of OpenAI’s GPT-4 or The Google Twins in their offices or basements. But the best versions of these proprietary models, housed in giant data centers owned by Microsoft and Google, are likely bigger, as well as more powerful, than anything Digits can handle.

Nvidia is one of the biggest beneficiaries of the AI boom. Its share price has soared in the past few years as technology companies clamored to buy huge quantities of the advanced hardware chips it makes, a crucial ingredient in the development of cutting-edge AI. The company has proven adept at creating AI-optimized hardware and software, and its product roadmap has become an important signal of where the industry is expected to be headed.

When released, Digits will be the most powerful consumer PC hardware Nvidia offers. It already sells a set of AI development chipsets known as Jetson that start at roughly $250. They can control smaller AI models and be used as a mini desktop computer or installed on a robot to test different AI programs.