Nvidia’s “AI personal supercomputer” costs R$3,000 (and wants to put an end to data centers)

Digits contains a “superchip” and has up to 4 terabytes of NVMe storage, intended for large projects. Nvidia wants to help build AI agents.

“Putting an AI supercomputer on the desks of every data scientist, AI researcher, and student allows them to engage and moldem to was from AI“, explains the founder and CEO of , Jensen Huang, during this Monday’s speech at the technology conference.

Later this year, the company will begin marketing a new “personal AI supercomputer” from 3,000 dollars (about 2,893 euros)which anyone You can use it in your own home or office.

The innovation contains “a ‘superchip’ from Nvidia called GB10 Grace Blackwelloptimized to accelerate the calculations needed to train and run AI models, and comes equipped with 128 gigabytes of unified memory and up to 4 terabytes of NVMe storage to handle especially large AI programs,” explains .

The new device is called , and according to the company, just connect two of these machines through a proprietary interconnect bond high-speed technology, Nvidia says they will be able to run the most capable version available from open source model Llama of Meta, which has 405 billion parameters.

“Os AI agents are the next frontier of AI development, and realizing this opportunity requires complete optimization in a system of LLMs to provide efficient and accurate AI agents,” he said Ahmad Al-Dahlevice president and director of GenAI at Meta, in a statement.

Nvidia says it expects companies to build and maintain AI agents using your technology. “In many ways, every company’s IT department will be the AI ​​agents’ HR department of the future,” said the CEO. “In the future, they will maintain, nurture, integrate and improve a series of AI agents.”