The CEO of Nvidia (NVDA), Jensen Huang, has major projects for the IA data center industry. During the company’s GTC conference, the executive and co-founder of the flea giant made its debut Spectrum-X and Quantum-X of Nvidia Silicon photonic networking systems.
The networking switches connect several NVIDIA graphic chips to act as a giant system. Huang says that Spectrum-X and Quantum-X photonics can reach hundreds and even thousands of GPUs and will allow data center companies to deploy up to 1 million GPU clusters.
It is a massive jump of the largest AI current AI supercomputer, XAI Colossuswhich is made up of 100,000 NVIDIA Hopper GPU. And according to Huang, these millions of clusters data centers will be connected to other data centers at a million dollars nearby to form massive installations of data centers.
“In the coming years, we will build giant AI factories. Not normal AI factories … those you see in space, “said Huang.
The CEO spent a good part of his time on stage during his Keynote GTC explaining how the latest advances of Nvidia, ranging from Spectrum-X and Quantum-X Photonics to its Blackwell Ultra and Vera Rubin Superchips, will help the AI customers of the company, including better income services.
“AI factories are directly linked to income, and if the flow is not good, your income is injured. If you do not have enough capacity, your income is injured. If you do not produce something of great value … Your income is injured,” said Huang.
Find out more: How does Nvidia earn money?
The CEO has assimilated AI factories to car factories, explaining how if car manufacturers do not have the right factories and products offered, their income will suffer.
“This could apply to a car business. If your capacity is not high enough, your income is injured. If your use is not good, your income is injured. If you build a car, no one wants to buy it, your income is injured,” he said.
“It’s a factory. Except that it is a token factory. It’s a new world. An industrial revolution. “
This industrial revolution will not only take place within data centers. In addition to its chips for massive servers systems, NVIDIA has also shown a handful of AI-centered computers, including its mini-desktop DGX Spark and full -scale officeDGX station the size of an operation.
Packed with NVIDIA Grace Blackwell Super Puces, computers are designed for AI developers and researchers. And when associated with the other AI platforms in the company, they constitute a new approach for the company that revolves around AI.
