(David Paul Morris/Bloomberg via Getty Images)
For PC builders, Nvidia has long been synonymous with graphics cards. But in a sign of the times, CEO Jensen Huang says his company has evolved from selling chips to constructing massive AI factories.
"Nvidia is an AI infrastructure company. We’re an infrastructure company, not just 'buy chips, sell chips,'" Huang told journalists at today's GTC event in San Jose, California.
His comments come a day after introducing not one but four GPU architectures to power next-generation AI services. Although chip makers have shared roadmaps before, Huang noted it’s still rare for a tech company to spill the beans on its future products.
"That’s kind of like someone goes ‘Hi, today I’m going to announce my next four phones.’ It makes no sense, right?” Huang said.
However, Nvidia is no longer operating like a typical electronics vendor. That’s because its major business of selling GPUs for AI development also involves companies spending tens of billions of dollars on data centers to power all the computing. In his keynote on Tuesday, Huang revealed Nvidia has sold 3.6 million Blackwell GPUs to the top four US-based cloud service providers this year, an increase from 1.3 million Hopper-based GPUs last year.
"AI infrastructure isn’t something you buy today and you deploy tomorrow," Huang said while speaking to the press on Wednesday. "It is something that you invest in two years in advance and you plan for the entire two years and hopefully you stand it up quickly."
It’s also why Huang introduced four GPU architectures that will fill Nvidia’s AI portfolio for 2026 and into 2028. “Everybody’s information has to be aligned and we have to plan together to build infrastructure for the world,” he added.
Although the AI demand has sent Nvidia’s financials soaring, it also means the company faces more scrutiny than ever, especially from its customers. “We’re an AI factory now. What that means is a factory helps customers make money. Our factories directly translate to customers’ revenues,” Huang said.
“The business bar is much, much, much higher than before,” he later added. “The competition bar is much higher than before. The risk tolerance is much lower than before for all of our customers. Does that make sense? Because their revenues directly translate to it. It’s a multi-year investment cycle because we’re talking about hundreds of billions of dollars.”
One factor that could increase the costs for Nvidia and its customers has been Trump’s brewing trade war, which has led to a 20% tariff on Chinese goods. But for now, Huang downplayed the threat of tariffs.
“We have a really agile network of suppliers. They are not just in Taiwan, or just in Mexico, or just in Vietnam. They are kind of distributed in a lot of places,” Huang told journalists.
“It depends on which countries get tariffed, so I think in the near-term based on what we know, we are not expecting a significant impact to our outlook and our financials,” he later added.
In the long-term, Nvidia also wants to expand its supply chain to the US, Huang said. This comes as Taiwan’s TSMC, Nvidia’s main chip manufacturer, has committed to investing $165 billion to build six new fabs in Arizona.
“We have the ability to manufacture a lot in the United States. Not all, but a lot,” he added.