For PC builders, Nvidia has lengthy been synonymous with graphics cards. However in an indication of the occasions, CEO Jensen Huang says his firm has developed from promoting chips to establishing large AI factories.
“Nvidia is an AI infrastructure firm. We’re an infrastructure firm, not simply ‘purchase chips, promote chips,'” Huang instructed journalists at at this time’s GTC occasion in San Jose, California.
His feedback come a day after introducing not one however four GPU architectures to energy next-generation AI providers. Though chip makers have shared roadmaps earlier than, Huang famous it’s nonetheless uncommon for a tech firm to spill the beans on its future merchandise.
“That’s sort of like somebody goes ‘Hello, at this time I’m going to announce my subsequent 4 telephones.’ It is senseless, proper?” Huang stated.
Nonetheless, Nvidia is now not working like a typical electronics vendor. That’s as a result of its main enterprise of promoting GPUs for AI improvement additionally includes firms spending tens of billions of {dollars} on information facilities to energy all of the computing. In his keynote on Tuesday, Huang revealed Nvidia has bought 3.6 million Blackwell GPUs to the highest 4 US-based cloud service suppliers this yr, a rise from 1.3 million Hopper-based GPUs final yr.
“AI infrastructure isn’t one thing you purchase at this time and also you deploy tomorrow,” Huang stated whereas talking to the press on Wednesday. “It’s one thing that you simply put money into two years prematurely and you intend for the whole two years and hopefully you stand it up rapidly.”
Jensen Huang (Credit score: PCMag/Michael Kan)
It’s additionally why Huang launched 4 GPU architectures that can fill Nvidia’s AI portfolio for 2026 and into 2028. “All people’s info needs to be aligned and now we have to plan collectively to construct infrastructure for the world,” he added.
Though the AI demand has despatched Nvidia’s financials hovering, it additionally means the corporate faces extra scrutiny than ever, particularly from its clients. “We’re an AI manufacturing facility now. What which means is a manufacturing facility helps clients earn a living. Our factories straight translate to clients’ revenues,” Huang stated.
“The enterprise bar is far, a lot, a lot increased than earlier than,” he later added. “The competitors bar is far increased than earlier than. The danger tolerance is far decrease than earlier than for all of our clients. Does that make sense? As a result of their revenues straight translate to it. It’s a multi-year funding cycle as a result of we’re speaking about tons of of billions of {dollars}.”
One issue that might enhance the prices for Nvidia and its clients has been Trump’s brewing trade war, which has led to a 20% tariff on Chinese language items. However for now, Huang downplayed the specter of tariffs.
Advisable by Our Editors
“We’ve a extremely agile community of suppliers. They don’t seem to be simply in Taiwan, or simply in Mexico, or simply in Vietnam. They’re sort of distributed in a variety of locations,” Huang instructed journalists.
“It is dependent upon which nations get tariffed, so I believe within the near-term primarily based on what we all know, we aren’t anticipating a major affect to our outlook and our financials,” he later added.
Within the long-term, Nvidia additionally needs to broaden its provide chain to the US, Huang stated. This comes as Taiwan’s TSMC, Nvidia’s essential chip producer, has dedicated to investing $165 billion to construct six new fabs in Arizona.
“We’ve the flexibility to fabricate loads in the US. Not all, however loads,” he added.
Nvidia AI and GPUs at GTC 2025: Every little thing You Have to Know in 16 Minutes
Get Our Finest Tales!
This text might include promoting, offers, or affiliate hyperlinks.
By clicking the button, you affirm you’re 16+ and conform to our
Terms of Use and
Privacy Policy.
You could unsubscribe from the newsletters at any time.
About Michael Kan
Senior Reporter
