Tipping point or bubble? Nvidia CEO sees AI transformation while skeptics count the risks

By Arsheeya Bajwa and Stephen Nellis

SAN FRANCISCO (Reuters) -Nvidia CEO Jensen Huang says he does not see an AI bubble, but rather a tipping point.

In his view, the kind of computing his company specializes in will come to suffuse everything from writing code to running legions of robots in the everyday world.

But a growing band of market skeptics are concerned the only way off a tipping point is down.

The chip giant on Wednesday produced results and forecasts that beat expectations, allaying immediate fears. But there are longer-term worries that Nvidia's growth could be crimped by factors beyond the control of even the most valuable listed company in human history, now worth more than $4.5 trillion. 

In a regulatory filing, Nvidia disclosed the majority of its booming business rests on four unnamed customers.

In the third quarter, 61% of its $57 billion in revenue came from those customers, up from a 56% concentration among four customers in the previous quarter. Past announcements suggest they could include Microsoft, Meta and Oracle.  

Nvidia also doubled the money it spends renting back its own chips from cloud customers to $26 billion, up from $12.6 billion in the second quarter, with those contracts stretching to at least 2031. The company in the past quarter said it would invest up to $100 billion in OpenAI and $10 billion in Anthropic, two major customers.

Its high reliance on just a few customers it is entangled with and the circular nature of some of its deals have raised concerns, particularly as none of the entities have reported massive profits from AI yet.

"A lot of this growth is coming from loss-making startups or loss-making projects, so most likely the cycle ends badly unless all these companies agree to stop spending together and let profits shine through, which is a near impossibility," said Chaim Siegel, an analyst at research firm Elazar Advisors.

HUANG BRUSHES OFF BUBBLE TALK

During an earnings call, Huang said Nvidia sees something "very different" from the market talk of an AI bubble. 

He outlined three transitions as part of a vision where Nvidia could reign supreme for years to come.

First, there is the shift of non-AI software such as engineering simulations and data science away from traditional central processors to Nvidia's high-powered chipsets. 

Then there is the invention of entirely new categories of software such as coding assistants. And later, he sees AI jumping from virtual applications such as chatbots to the physical world of cars, robots and more.

"These three fundamental dynamics each will contribute to infrastructure growth in the coming years. Nvidia is chosen because our singular architecture enables all three transitions," Huang said.

But building all of the required data centers to meet that vision will require an enormous amount of land and power, concerning even Nvidia bulls such as Ivana Delevska, chief investment officer of Spear Invest, which holds the company's shares in an actively managed exchange-traded fund.

Huang addressed those concerns on the conference call, saying Nvidia was working diligently to make sure factors beyond the chip supply chain would not stand in its way.

"We've now established partnerships with so many players in land and power and (data center buildings), and of course, financing these things," he said. "None of these things are easy, but they're all tractable, and they're all solvable things."

But as companies like Google parent Alphabet and Amazon design their own AI chips and begin to sell them to a similar customer base, some analysts said a looming era of Nvidia maintaining its dominance is far from certain. 

"They have said they are sold out for the year and probably next, which leads me to wonder what possible upside surprises they could offer," said Jay Goldberg, senior analyst at Seaport Research Partners, which has a "sell" rating on Nvidia. "The list of things that could go wrong for Nvidia is longer than the list of things that could go right."

(Reporting by Arsheeya Bajwa in Bengaluru and Stephen Nellis in San Francisco; Editing by Jamie Freed)