What concerns are there about an AI bubble? : NPR

Perhaps no one embodies the artificial intelligence craze more than Jensen Huang, chief executive of chip giant Nvidia, which has seen its value surge 300% over the past two years.

These are certainly heady times for Huang, which makes it all the more understandable that his first statement to investors on the recent earnings call was an attempt to downplay bubble fears.

“There has been a lot of discussion about an AI bubble,” he told shareholders. “From our vantage point, we see something very different.”

Take in the AI ​​bubble discourse and something becomes clear: Those who stand to gain most from artificial intelligence spending are never slowing down, declaring that critics who worry about an over-hyped investment frenzy are all wrong.

“I don’t think this is the beginning of a bearish cycle,” White House AI czar and venture capitalist David Sachs said on his podcast All-In. “I think we’re in a boom. We’re in a super-cycle of investment.”

“To me the idea that five years from now we’re going to have a demand problem seems pretty absurd,” said Ben Horowitz, a prominent Silicon Valley investor. He added: “If you look at demand and supply and what’s going on and the multiples against growth, it doesn’t look like a bubble at all to me.”

Appearing on CNBC, JPMorgan Chase executive Mary Callahan Erdoz said that calling the amount of money flowing into AI right now a bubble is “a crazy concept,” declaring that “we are on the verge of a big, big revolution in the way companies operate.”

Still, one look at what’s actually going on in the AI ​​industry right now is enough to generate serious skepticism, said Paul Kedrosky, a venture capitalist who is now a research fellow at MIT’s Institute for the Digital Economy.

He said huge amounts of capital were being invested in a “revolution”, mostly based on speculation.

“The technology is very useful, but the pace at which it is improving has more or less stalled,” Kedroski said. “So the notion that the revolution will continue to the beat of the same drum for the next five years is tragically wrong.”

huge cash flow

The flow of money is happening so fast that financial experts are amazed.

Take OpenAI, the creator of ChatGPT, which started the AI ​​race in late 2022. Its CEO Sam Altman has said that the company is generating $20 billion in revenue per year, and it plans to spend $1.4 trillion on data centers over the next eight years. Of course, this growth will depend on increasing sales as more people and businesses purchase its AI services.

There is reason to doubt. A growing body of research indicates that most companies aren’t seeing chatbots impacting their bottom lines, and only 3% pay for AI, according to one analysis.

“These models are being hyped up, and we are investing more than we should,” said Daron Acemoglu, an economist at MIT who was awarded the 2024 Nobel Memorial Prize in Economic Sciences.

He said, “I have no doubt that in the next ten years AI technologies will emerge that will add real value and increase productivity, but everything we hear from the industry now is hyperbole.”

Nonetheless, Amazon, Google, Meta and Microsoft are set to collectively invest nearly $400 billion on AI this year, mostly to finance data centers. Some companies are willing to devote nearly 50% of their current cash flow to data center construction.

Or to put it another way: Every iPhone user on Earth would have to pay more than $250 to cover that expense. “That’s not going to happen,” Kedroski said.

To avoid spending too much of the cash they have, big Silicon Valley companies like Meta and Oracle are tapping private equity and debt to finance the industry’s data center building spree.

Paving the AI ​​future with loans and other risky financing

An assessment by Goldman Sachs analysts found that hyperscaler companies – technology companies that have massive cloud and computing capabilities – have taken on $121 billion of debt in the past year, more than 300% the industry’s typical debt load.

Gil Luria, an analyst at the DA Davidson investment firm who has been tracking Big Tech’s data center boom, said some of the financial maneuvering being done by Silicon Valley is structured to keep the presence of debt off the balance sheet, using what are known as “special purpose vehicles.”

The tech firm invests in the data center, outside investors put up most of the cash, then the special purpose vehicle borrows the money to buy the chips housed inside the data center. The tech company benefits from increased computing capacity but does not burden the company’s balance sheet with debt.

For example, a special purpose vehicle was recently funded by Wall Street firms Blue Owl Capital and Meta for a data center in Louisiana.

The design of the deal is complicated but it goes something like this: Blue Owl borrowed $27 billion for the data center. That loan is backed by Meta’s payment for leasing the facility. Meta essentially has a mortgage on the data center. Meta owns 20% of the entity, but receives all the computing power generated by the data center. Due to the financial structure of the deal, the $27 billion loan never appears on Meta’s balance sheet. If the AI ​​bubble bursts and the data center goes dark, Meta will be forced to pay Blue Owl several billion dollars for the value of the data center.

According to Luria, such financial arrangements have had a checkered past.

“The term special purpose vehicle came up about 25 years ago with a little company called Enron,” Luria said, referring to the energy company that collapsed in 2001.

Huge expenses dependent on returns may be a fantasy

Silicon Valley is taking on all this new debt with the assumption that massive new revenues from AI will cover the tab. But still, there is reason for doubt.

Morgan Stanley analysts estimate that Big Tech companies will spend about $3 trillion on AI infrastructure by 2028, with only half of that coming from their own cash flow.

“If the artificial intelligence market remains steady in its growth, very soon we will have overbuilt capacity, and debt will become worthless, and financial institutions will lose money,” Luria said.

Twenty-five years ago, when the original dot-com bubble burst, among other factors, debt financing created fiber-optic cables for a future that had not yet come, Luria said, a lesson, it appears, tech companies aren’t worried about repeating.

“If we get to the point after spending hundreds of billions of dollars on data centers that we don’t need in a few years from now, we’re talking about another financial crisis,” he said.

Circular deals raise even more concerns

Another aspect of the over-heated AI landscape that is raising eyebrows is the circular nature of the investment.

Take the recent $100 billion deal between Nvidia and OpenAI.

Nvidia will pump that amount into OpenAI to bankroll the data centers. OpenAI will then feed those features to Nvidia’s chips. Some analysts say this structure, where Nvidia is essentially subsidizing one of its largest customers, artificially inflates the real demand for AI.

“The idea is that I’m Nvidia and I want OpenAI to buy more of my chips, so I pay them money to do it,” Kedrosky said. “It’s quite common on a small scale, but it’s unusual to see it in the tens and hundreds of billions of dollars,” noting that the last time it was prevalent was during the dot-com bubble.

Lesser-known companies are also getting in on the action.

CoreWeave, once a crypto mining startup, pivoted to data center building to ride the AI ​​boom. Major AI companies are turning to CoreWeave to train and run their AI models.

OpenAI has struck billion-dollar deals with CoreWeave, in which CoreWeave’s chip capacity in data centers is rented to OpenAI in exchange for stock in CoreWeave, and OpenAI, in turn, can use that stock to pay its CoreWeave rental fees.

Meanwhile, Nvidia, which is also part of CoreWave, has a deal that guarantees Nvidia will gobble up any unused data center capacity through 2032.

“The danger is that these types of deals ultimately reveal the deck of cards,” said Acemoglu, an MIT economist.

Some high profile investors see bursting bubbles on the horizon

Some influential investors are showing signs of bubble panic.

Tech billionaire Peter Thiel sold his entire stake in Nvidia earlier this month, worth about $100 million. This happened after SoftBank sold a stake in Nvidia worth about $6 billion.

And in recent weeks, AI bubble pessimists have rallied around hedge-fund investor Michael Burry, who bet hundreds of millions of dollars against the housing market in 2008. He was the subject of the 2015 film The Big ShortHowever, since then, he has had a mixed reputation for market predictions, with warnings about impending collapse that never materialized,

For what it’s worth, Barry is now betting against Nvidia, and accusing the AI ​​industry of hiding behind fancy accounting tricks. He is involved in circular deals between companies.

“True final demand is ridiculously small. Almost all customers are financed by their dealers,” Bury wrote on X. He later wrote: “OpenAI is the lynchpin here. Can anyone name their auditor?”

As tech companies sink billions of dollars into data centers, some executives are admitting to themselves that there seems to be some overzealousness.

Sam Altman, CEO of OpenAI, told reporters in August: “Are we in a phase where investors overall are extremely excited about AI? My opinion is yes. Is AI the most important thing to happen in a very long time? My opinion is also yes.”

And Google Chief Executive Sundar Pichai recently told the BBC that there are “elements of irrationality” in the AI ​​market right now.

Asked how Google would fare if the bubble bursts, Pichai responded: “I think no company, including us, will be immune.”