What concerns do you have about the AI ​​bubble?: NPR

Perhaps no one embodies the artificial intelligence mania more than Jensen Huang, chief executive officer of semiconductor giant Nvidia, whose value has soared 300% in the past two years.

It’s certainly been a turbulent time for Mr. Hwang, but this makes it all the more understandable why his first comments to investors during a recent earnings call were aimed at allaying bubble fears.

“There’s been a lot of talk about an AI bubble,” he told shareholders. “From our perspective, we see something completely different.”

If we incorporate the AI ​​bubble argument, something becomes clear. The people who stand to gain the most from spending on artificial intelligence never decreasing are the ones who are claiming that critics worried about an overhyped investment frenzy are all wrong.

“I don’t think this is the start of a bust cycle,” White House AI czar and venture capitalist David Sachs said on the “All-In” podcast. “I think we’re in a boom right now. We’re in a supercycle of investing.”

“The idea that we’re going to have a demand problem five years from now seems absolutely ridiculous to me,” said Ben Horowitz, a prominent Silicon Valley investor, adding, “If you look at supply and demand, what’s going on, and the multiples of growth, it doesn’t seem like a bubble to me at all.”

JPMorgan Chase executive Mary Callahan Erdoes appeared on CNBC and said it’s a “crazy concept” to call the money currently flowing into AI a bubble, declaring: “We’re on the precipice of a big, big revolution in how companies operate.”

But just an inside look at what’s actually going on in the AI ​​industry right now is enough to raise serious questions, said Paul Kedrowski, a venture capitalist and now a fellow at MIT’s Institute for the Digital Economy.

He said an astonishing amount of capital is being poured into a “revolution” that remains largely speculative.

“This technology is very useful, but the pace of its advancement has more or less stalled,” Kedrowski said. “So the idea that the revolution will continue playing to the same drum beat for the next five years is sadly false.”

huge cash injection

Huge sums of money are flowing in at an alarming rate for financial professionals.

Take OpenAI, the ChatGPT maker that launched its AI race in late 2022. CEO Sam Altman said the company generates $20 billion in annual revenue and plans to spend $1.4 trillion on data centers over the next eight years. Of course, that growth will depend on sales continuing to swell from more and more people and businesses buying the company’s AI services.

There are reasons to be skeptical. A growing body of research shows that most companies don’t think chatbots have an impact on their bottom line, and one analysis shows that only 3% of people pay for AI.

“These models are overhyped and we are investing more than we should,” said Daron Acemoglu, a Massachusetts Institute of Technology economist who won the 2024 Nobel Prize in economics.

“There is no doubt that within the next 10 years we will see AI technologies that add real value and improve productivity, but much of what we are currently hearing from the industry is exaggeration,” he said.

Nevertheless, Amazon, Google, Meta, and Microsoft will collectively spend about $400 billion on AI this year, much of which will go toward funding data centers. Some companies plan to use about 50% of their current cash flow to build data centers.

In other words, every iPhone user on the planet would have to pay more than $250 for that amount of spending. “That’s not going to happen,” Kedrowski said.

Silicon Valley giants like Meta and Oracle are turning to private equity and debt to finance the industry’s data center construction rush to avoid burning through too much cash.

Opening up your AI future with debt and other risky financing

One review by Goldman Sachs analysts found that hyperscaler companies (high-tech companies with large cloud and computing capabilities) had accumulated $121 billion in debt over the past year, an increase of more than 300% compared to the industry’s typical total debt.

Gil Luria, an analyst at investment firm DA Davidson who has been tracking Big Tech’s data center boom, said some of Silicon Valley’s monetary policy is structured to keep debt off its balance sheet, using so-called “special purpose vehicles.”

A tech company invests in a data center, outside investors contribute most of the cash, and then a special purpose vehicle borrows money to buy the chips inside the data center. The technology company benefits from increased computing power, but debt doesn’t weigh on the company’s balance sheet.

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

The contract design is complex, but looks like this: Blue Owl took out a $27 billion loan for its data centers. This debt is supported by Meta’s facility lease payments. Meta basically has a lien on the data center. Meta owns 20% of the entity but gets all the computing power that the data center generates. Due to the financial structure of the transaction, the $27 billion loan will not appear on Meta’s balance sheet. If the AI ​​bubble bursts and the data centers go dark, Meta would have to pay Blue Owl billions of dollars in value for the data centers.

Luria said such financing has a checkered past.

“The term special purpose vehicle was coined about 25 years ago by a small company called Enron,” Luria said, referring to the energy company that went bankrupt in 2001. “What’s different now is that companies aren’t hiding it. But that doesn’t mean we should rely on it to build our future.”

It may be an illusion that huge expenditures are influenced by profits.

Silicon Valley is taking on all of this new debt, with the assumption that the huge new revenues from AI will cover the burden. But here too there is room for doubt.

Morgan Stanley analysts predict that Big Tech companies will spend about $3 trillion on AI infrastructure by 2028, but only half of that will be covered by their own cash flow.

“Even if the artificial intelligence market grows steadily, capacity will soon become overbuilt, debt will become worthless, and financial institutions will suffer losses,” Luria said.

The first dot-com bubble burst 25 years ago, Luria said, due, among other things, to debt financing to install fiber-optic cable for a future that wasn’t yet here. Learning from that lesson, tech companies don’t seem worried about it repeating itself.

“If you spend hundreds of billions of dollars on data centers and a few years later you don’t need them, you’re going to have another financial crisis,” he said.

Circular trading raises further concerns

Another aspect of the heated AI landscape that is raising eyebrows is the cyclical nature of investing.

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

Nvidia will funnel that money into OpenAI to fund its data centers. OpenAI will then equip these facilities with Nvidia chips. Some analysts say this structure, in which NVIDIA is effectively subsidizing one of its largest customers, artificially inflates the real demand for AI.

“My idea is that I’m Nvidia and I want OpenAI to buy more of my chips, so I’m going to give them money to do that,” Kedrosky said. “It’s pretty common on a small scale, but it’s unusual for it to be in the tens of billions or hundreds of billions of dollars,” he said, noting that the last time it was popular was during the dot-com bubble.

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

CoreWeave was once a cryptocurrency mining startup that pivoted to building data centers to ride the AI ​​boom. Leading AI companies turn to CoreWeave to train and run their AI models.

OpenAI has a contract with CoreWeave worth tens of billions of dollars in which CoreWeave’s chip capacity in data centers is leased to OpenAI in exchange for CoreWeave stock, which OpenAI can use to pay CoreWeave’s rental fees.

Meanwhile, Nvidia, which also owns part of CoreWeave, has a deal that guarantees it will buy up all unused data center capacity by 2032.

“The danger is that these kinds of deals end up revealing a house in the sand,” said Acemoglu, an economist at the Massachusetts Institute of Technology.

Some prominent investors believe the bubble is about to burst.

Some influential investors are showing signs of bubble anxiety.

Earlier this month, tech billionaire Peter Thiel sold all of his Nvidia stock worth about $100 million. This comes after SoftBank sold nearly $6 billion in Nvidia stock.

And in recent weeks, AI bubble pessimists have rallied around Michael Varley, a hedge fund investor who made hundreds of millions of dollars betting on the housing market in 2008. He is the subject of a 2015 film. big short. But since then, he has had a mixed reputation for his market predictions, warning of an impending collapse that never happened.

Unsurprisingly, Burley is now betting on NVIDIA, accusing the AI ​​industry of hiding behind a number of fancy accounting tricks. He is based in inter-company circular trading.

“True final demand is ridiculously small; almost all customers are financed by dealers,” Barry wrote of X. He then wrote, “OpenAI is the linchpin here. Can anyone name their auditors?”

As tech companies spend billions of dollars on data centers, some executives themselves freely admit that there seems to be too much momentum.

“Are we at a stage where investors as a whole are getting too excited about AI? My opinion is yes. Is AI the most important thing that’s going to happen for a very long time? My opinion is also yes,” OpenAI CEO Sam Altman told reporters in August.

And Google CEO Sundar Pichai recently told the BBC that there is “an element of irrationality” in the current AI market.

When asked what would happen to Google if the bubble bursts, Pichai said, “I don’t think any company, including us, will be unaffected.”