As artificial intelligence companies pour hundreds of billions of dollars into data centers, fears are growing that the industry is inflating a financial bubble that could harm the economy.
AND MARTINEZ, GUEST:
Tech companies are pouring hundreds of billions of dollars into artificial intelligence data centers, and the big question is whether all that money represents a financial bubble. NPR's Bobby Allyn watched. So, Bobby, bubble or no bubble?
BOBBY ALLYN, BYLINE: That's the question, isn't it? And it really depends on who you ask, right? So to tech executives and investors, who of course have the most to gain from this momentum continuing, they say no, absolutely not. There is no bubble visible here. But for the financial analysts, economists and other experts I spoke with, the situation is quite alarming.
Let's start with the money, right? Amazon, Google, Meta and Microsoft are set to spend $400 billion on AI this year alone, and that's mostly on data centers. OpenAI says it wants to spend more than $1 trillion on data centers in the coming years. But the catch is that most AI chatbots, including ChatGPT and the rest, do not make money. An analysis by Menlo Ventures found that only 3% of customers pay for AI. And OpenAI, a leading artificial intelligence company, doesn't expect to be profitable in the next five years. Therefore, some view this fountain of money as a huge overinvestment. There's also the technology itself, which has all but stopped, according to Paul Kedroski. He is a venture capitalist who researches artificial intelligence at the Massachusetts Institute of Technology.
PAUL KEDROSKI: The technology is very useful, but the pace of improvement has more or less slowed down. So the idea that the revolution will continue with the same drum music for the next five years is unfortunately wrong.
MARTINEZ: Okay. So how do tech companies protect their costs?
ELLIN: Yes. Silicon Valley proponents say AI is a once-in-a-generation technology. They compare it to the early Internet or even the railroad system. And they say all this infrastructure is necessary to remain competitive in the global market, especially with China. White House AI czar and venture capitalist David Sachs recently wrote on X that slowing AI spending could even lead to a recession, writing, quote, “We can't afford to go backwards.” But if you look under the magnifying glass A at the details of some of these expenses, there are many reasons for doubt.
MARTINEZ: Oh. So what are the doubts? I mean, what about expenses is questionable?
ELLIN: Yes. Mainly two things. The first is what is known as circular costs. We see a lot of this. One company gives another company billions of dollars to buy their own products, usually semiconductor chips. Investments like this have been happening for some time, but we've never seen them happen on a scale of hundreds of billions of dollars. There is concern that these types of deals essentially artificially inflate the real demand for AI. The second problem is the so-called special-purpose vehicles. Big tech companies are teaming up with Wall Street giants to build these data centers, and the debt they use doesn't end up on the tech companies' balance sheets. Analyst Gil Luria of investment firm DA Davidson told me that the industry currently faces up to $100 billion in debt to build AI-powered data centers.
GILE LURIA: At the moment, if the artificial intelligence market grows at least steadily, then quite quickly we will have outgrown capacity. And that debt will become worthless and those financial institutions will lose money.
ELLIN: Yes. And Luria says that this situation will not only be a problem for investors, but will affect all of us, right? The stock market could fall and this could hit pensions, superannuation funds and the labor market. Borrowing may become more difficult and expensive. So if this happens, A, the economic consequences will be widely felt.
MARTINEZ: This is NPR's Bobby Allyn. Thank you very much, Bobby.
ALLYN: Thank you, A.
© 2025 NPR. All rights reserved. Visit our website terms of Use And permissions pages in www.npr.org for more information.
The accuracy and availability of NPR transcripts may vary. Transcript may be modified to correct errors or accommodate audio updates. Audio on npr.org may be edited after it is originally broadcast or published. The authoritative recording of NPR programs is the audio recording.
