You may remember the recession that followed the collapse of dot-com stocks in 2001. Or, worse, the housing crisis of 2008. Both times, a new idea — the internet, mortgage-backed securities and the arcane derivatives they unleashed — convinced investors to plunge so much money into the stock market that it inflated two speculative bubbles whose inevitable bursting created much economic pain.
We believe it’s time to call the third bubble of our century: the A.I. bubble.
While no one can be certain, we believe this is more likely the case than not. Investment in artificial intelligence has been so huge — with venture capitalists investing nearly $200 billion in the sector this year alone. Additionally, data-center investment has tripled since 2022. Together, these investments are driving growth across the entire economy, pumping up the stock market and generating increasingly eye-popping valuations of the technology firms driving the A.I. revolution.
In financial markets, a bubble occurs when the level of investment in an asset becomes persistently detached from the amount of profit that asset could plausibly generate. While investors are always making bets on an unknown future, bubbles form when large swaths of investors continuously pour ever more into an asset, with seemingly little regard for how much it could earn and when.
A.I. investment fits that pattern. OpenAI says it needs at least $1 trillion to invest in data centers that provide the electricity, computing power and storage to train and run A.I., yet the company’s revenues are expected to amount to a mere $13 billion this year. And since the debut of ChatGPT, an easily accessible A.I. chatbot, in late 2022, the S&P 500 has swelled by nearly two-thirds, with just seven firms — all of whom have invested heavily in A.I. — driving more than half of that growth.
Or take a look at the price-to-earnings ratio — a common measure of how much the future profits of a company are valued over current ones — of the stocks of companies heavily invested in A.I. They are at levels not seen since the dot-com bubble of 2000. Shares of the A.I. chipmaker Nvidia are trading at roughly 55 times earnings, nearly double what they were a decade ago. And by our own estimates, the share of the economy devoted to A.I. investment is nearly a third greater than the share of the economy devoted to internet-related investments back then. All this points to one conclusion: Should lackluster A.I. performance or sluggish adoption cause investors to doubt these lofty profit expectations, this probably-a-bubble will pop. And a lot of people, not just wealthy investors, will get hurt. Adoption, both by firms and individuals, is clearly growing, but whether this adoption is generating massive productivity benefits or profits remains to be seen.
Of course, we cannot rule out the possibility that this time is different, and unlike the railroad and internet bubbles, A.I. is an epoch-shifting technology that generates its promised economic benefits relatively quickly. If that occurs, say, over the next five to 10 years, the future profits generated by A.I. could justify the levels of investment we’re observing today (it was in this spirit that Microsoft’s chief executive, Satya Nadella, recently said, “I hope we don’t take 50 years”). It is also impossible to know when we’re at the top of a bubble, which is one reason investors tend to keep piling in.
But we’re skeptical. Look at what happened with the internet. In the dot-com bubble of the late 1990s, hype around that revolution allowed companies like pets.com to raise over $80 million in an initial public offering, despite the fact that its business model, which involved spending too much money to sell unprofitable pet supplies, was questionable at best. Less than nine months after their I.P.O., the firm went bankrupt — and many other busts soon followed. The belief that the internet would become a transformative technology was eventually correct, but investors during the dot-com bubble were wrong about the winners and their timing.
The economic impact generated by a bursting of the A.I. bubble would be greater than the loss of the trillions currently being invested to build the technology itself. The stock market, one of the brightest parts of the current economy and heavily dependent on A.I. ebullience, would also tumble. That, in turn, will diminish the “wealth effect,” or the way that stock market gains support consumer spending.
Using data from the economist Mark Zandi, we found that over the past two years, real consumer spending is up 17 percent for the wealthiest households, who disproportionately hold stocks, but flat for the middle class. Mr. Zandi estimates that the A.I. wealth effect is boosting current real gross domestic product growth by about 0.4 percentage points (just under $100 billion), comparable to the peak of the dot-com bubble, when the wealth effect was 0.6 percentage points.
There is a bit of a silver lining. As best we can tell, the damage of a potential A.I. bubble would not approach the carnage that resulted from the burst of the housing bubble and the financial crisis of 2007 and 2008. While banks, private credit and private equity are all lending heavily to companies that are building and leasing A.I. data centers, this debt appears less distributed and embedded in global finance than it was back then. Plus, the risks are not obviously or systemically underpriced, a factor that played a key role in spreading the contagion across the globe during the housing bubble. Prominent borrowers in this space, like CoreWeave, are paying 9 percent on their debt, well above the current risk-free rate on 10-year Treasuries of around 4 percent.
Our economy faces real risks. If A.I. is in a bubble, and its valuations relative to its expected payouts start to alarm investors, the bubble will burst. The ensuing wealth losses and consumer spending impact could, once again, be recessionary, though there’s a good chance the impact won’t be nearly as bad as the last bubble. Granted, that’s not good news. But it could be worse.
Jared Bernstein is a distinguished policy fellow at the Stanford Institute for Economic Policy Research, where Ryan Cummings is the chief of staff.
The Times is committed to publishing a diversity of letters to the editor. We’d like to hear what you think about this or any of our articles. Here are some tips. And here’s our email: [email protected].
Follow the New York Times Opinion section on Facebook, Instagram, TikTok, Bluesky, WhatsApp and Threads.
The post A.I. Sure Looks Like a Bubble. Watch Out When It Pops. appeared first on New York Times.
