Fears of an AI bubble are growing
CFOTO/SIPA United States/ALAMY
The hundreds of billions of dollars spent on AI appear to have inflated a global financial bubble that is now ready to burst, leaving companies and investors at risk of incurring massive debt that cannot be serviced by the meager revenues generated by current AI services. But what does this mean for the future of the technology behind this financial frenzy?
In recent weeks, the International Monetary Fund, the Bank of England, the president of the largest US bank and even OpenAI boss Sam Altman have warned of a potential AI bubble. “It’s not just a stock market bubble, but also an investment bubble, but also a public policy bubble,” says David Edgerton of King’s College London.
The circular nature of some agreements between major AI players also raises eyebrows. For example, Nvidia, which makes the GPU chips that are powering the AI boom, recently invested up to $100 billion in OpenAI so that the company could build a new data center filled with Nvidia’s own chips. OpenAI, in turn, struck a deal that could ultimately allow it to take a 10% stake in Nvidia’s rival chipmaker AMD.
The concern over the bursting of the AI bubble is also highlighted when we measure its scale: at least $400 billion is spent in data centers each year, according to Morgan Stanley Wealth Management. And while U.S. GDP grew 3.8 percent in the second quarter of the year, Harvard University’s Jason Furman estimates that if data centers were taken out of the equation, growth would have been just 0.1 percent for the entire first half.
Carl-Benedikt Frey of the University of Oxford says these kinds of exuberant deals are not unusual in the history of technology. In fact, it would be unusual for the global economy to manage to invest in infrastructure for a new technology at precisely the right pace to meet demand. “It’s pretty common to overbuild: the same thing happened with the rail boom, the same thing with the dot-com bubble,” he says.
The question is whether the consequences of an AI bubble would simply harm the companies involved or could have broader impacts. Frey points out that many of these extremely expensive data centers are actually built “off balance sheet.” This involves the creation of new companies backed by external investors or banks that build and own the assets, assuming both the risks and potential rewards.
Therefore, we do not know enough about who is at risk. A data center could be financed by a dozen tech billionaires, or by big banks – and if their losses are big enough, then a banking crisis could send shockwaves through the entire economy. “That doesn’t mean there’s a financial crisis looming, but that it’s a little opaque. And when things are opaque, there’s usually a risk,” says Frey.
Benjamin Arold of Cambridge University says the telling factor is the relationship between profits and company valuations, which indicates how disconnected public opinion is from the real money companies make. He believes these numbers for today’s tech companies are a wake-up call.
“The last time it was this low was 25 years ago, and if you remember, 25 years ago we had the dot-com bubble,” says Arold. “It’s possible it will go well, but I wouldn’t bet my money on it.”
James Poskett of the University of Warwick, UK, believes we’re heading for a correction in the AI sector that could mean the end of many companies, but he says it’s certainly not the end of the technology itself. “It’s important not to confuse this with the idea that the technology is broken or will disappear,” Poskett says. “There might be an AI failure, but that doesn’t mean we won’t have AI.”
Just as the consolidation of many railroad companies after bankruptcy left us with a rail network, and the collapse of technology companies during the dot-com bankruptcy left us with a legacy of vast fiber-optic networks, we will end up with useful technology, Poskett says.
For consumers, the bursting of the AI bubble will likely mean a little less choice, perhaps a slightly higher price of access, perhaps a slower pace of updates. This might force us to face the reality that using an extremely expensive tool like GPT-5 to write an email is like using a hammer to crack a walnut, and that the true cost of using it had previously been hidden by the frenzied AI arms race. “Right now there’s a lot of free stuff, but at some point these companies have to make a profit,” Poskett says.
Topics:
