The debate rages on over whether AI is in a bubble, and top economist Jeremy Siegel thinks a less-remembered takeaway from the dot-com frenzy offers a clue about a budding risk to the current craze.
Siegel, a senior economist at WisdomTree and Professor Emeritus of Finance at the Wharton School of the University of Pennsylvania, is bullish on AI. However, he thinks there's still risk for companies committing enormous amounts of capex to pursue their ambitions in the space.
Speaking with CNBC, Siegel highlighted a key point: companies are spending unprecedented amounts of money to scale their AI programs, but it's possible that soon there will be much less expensive ways to power the technology.
"I think the biggest risk with AI isn't if it's going to work, I think it's going to work, but it can be done more cheaply," Siegel said.
He compared the current AI boom to the dot-com bubble, arguing that the most important takeaway from that era was that plummeting costs are among the top reasons the bubble ultimately popped.
"What really caused the bust was that they figured out how to put 1,000 times more data through the fiber optic cables they were laying," he noted. "All of a sudden, all the companies that were going to wire the universe said, you know what, we don't need that much."
Siegel raised the question of whether companies can continue scaling AI without spending billions of dollars on data centers. In his view, the answer is yes, and the result will be a poor return on investment for the company and ultimately investors.
Other commentators have begun to express worry about the size and scope of the AI investment boom. In a note on Monday, Neil Shearing, group chief economist at Capital Economics, made a similar argument about the risk of companies overextending themselves.
"The scale of current investment is comparable to past technological surges — from railways to the internet," Shearing said. "As history shows, booms like these almost always involve some degree of overreach. Some projects will fail to pay back, leaving investors nursing losses."
Still, the bullish mood and the optimism over the transformative power of AI endure among commentators. Shearing added that although many tech companies did not survive the dot-com bubble bursting, the fibre optic network that they left behind proved valuable. AI data centers may have the same impact.
"What may look like wasteful investment from a financial standpoint — and indeed is a waste for some firms, with consequences for their investors — can be an important part of the economic diffusion process," Siegel said.
