The artificial intelligence investment boom that has driven U.S. stocks to record highs is showing signs of strain, according to economists at Capital Economics. The research firm now estimates the bubble will begin to burst in 2027, with a correction of at least 20% in the S&P 500 likely arriving next year.
"There are plenty of signs that we are now in the late stages of a bubble in AI," John Higgins, chief economic adviser for financial markets at Capital Economics, said in a report Monday. The warning comes as global capital expenditures on AI-related projects are projected to reach $1 trillion in 2026, including $581 billion in the U.S., according to Goldman Sachs.
The earnings expectations problem
Capital Economics senior markets economist James Reilly drew a direct comparison to the dot-com era. "If you look at the rate at which leading AI firms' earnings are expected to grow, versus how fast the U.S. economy has been growing, by many measures they look really stretched - as in this is the dot-com bubble all over again," Reilly said. He added that while AI will likely be transformative and generate profits, the firm does not believe returns will match current analyst expectations.
The two-year stock rally has been fueled by investors betting heavily on robust future profits from AI leaders. But the gap between those expectations and economic reality is widening, the firm's analysis suggests.
Why bubbles are hard to call
Other economists urge caution about labeling the current moment a bubble. Gregory Daco, chief economist at EY-Parthenon, pointed out that technological revolutions typically involve heavy upfront investment phases. "We have to be very careful with the bubble terminology, specifically, because every type of technological revolution tends to have a great dose of investment in the first phase," Daco said. "But at the same time, there are often excesses. There is often exuberance because a new technology is very attractive and promises to revolutionize the way we do things."
Kenneth R. French, an investment strategist at Dartmouth College's Tuck School of Business, takes a different view. He argues that investors routinely pile into hot tech stocks long before a technology's full economic impact becomes clear. "People get optimistic, and it's conceivable that five years from now, we'll be looking back and saying people were pessimistic about AI, and that it was more important than we expected," French said. "It's already having a huge impact on earnings, so it could really be that we have underestimated AI's positive impact."
French was blunt about the limits of current analysis: "We don't have enough information to judge if these prices are right or wrong, too high or too low."
A separate risk: guardrails, not returns
The public conversation around AI is shifting as researchers and corporate leaders issue warnings about the technology's risks. Those concerns, combined with calls from industry figures for slower AI development, could weigh on tech stocks. But Daco drew a distinction between fears about AI safety and fears about investment returns. "That's slightly different than a bubble fear," he said. "It's more fear of not having the right guardrails to control tech and avoid excesses of the tech itself - not about investments and returns."
Why this matters for finance professionals
A potential AI-driven correction in the S&P 500 would hit portfolios across sectors, not just tech. Capital Economics is putting a timeline on it: a pullback next year, with the bubble deflating fully by 2027. The debate among economists also underscores a practical problem - no reliable method exists to confirm a bubble in real time. For investment committees and advisors, the immediate question is whether current AI exposure aligns with a client's actual risk tolerance, not whether the technology will eventually deliver on its promise.
Your membership also unlocks:
