The Great AI Bubble: The Start of a Major Bear Market?

What follows is a translated summary of Owen Lamont’s views on the AI bubble. Lamont is an Executive Vice President and portfolio manager at Acadian Asset Management, a Boston-based boutique hedge fund.

U.S. big tech companies have announced that they will undertake “massive AI spending” in 2025. Shareholders are left wondering whether to cheer this news or view it with concern. No one knows yet whether it will lead to explosive future profits and soaring stock prices, or turn into pouring money into a bottomless pit.

Looking back at historical data, periods following announcements of large-scale investment plans have often been followed instead by low equity returns, both at the market level and for individual firms. When, as now, both investment plans and valuations for U.S. mega-cap growth stocks are elevated at the same time, it strengthens the case for a more cautious stance.

In academia, one interpretation is that high corporate capital expenditures (CAPEX) imply low expected returns.

Both interpretations ultimately lead to the same outcome: “because the current cost of capital (discount rate) is low, firms invest aggressively” (Cochrane 1991). In other words, when discount rates are low, investment rises, and then in the following year stock returns tend to be low—a pattern that repeats (Lamont 2000).

There is even research showing that “planned future capital expenditures” are a better predictor of future stock prices than “capital expenditures already incurred.” From that perspective, the wave of large investment plans being announced as of February 2025 can be read as signaling low market returns over the next 12 months.

Behavioral finance views high capital expenditure as a potential “overvaluation signal.” When stock prices are excessively high, firms have a strong incentive to issue new shares and use the proceeds to fund investment (Baker & Wurgler 2013). Historical data indeed show that “firms or markets that issue a lot of equity today tend to deliver low returns in the future” (Daniel, Hirshleifer, and Sun 2020; Baker & Wurgler 2000).

Today, however, there is no notable IPO boom, and existing listed companies are focused on share buybacks. This opens the door to a counterargument: “If the market is truly overvalued, why aren’t we seeing a rush of equity issuance?”

Even so, there is a reason why high CAPEX can still point to market overvaluation even in the absence of equity issuance.

When investors favor a particular trend, firms tend to follow that trend aggressively in hopes of boosting their stock prices (Stein 1996). If, as today, AI investment is being cheered by the market, companies can sharply ramp up spending even without actually issuing new shares. Polk and Sapienza (2009) find that “even when firms spend money without issuing equity, high CAPEX is followed by low subsequent returns.”

In other words, the logic that “investors get excited about AI investment → firms respond by increasing AI CAPEX → the stock prices of those firms become excessively inflated” can also apply to the situation in 2025.

Another scenario is that both firms and investors are simultaneously overly optimistic about the AI future, leading to vigorous investment even without equity issuance. With big tech profits having surged recently, an unfounded belief can take hold that “this growth will persist indefinitely” (Gennaioli, Ma, and Shleifer 2016).

In such a case, whether CAPEX is being deployed efficiently becomes secondary. What matters is that “misplaced optimism” is what simultaneously produces today’s high stock prices and heavy capital spending (Arif & Lee 2014).

We can recall the dot-com bubble era, when the telecommunications services industry went bankrupt in droves after pouring money explosively into infrastructure (Doms 2004). When firms have abundant free cash flow, they are prone to overinvest (Richardson 2006), and in such cases investors often underestimate the problems caused by “empire building,” leading them to price stocks too high (Titman, Wei, and Xie 2004).

Ignoring competition during boom times also fuels “overbuilding” (Greenwood & Hanson 2015). Each company thinks, “If we invest first, we can secure the lead,” but at the industry level this can easily result in excessive, overlapping investment.

There is ample evidence that high CAPEX like this sends a negative signal about future stock returns. Still, it seems a stretch to declare the current situation a “fully red light” where the market has clearly crossed into the danger zone.

For this reason, I see the current “large-scale AI investment” as more of a “yellow light” for the market. It is too early to call it a clear signal of an imminent crash, but it does suggest that some degree of correction may be warranted.

What really matters is whether this investment can actually translate into higher corporate earnings and stronger fundamentals. More than the uncertainty itself, how that uncertainty is perceived and managed will determine where the market goes from here.