U.S. technology firms have announced that they will greatly increase capital expenditures (capex) in 2025 in the form of “massive AI spending.”[1] How should shareholders feel about this tidal wave of investment? Will it be spent productively, generating huge profits and soaring stock prices in coming years? Or will it be wasted, destroying shareholder value?
I don’t know if it’ll be wasted, but I do know that historically, high planned investment has been followed by low stock returns in subsequent years. This negative forecasting relation is true for both the aggregate stock market and for individual firms. Today’s high investment plans, like today’s high valuations, is a contrarian signal for U.S. megacap growth stocks.
Let me review the arguments for why high capex today suggests low returns going forward. We have many different theories in the academic literature, reflecting both rigorous rational modeling and loosey-goosey behavioral approaches.
Time-varying expected returns
One explanation for high capex is that expected returns on the U.S. stock market are low. Let me give two possible interpretations of what it means to have “low expected returns”:
- Behavioral: The U.S. stock market is currently overvalued, and therefore it will only earn 2% real returns going forward.
- Rational: The U.S. stock market is currently fairly valued given that the rational discount rate is 2%.
These two versions might sound different, but they both have the same implication that the cost of capital is low today. Both interpretations are consistent with the notion that if firms are rationally maximizing shareholder value, then they should do more capex when the discount rate is low. That’s the basic idea of Cochrane (1991). Sometimes discount rates are low, and firms invest a lot. Sometimes the discount rate is high, and firms invest less.
Thus as long as you believe that firms invest more when the discount rate is low, you should agree that high capex today predicts low future returns. Here’s Lamont (2000):
The result that higher discount rates cause lower investment is a minimal condition that should be satisfied by any rational model and by most irrational models.
The ideal measure of forward-looking returns is not past capex, but future expected capex. As shown in Lamont (2000), if your goal is to predict aggregate U.S. stock returns in the coming year, capex plans work better than trailing capex. Thus as of February 2025, we expect low stock returns over the next 12 months. You could say that stocks are “overpriced” or you could say they are “fairly priced to deliver low returns,” but either way the forecast is for below-average returns.
Issuance
According to the behavioral corporate finance view, capex might be a sign of overvaluation because capex is often financed by issuing equity. That is, if the stock price of company XYZ is too high, the management of XYZ might sell equity and spend the proceeds.
As discussed in Baker and Wurgler (2013): “market timing and attempted market timing play a considerable role in equity issuance decisions.” Generally speaking, firms are skilled market timers. The evidence suggests that both for individual stocks (see Daniel, Hirshleifer, and Sun (2020)) and the whole market (see Baker and Wurgler (2000)), high equity issuance today predicts low future returns. That’s why I call issuance the Third Horseman of the Bubble.
Right now, we have high capex, but not high issuance. We’ve seen no wave of IPOs, and existing U.S. publicly traded firms are still merrily distributing cash to shareholders via repurchases. That’s the bull case for the U.S. stock market: if the market is overvalued, why don’t we see issuance?
There are, however, some situations where high capex indicates overvaluation even in the absence of issuance, and I turn to those next.
Catering
If investors prefer firms that do X, then firms might do X in order to boost their stock price. Here’s Stein (1996):
… if the manager is interested in maximizing the current stock price, he must cater to any misperceptions that investors have. Thus if investors are overly optimistic about the prospects for the firm’s assets … the manager should be willing to invest very aggressively …
Catering can occur in the domains of capex, splits, firm naming, dividends, or anything else investors might take a fancy to. A recent form of catering involves the dubious practice of “bitcoin treasury, ” where firms invest money in bitcoin, as opposed to capex, in an attempt to cater to bitcoin-loving investors.
Looking at catering via investment spending at the individual firm level, Polk and Sapienza (2009) find that high capex predicts low subsequent returns, even when the company is not issuing equity. The implication for 2025 is that investors are exuberant about AI capex, therefore firms are doing AI capex to cater, therefore these firms are exuberantly overpriced.
Under this catering story, prices today are too high, and that fact does not necessarily hinge on whether the AI capex is wasted, or useful, or even whether it occurs at all. What matters is whether investors react to the planned capex by overvaluing the company.
Everyone’s overoptimistic
A different view is that the high level of capex today reflects general overoptimism about AI. Here, the reason we don’t see issuance as of 2025 is that both firms and investors are equally exuberant about the benefits of AI. Again, it’s not especially important whether the capex is wasted or not; it’s just a symptom of shared overoptimism.
Overoptimism can arise from extrapolation: corporate profits for big tech firms have grown dramatically in recent years. If (hypothetically) everyone wrongly expects that profit growth will continue at its current unsustainable pace, we’ll observe overvalued stock prices along with high capex. Here’s Gennaioli, Ma, and Shleifer (2016):
… expectations depart from rationality in the direction of being extrapolative: when CFOs or analysts observe good or bad earnings realization, they think that similar realization persists into the future and fail to correct for mean reversion.
They show that around 1999/2000, planned capex was very high, while expected profits were too optimistic compared to subsequent outcomes. In contrast, during the GFC in 2009, both planned capex and expected profits were very low.
Similarly, Arif and Lee (2014) find that times of high capex are times of high sentiment:
… corporate investments peak during periods of positive sentiment, yet these periods are followed by lower equity returns. … Higher aggregate investments also precede greater earnings disappointments, lower short-window earnings announcement returns, and lower macroeconomic growth.
The money will be wasted
The 1999/2000 bubble seemingly resulted in overbuilding. Here’s Doms (2004):
… the telecom service sector went on a capital expenditure binge … capital investment by publicly traded telecom service companies rose sharply in the late 1990s, starting at $47 billion in 1995 and peaking at $121 billion in 2000. Since then, however, the telecom service industry landscape is littered with the wrecks of overly optimistic expectations, as witnessed by the bankruptcies by WorldCom, Global Crossing, and numerous smaller firms.
Perhaps high capex plays a causal role in producing bad outcomes. Perhaps CEOs usually overspend, and when we see more spending, that just implies more waste. Richardson (2006) looks at firms with high free cash flow, and finds that they appear to overinvest. According to this story, the fact that firms are not issuing equity today is not especially comforting; thanks to high profitability, they don’t need to issue equity in order to waste money. Similarly, Titman, Wei, and Xie (2004) find that:
Firms that substantially increase capital investments subsequently achieve negative benchmark-adjusted returns … investors tend to underreact to the empire building implications of increased investment expenditures.
Perhaps today, firms are overspending but investors are oblivious, and thus stock prices today are too high.
A more complex story involves “competition neglect” in booms. If hyperscaler XYZ spends $100B on data centers, XYZ may neglect the fact that hyperscaler ABC is also spending $100B on data centers, and we end up with too many data centers. Historical accounts of asset bubbles often describe overbuilding: too many railroads, too many houses, and too many factories.
Here’s Greenwood and Hanson (2015) describing overbuilding due to competition neglect:
… heavy investment during booms predictably depresses future earnings and the price of capital, leading prices to overshoot their rational-expectations levels … particularly in markets—such as industries with long time-to-build delays—where feedback is delayed and learning is slow.
Even if firms are making rational calculations about spending, investors may not be. When firms compete for dominance, investors may suffer from what Cornell and Damodaran (2020) call "the big market delusion;” not every existing firm will win the race to dominate AI, and prices need to reflect that reality.
Let me conclude by giving my overall assessment. Does today’s high planned capex mean the market is dangerously overvalued? I think not. Maybe the market is indeed dangerously overvalued, but I think that high planned capex is only a mildly bad sign taken in isolation. Capex is a negative predictor of future returns, but I wouldn’t put it at the top of my list of super-reliable signals. On the one hand, today’s high planned capex is part of a mosaic of information suggesting the market is overvalued today. On the other hand, it has not been accompanied by a wave of equity issuance. So on balance, I say the massive AI spending is only a yellow flag for the stock market, not a red flag.
Endnote
[1] “Tech Giants Double Down on Their Massive AI Spending.” The Wall Street Journal, February 6, 2025.
References
Arif, Salman, and Charles M.C. Lee. "Aggregate investment and investor sentiment." The Review of Financial Studies 27, no. 11 (2014): 3241-3279.
Baker, Malcolm, and Jeffrey Wurgler. "The equity share in new issues and aggregate stock returns." The Journal of Finance 55, no. 5 (2000): 2219-2257.
Baker, Malcolm, and Jeffrey Wurgler. "Behavioral corporate finance: An updated survey." In Handbook of the Economics of Finance, vol. 2, pp. 357-424. Elsevier, 2013.
Cochrane, John H. "Production‐based asset pricing and the link between stock returns and economic fluctuations."The Journal of Finance 46, no. 1 (1991): 209-237.
Cornell, Bradford, and Aswath Damodaran. "The big market delusion: Valuation and investment implications." Financial Analysts Journal 76, no. 2 (2020): 15-25.
Daniel, Kent, David Hirshleifer, and Lin Sun. "Short-and long-horizon behavioral factors." The Review of Financial Studies 33, no. 4 (2020): 1673-1736.
Doms, Mark. "The boom and the bust in information technology investment." Economic Review-Federal Reserve Bank of San Francisco (2004): 19-34.
Gennaioli, Nicola, Yueran Ma, and Andrei Shleifer. "Expectations and investment." NBER Macroeconomics Annual 30, no. 1 (2016): 379-431.
Greenwood, Robin, and Samuel G. Hanson. "Waves in ship prices and investment." The Quarterly Journal of Economics 130, no. 1 (2015): 55-109.
Lamont, Owen A. "Investment plans and stock returns." The Journal of Finance 55, no. 6 (2000): 2719-2745.
Polk, Christopher, and Paola Sapienza. "The stock market and corporate investment: A test of catering theory." The Review of Financial Studies 22, no. 1 (2009): 187-217.
Richardson, Scott. "Over-investment of free cash flow." Review of Accounting Studies 11 (2006): 159-189.
Stein, Jeremy C. "Rational capital budgeting in an irrational world." Journal of Business (1996): 429-455.
Titman, Sheridan, KC John Wei, and Feixue Xie. "Capital investments and stock returns." Journal of Financial and Quantitative Analysis 39, no. 4 (2004): 677-700.
