The S&P 500's cyclically adjusted price-to-earnings ratio has climbed to 40.5, a level exceeded only twice before in over a century of market history. Both prior instances — 1929 and 1999 — preceded major market declines, and the current run-up is being driven by generative AI spending that one Wall Street Journal study says may not be economically justified.
The S&P 500 slipped 0.22% to 7,801.77 on the day, a pullback that sits against a much larger backdrop: the index's cyclically adjusted price-to-earnings (CAPE) ratio has reached one of only three elevated extremes in more than 156 years of data. History shows that significant drawdowns often follow such periods, as the market reverts toward its long-running mean.
A valuation metric flashing a rare warning
The CAPE ratio compares the S&P 500's price with its average inflation-adjusted earnings over the past decade, smoothing out short-term swings to show how the market is priced against historical norms. The CAPE ratio now stands at 40.5, well above its long-term average of 17.4.
Only two prior peaks reached this scale: before the Great Depression in 1929 and during the dot-com bubble in 1999, when the ratio hit its all-time high of 44.19. Both episodes were followed by substantial declines in equity prices as the speculative run-ups deflated. The current generative AI boom shows strong similarities to the dot-com era, driven by a transformational technology that may not pay off as quickly as its biggest backers expect.
The AI spending question
The generative AI boom looks safer than the dot-com bubble in one respect: it's being driven by stable, profitable companies rather than speculative start-ups. Valuations also look reasonable on the surface, with Nvidia and Micron Technology carrying forward price-to-earnings multiples of just 25 and 6, respectively, because their revenue is growing faster than their stock prices.
But that revenue increasingly comes from a source that may not be sustainable. Goldman Sachs estimates hyperscalers could spend $800 billion on AI-related capital expenditures in 2026 alone, with that figure potentially rising to $1.4 trillion by 2028 if current trends continue. Those funds could otherwise flow back to shareholders through buybacks or dividends, and investors may eventually start pressuring management teams to spend more conservatively.
A Wall Street Journal study found that American businesses and consumers would need to spend the equivalent of 8.8% of the country's GDP annually on AI to justify the industry's spending, raising the possibility that many current data center investments never pay off. Cloud computing giant Oracle illustrates what happens when shareholders lose patience: its shares have fallen 50% over the last 12 months alone.
What comes next for investors
Timing the stock market is notoriously difficult, since even a correctly identified problem can take months or years before the broader market reacts to it. Rather than selling everything, investors should instead focus on taking some profits off the table and diversifying away from stocks heavily exposed to generative AI spending.
Source: The Motley Fool
Trading involves risk.
Risk Disclaimer
