Bill Dudley, former President of the Federal Reserve Bank of New York, wrote in a commentary on Thursday that US stock market valuations have clearly entered bubble territory. Should the AI investment cycle slow down, the earnings, profit margins, and financing logic currently supporting the equity rally could reverse simultaneously. He projects that this bubble could burst before the end of 2027.
Dudley first pointed out that multiple indicators have already revealed the high valuations in US stocks: the Shiller Cyclically Adjusted Price-to-Earnings Ratio (CAPE) currently stands at around 41 times, far above the long-term average of approximately 17 times and only slightly below the all-time high of 44 times recorded in December 1999 during the dot-com bubble. The real equity risk premium is roughly 1.1%, less than half of the average since 2010. Meanwhile, the "Buffett Indicator" — the ratio of total US stock market capitalization to GDP — has reached approximately 240%, far exceeding the level above 100% that Warren Buffett previously deemed as already expensive.
However, Dudley emphasized that high valuations do not mean the bubble will burst immediately. Bubbles often continue to inflate because rising asset prices in turn stimulate investment and earnings growth, while reinforcing the optimistic expectations of market participants. The risk lies in the possibility that this positive feedback loop could eventually reverse.
AI Investment: A Turning Point May Arrive in 2027
Dudley believes the most critical inflection point stems from AI capital expenditure. The impact of AI investment on the economy and corporate earnings depends on the incremental increase in investment rather than its absolute scale. Investment growth in 2026 is likely to peak, and by 2027, resource constraints such as construction workers, electricity supply, and chip manufacturing capacity will make it difficult for companies to continue expanding investment at the same pace. At the same time, the free cash flow and balance sheets of major hyperscalers may not be sufficient to support investment growth in 2027 that exceeds that of 2026.
Once the growth rate of AI capital spending decelerates, upstream "picks and shovels" companies will feel the pressure first: slowing demand growth, declining earnings expectations, and potentially shrinking profit margins, ultimately resulting in a "double blow" to both valuations and earnings. On a deeper level, the market will eventually shift its focus from "how much money is being invested" to "how much money these investments can generate." Dudley argues that AI giants will need to generate US$2 trillion or more in annual revenue in the future to provide sufficient returns on a potential US$5 trillion AI capital base. As investment in data centers and other infrastructure expands, the associated financing is becoming increasingly complex, interconnected, and opaque. Chip makers like Nvidia (NVDA) have even begun offering financing to cloud service providers, which could create a new circular financing structure that amplifies negative feedback once the AI boom reverses.
Multiple Triggers Could Spark the Bubble's Collapse
Beyond the AI investment cycle, Dudley also identified several other pressures. First, the supply of US stocks is increasing. The number of IPOs has risen noticeably, and as insider lock-up periods expire, increased stock sales could further depress valuations. Second, the macroeconomic environment is deteriorating. Both real and nominal long-term interest rates in the US have risen significantly this year, with the 30-year Treasury yield at one point reaching its highest level since 2007. Higher risk-free rates directly compress stock valuations, and the persistent lack of a political solution to the US federal debt problem means long-term yields still face further upward risk.
Therefore, once AI investment slows down, the positive feedback loop of "investment growth — earnings growth — margin expansion — valuation appreciation" could quickly reverse into "declining demand — deteriorating cash flow — rising financing risk — valuation contraction." History offers a warning. Dudley used the 2008 global financial crisis to illustrate this mechanism. Subprime mortgage expansion initially drove housing demand and home prices higher, and rising prices made it easier for borrowers to avoid default through refinancing, further reinforcing the market's belief that subprime risk was manageable. But as housing supply increased and prices stopped rising, refinancing channels narrowed, default rates surged, and the subprime financing system collapsed, which in turn hit housing demand and prices.
Dudley believes AI could follow a similar path: the technology itself will deliver long-term productivity gains, but the investment boom will inevitably create excess capacity, and that overcapacity will ultimately erode corporate profits and stock prices. He therefore concludes that AI, like the railway and internet investment booms before it, may have a positive long-term impact, but the short-term prosperity will ultimately translate into overinvestment and market correction.
Exiting Early During a Bubble Era Is the Hardest Challenge
Dudley finally reminded readers that the truly difficult task is not identifying a bubble, but choosing to exit while it continues to inflate. In the late 1990s, legendary investor and GMO co-founder Jeremy Grantham issued an early warning about the Nasdaq bubble. Although his judgment ultimately proved correct, his premature bearish stance caused his firm's assets under management to shrink by approximately one-third. In contrast, Chuck Prince, former CEO of Citigroup, said in 2007 that "as long as the music is playing, you've got to get up and dance." The global financial crisis subsequently proved that dancing along with the bubble may yield short-term gains, but once the music stops, the cost could far exceed that of leaving early. Dudley believes the greatest risk of the current AI boom lies precisely in the market mistaking the growth generated by a short-term investment frenzy for a long-term profitability that can be sustained indefinitely.
Comments