A former high-ranking official from the U.S. Federal Reserve, often referred to as the institution's number three figure, has issued a stark caution to investors caught up in the global excitement surrounding artificial intelligence stocks. Bill Dudley, who previously served as the President of the Federal Reserve Bank of New York from 2009 to 2018 and was once a partner at Goldman Sachs, stated bluntly that the current AI investment wave has reached its zenith. He predicts that the speculative bubble could burst before the conclusion of 2027.

Dudley, who now sits on the advisory board for Coinbase Global, pointed to several key indicators suggesting the U.S. stock market is currently in bubble territory. He noted that the Shiller CAPE ratio, a cyclically adjusted price-earnings ratio, stands at 41.96, which is perilously close to the 44.19 peak witnessed during the dot-com bubble in December 1992, and significantly higher than the long-term average of roughly 17.4. Furthermore, the current real equity risk premium, which measures the expected return of holding stocks compared to inflation-indexed government bonds, is a mere 1.1%. This figure is less than half of the average seen since 2010. He also highlighted that the "Buffett Indicator," which compares the total stock market capitalization to GDP, has soared to 240%, a level far exceeding the threshold of 100% that investing legend Warren Buffett considers overvalued.

While these statistics are stark, the underlying risk factors accumulating behind the market's rise are concrete and could potentially trigger a collapse by the end of 2027. Dudley elaborated on five critical elements that could cause the equity bubble, particularly in the AI sector, to deflate. The first factor is a significant slowdown in AI investment growth. He argues that the positive impact of the AI spending boom on economic activity and corporate earnings is likely to diminish considerably by 2027. Investment growth in 2026 will almost certainly be the peak, as by the following year, the availability of construction workers, power generation capacity, and chip manufacturing capability will be insufficient to sustain an equivalent expansion. Additionally, the dominant hyperscale data center operators may lack the necessary free cash flow and balance sheet capacity to fund even more aggressive investment.

Second, earnings growth is set to face a dual challenge. As the pace of investment spending decelerates, the earnings growth of hyperscale data center suppliers will cool down, leading to reduced earnings expectations and lower price-to-earnings multiples. These companies, often described as the "picks and shovels" providers, will encounter the combined pressure of slowing demand growth and shrinking profit margins. Third, the return on investment is questionable. As the investment cycle matures, market attention will shift to whether hyperscale data center operators can generate sufficient returns from their massive expenditures. Dudley believes it will be difficult for these operators to produce the $2 trillion or more in annual revenue necessary to justify an AI capital base that could reach $5 trillion. Simultaneously, the risks associated with loans funding data centers and AI infrastructure are escalating, with financing mechanisms becoming increasingly complex and opaque.

The fourth factor is an increase in stock supply. A surge in initial public offerings, such as the anticipated listings of Anthropic and OpenAI, coupled with insider selling after lock-up periods expire, will add to the supply of U.S. equities, placing downward pressure on market valuations. The fifth element is a tightening macroeconomic environment. Both real and nominal long-term interest rates have climbed significantly this year, with the 30-year Treasury yield reaching its highest level since 2007. This puts further strain on equity valuations. Given the lack of political will and progress in addressing the unsustainable growth of U.S. federal debt, the risk of yields rising further is substantial.

Dudley drew parallels between the current AI craze and historical market booms and busts. He explained that in the early stages of a bubble, growth can be self-reinforcing, as demand from the investment boom supports rapid profit growth and rising stock valuations. However, this feedback loop can reverse violently, where a collapse in demand leads to reduced cash flow and a reassessment of lending risks, which in turn hampers further expansion. He referenced the subprime lending boom before the 2008 financial crisis to illustrate this mechanism. During that period of market exuberance, innovative financial instruments fueled housing demand and drove up prices, making loans appear low-risk. But when demand growth met increased supply and housing price appreciation stalled, refinancing became difficult, delinquency and default rates soared, ultimately triggering a full-scale crisis.

Dudley stated his expectation that the trajectory of artificial intelligence will follow a similar pattern to other major tech industry boom-and-bust cycles. He noted that like the railway and internet booms, AI will have a significant long-term impact on productivity and economic growth. However, overcapacity is inevitable and will pressure profits and share prices, potentially converting the investment boom into a downturn more rapidly than expected.

To drive his point home, Dudley cited two contrasting historical examples. Jeremy Grantham of GMO Investment warned about the Nasdaq bubble in the late 1990s. Although his prediction was premature, which caused his firm's assets under management to shrink by roughly a third, he was ultimately vindicated. Grantham emerged from the dot-com crash unscathed with an enhanced reputation. In stark contrast is the fate of Chuck Prince, the former CEO of Citigroup. In 2007, Prince famously remarked that "as long as the music is playing, you've got to get up and dance." By November of that year, he was forced to resign as Citigroup suffered $6.5 billion in losses due to the subprime mortgage crisis. He was later listed by Fortune magazine as one of the executives who failed to foresee the financial crisis. Dudley uses these examples to remind investors that while resisting the temptation of a growing bubble is difficult, heeding early warnings is far preferable to being caught off guard when the music stops.