📲 AI bubble: Could it burst?

A 2026 paper from Brookings calculates that expanding US data-center capacity could require as much as $10.3 trillion in investment by 2032.

One of the study's central assumptions is that infrastructure would have to bring in roughly $3.7 trillion a year to generate a 10% unlevered return.

Experts caution that a single serious safety incident, a failed IPO, or weaker-than-expected AI revenues could trigger a significant market shock.

The AI spending story could unravel in several ways:

🔴 Revenue may grow too slowly to justify the enormous capital outlays, leaving data-center operators with returns that fail to cover costs

🔴 Competition could compress margins across the industry, eroding the profitability that investors have priced in

🔴 The buildout may outpace actual demand, leaving swaths of computing capacity idle

If spending slows, chipmakers, cloud providers, and investors could all absorb losses, with the shock rippling through equity and credit markets.

History offers three warnings:

🔶 The Nifty Fifty of the 1970s: fifty blue-chip American companies, including Coca-Cola, IBM, and Xerox, traded at price-to-earnings ratios of 80 to 90 on the belief that their growth would never falter, only to lose 60% to 90% of their value in the 1973–1974 bear market — proof that even excellent companies become disastrous investments when bought at excessive prices

🔶 Japan in the late 1980s: property and stock prices soared, then collapsed, leaving damaged balance sheets and years of economic stagnation in their wake.

🔶 The dot-com crash of 2000–2002: the internet revolution was real, yet the Nasdaq still lost roughly 77% from its peak. Technological progress offered investors no protection from overvaluation.

👍 Substack | Chat | @geopolitics_prime