The AI Gold Rush: Billions in Debt, Uncertain Returns
The artificial intelligence revolution is unleashing one of the largest capital expenditure booms in modern history, but beneath the surface lies a growing mountain of debt that has investors and credit markets increasingly nervous. According to analysis from BCA Research, the current AI investment frenzy shows striking parallels with historical capital-expenditure cycles that typically ended in painful busts.
The hundreds of billions being spent to build and equip data centers have become a central focus for both equity and credit markets in recent weeks. What makes this boom particularly concerning for fixed-income investors is the fundamental mismatch between risk and reward. Unlike equity investors who can enjoy unlimited upside from successful bets, bondholders face capped returns with potentially total loss exposure.
Credit Markets Sound Alarm Bells
The complexity and scale of recent AI infrastructure deals are testing even the most experienced credit analysts. Dan Fuss, the 92-year-old vice chairman of Loomis Sayles often called the "Buffett of bonds," expressed serious concerns about current data-center financing arrangements. Fuss believes these deals are too speculative, with insufficient yields to compensate for the substantial risks and uncertain future revenue streams.
The market is already showing signs of stress. Credit default swaps for major players have surged dramatically, with Oracle's protection costs more than doubling since September following its $18 billion public bond issuance and $38 billion private loan. Similarly, CoreWeave's credit default swaps gaped higher recently, mirroring the data-center company's stock decline.
Historical Patterns Suggest Caution
BCA Research's analysis reveals uncomfortable similarities between the current AI investment boom and previous technological revolutions. The research firm notes that investors consistently fail to appreciate the "S-shaped nature of technological adoption" - an initial explosive growth phase typically followed by more moderate adoption rates.
Historical precedents are sobering: Railroad securities peaked before spending declined in the 1800s, utility stocks and bonds topped out ahead of capex reductions in the 1920s, and dot-com stocks cratered in 2000 before investment slowed. The oil industry has experienced similar boom-bust cycles repeatedly over the past half-century.
BCA warns that if no new bubble emerges to cushion the eventual AI downturn, the resulting recession could be more severe than the 2001 dot-com bust. The firm points to Cisco Systems as a cautionary tale - the company's stock only recently recovered its March 2000 peak of $78 per share, after bottoming in single digits during October 2002.
Meanwhile, Federal Reserve officials have cast doubt about another rate cut at the December 9-10 policy meeting. Odds of a quarter-percentage-point reduction from the current 3.75%-to-4% target range for federal funds had dropped to 45.8% from over 90% before the previous meeting ended October 29.
J.P. Morgan credit strategists project investment-grade corporate-bond borrowing will rise to a record $1.81 trillion in 2026, surpassing the previous peak of $1.76 trillion in 2020. Technology companies specifically are expected to boost borrowings to $252 billion, representing a 61% increase over current levels.
As the Latin root of "credit" means "to believe," the AI debt boom leaves little room for doubt - and growing concern that belief may be outpacing reality.
