Debt-Funded AI Build-Out Draws Warnings as Borrowing Costs Climb

"Cracks are appearing in America's debt-laden $10tn AI infrastructure boom as project delays, public backlash and shaky investor confidence bring a reckoning for the data centre build-out underpinning the US economy," the Financial Times wrote on October 18. Mohamed El-Erian shared a chart from The Economist on "the fast-growing stock of AI-related debt," warning that without a matching rise in loanable funds, persistent rate pressure "will likely crowd out rate-sensitive, non-tech sectors".

The spending is enormous. Citing Vanguard research, The Motley Fool reports that GOOGL, MSFT, AMZN, META and ORCL will spend almost $780 billion on capital expenditure in 2026, up 87% year over year, and more than $1 trillion annually from 2027 through 2030. The same piece notes Alphabet's long-term debt reached $98.2 billion as of June 30 and Meta's $83.7 billion, and that Alphabet posted negative free cash flow of $5.9 billion in the second quarter. ZeroHedge highlighted an estimate that $1.6 trillion of projected AI capex drives 5% of 2027 sales and 25% of sales growth, with half of it debt-funded, though the post did not identify the source of the estimate.

Borrowing costs are highest for the neoclouds. CRWV bonds have reached 15% yields, according to The Motley Fool, and TheStreet reports that CoreWeave and Oracle are borrowing billions for AI just as rates rise. The payback math is also stretched: a Columbia University estimate puts cumulative AI infrastructure spending at about $9 trillion from 2025 to 2032, requiring more than $3.5 trillion in annual revenue by 2032 just to break even.

The warnings follow last week's sharp selloff in chip and AI infrastructure stocks. The next test comes with third-quarter hyperscaler earnings, where capex guidance, free cash flow and new debt issuance will show whether the build-out is still being financed comfortably or is starting to strain balance sheets.