AI companies raise $236 billion in debt as bond investors show strain

AI companies collectively issued $236 billion in debt over a five-month period ending in mid-2026, a record borrowing spree that is testing the limits of bond investor appetite for the sector's infrastructure buildout.

"The sheer volume is starting to create indigestion in the credit markets," said John Atkins, senior fixed-income analyst at Morningstar. "New issues from hyperscalers and data-center operators are trading below par within weeks of pricing."

Investment-grade bonds from traditional hyperscalers — Amazon, Alphabet, Meta, Oracle — plus data-center developers and AI-focused companies reached $218 billion through July 8, more than doubling the $80.5 billion issued in all of 2025, according to Morningstar data. Amazon alone borrowed $25 billion in July. The US high-yield market absorbed $31.9 billion of AI-related bonds through the same date, nearly all backing new data centers.

Signs of buyer fatigue are emerging. SpaceX's 6.65% 30-year bonds, priced at T+175, traded above T+200 this week. Meta's 6.3% 2056 bonds, part of a $25 billion April package, widened to T+145 from T+120 a month ago. CoreWeave's 9.625% six-year senior notes, priced at par in June, slumped to 96.50, pushing the yield above 10%.

The debt-funded AI buildout is a bet on future revenue that has yet to materialize at scale

The borrowing binge reflects an industry racing to secure computing capacity before rivals. OpenAI is targeting nearly $600 billion in compute spending through 2030, while Oracle plans to ramp AI capital expenditure toward $95 billion in fiscal 2027. Microsoft, Alphabet, Meta, Oracle and Nvidia together hold more than $460 billion in outstanding debt, with nearly $100 billion in new issuance planned for 2026 alone, according to MarketWatch data.

The financing chain is drawing scrutiny. Private-credit AI loans may have nearly doubled, and Morgan Stanley estimates private credit could fund over half of the $1.5 trillion data-center buildout through 2028, Reuters reported. The circularity concern: companies borrowing to buy Nvidia chips, whose soaring revenue and market cap then back further borrowing.

OpenAI illustrates the cash-burn dynamic. The company posted $13.07 billion in 2025 revenue against $34 billion in costs and expenses, according to leaked financials cited by tech critic Ed Zitron, with losses piling up sharply. The gap between revenue and spending underscores the pressure on AI companies to demonstrate that the infrastructure investment can generate returns before financing conditions tighten.

For investors, the question is whether the debt markets can sustain the pace. The $236 billion raised in five months compares with $12.1 billion in the second half of 2025. If yields continue to widen and secondary prices keep slipping, the cost of capital for the next wave of AI infrastructure could rise sharply, squeezing the very companies that need it most.

This article is for informational purposes only and does not constitute investment advice.