Jitters over AI spending set to grow as US tech giants flood bond market

Specialists John O'Hara, left, and Glenn Carell work on the floor of the New York Stock Exchange, Thursday, Nov. 20, 2025. (AP Photo/Richard Drew)

NEW YORK — Investors are growing uneasy that the rapid rise in public debt used to bankroll AI investments could strain the U.S. corporate bond market and eventually dampen the appeal of tech stocks, despite leverage across most major companies remaining low for now.

Big tech firms are turning aggressively to the debt markets in their race to build AI-ready data centers, a shift for Silicon Valley firms that typically relied on cash to fund their investments.

Since September, public bond issuance by four of the major cloud computing and AI platform companies known as “hyperscalers” has hit nearly US$90 billion, with Google owner Alphabet selling $25 billion in bonds, Meta $30 billion, Oracle $18 billion, and Amazon, the most recent, $15 billion, according to Reuters calculations of publicly available data. Only Microsoft, the fifth one, has not tapped the debt market in recent weeks.

Investors say that, so far, they are not overly concerned about the impact on stock valuations because of the recent fundraising, since these companies remain lightly leveraged relative to their scale.

But the sudden pickup in public debt issuance has raised questions about the market’s ability to absorb the surge in supply and is feeding into growing worries over AI-related spending that have helped trigger a sharp pullback in U.S. stocks this month after six months of gains. The S&P 500 is still up 11 per cent this year, with tech stocks among the main contributors to gains.

“You have all these hyperscaler issuance coming out, and I think the market woke up to the fact that it’s not going to be private credit markets that are going to fund AI, it’s not going to be free cash flow. It’s going to have to come from the public bond markets,” said Brij Khurana, portfolio manager at Wellington Management Company.

“You need capital to come from somewhere to finance this,” he said. “What’s happening is a recognition that you need money almost coming out of stocks into bonds.”

Including a $27 billion financing deal Meta struck with Blue Owl Capital in October to fund its biggest data center project, hyperscaler debt issuance has jumped to over $120 billion this year from an average of $28 billion over the past five years, analysts at BofA Securities said in a recent note.

Rising debt at tech companies adds a new layer of concern to a market that, despite being fueled by the promise of high AI returns, remains wary that the technology has yet to deliver the profits needed to justify such large capital spending.

“There are doubts that have emerged in the last few weeks around the AI spend story that are related to ... the need for firms to be able to finance that, and that includes through debt finance,” said Larry Hatheway, global investment strategist for Franklin Templeton Institute.

AI capital expenditure is projected to increase to $600 billion by 2027, up from over $200 billion in 2024 and just under $400 billion in 2025, and net debt issuance is expected to reach $100 billion in 2026, Sage Advisory, an investment management firm, said in a recent note.

While hyperscalers have been ramping up borrowing, Nvidia NVDA.O, a major supplier of computing power to them, has trimmed its long-term debt from $8.5 billion in January to $7.5 billion by the end of the third quarter. Credit ratings agency S&P Global Ratings last month revised its outlook on the company to “positive” from “stable,” citing revenue growth and robust cash flow.

Microsoft and Oracle declined to comment. An Amazon spokesperson said proceeds from its recent bond sale will fund business investments, future capex, and repay upcoming debt maturities, noting that such financing decisions are part of routine planning. Alphabet and Meta did not immediately comment.

Market constraints

Demand for recent tech bond deals has been strong, but investors demanded sizeable new issue premiums to absorb some of the new securities. Alphabet and Meta paid about 10 to 15 basis points over the companies’ existing debt with their most recent debt issues, said Janus Henderson in a note.

While U.S. investment grade credit spreads - which indicate the premium highly rated companies pay over Treasuries to attract investor demand - remain historically low, they have ticked up in recent weeks, partly reflecting concerns over the new wave of bond supply hitting the markets.

“For much of the year, credit spreads have been grinding tighter ... But the recent deluge of supply – particularly from tech – may have changed the game," said Janus Henderson.

To be sure, the shift to debt is expected to remain a small portion of total AI spending by large tech companies, with UBS recently estimating that about 80 to 90 per cent of their planned capital expenditure still comes from cash flows. Sage Advisory’s research said the top hyperscalers are expected to move from having more cash than debt to just modest levels of borrowing, still keeping leverage below one x, meaning their total debt would be less than what they earn.

“Supply bottlenecks or investor appetite are more likely to act as constraints on near-term capex than cash flows or balance sheet capacity,” analysts at Goldman Sachs said in a note this week.

Excluding Oracle, hyperscalers could absorb up to $700 billion in additional debt and still be viewed as safe, keeping leverage below that of the typical A+ rated firm, they said.

“These companies still have very solid business lines that are just spinning off tons of cash,” said Garrett Melson, portfolio strategist at Natixis Investment Managers Solutions.

(Reporting by Davide Barbuscia in New York; Additional reporting by Lewis Krauskopf, Anirban Sen and Chuck Mikolajczak in New York; Editing by Megan Davies and Matthew Lewis)