Big Tech faces rising financing costs amid AI spending surge

For years, Big Tech companies printed cash faster than they could spend it. Those days are over. The AI infrastructure buildout has turned the industry’s most profitable companies into some of the bond market’s most prolific borrowers, and the cost of all that debt is climbing.

The combined debt of the five largest US data-center spenders, Microsoft, Alphabet, Amazon, Meta, and Oracle, has ballooned to roughly $350 billion over the past five years. Their collective interest expenses topped $10 billion in the last year alone, more than double what they were paying in 2019.

The debt machine revs up

Goldman Sachs estimates that AI-related corporate debt issuance will hit approximately $141 billion in 2025. That figure is being driven by blockbuster bond deals from Meta, Alphabet, and Oracle, companies that historically financed their growth through cash flows rather than credit markets.

Meta’s recent $12.5 billion bond offering for a Texas data center carried notably higher yields than Meta’s comparable bond issuance earlier in 2025. Oracle’s credit default swap spreads have also been widening, signaling that the market views the risk of lending to these companies as incrementally higher than before.

UBS projects that new tech debt supply could rise to $900 billion by 2026. Morgan Stanley and JPMorgan have penciled in total AI infrastructure borrowing needs of $1.5 trillion by 2028.

Where all the money is going

Hyperscalers are projected to spend approximately $490 billion on infrastructure in 2026, a figure that has been revised upward from earlier estimates. Combined capital expenditures for Microsoft, Amazon, Alphabet, and Meta alone may reach around $725 billion in the near future.

Winners, losers, and what investors should watch

Oracle’s credit spreads have widened more noticeably, reflecting investor concern about a company with a smaller revenue base taking on substantial debt to compete in the AI infrastructure race.

For credit market investors, the flood of tech debt creates a saturation problem. When $141 billion in AI-related bonds hits the market in a single year, supply starts to overwhelm demand. That dynamic pushes yields higher across the board, meaning even the strongest borrowers pay more than they otherwise would.

Traders should be watching credit spreads as a leading indicator. Widening CDS spreads and higher yields on new issuances are the market’s way of saying it’s getting nervous. That nervousness hasn’t reached alarm levels yet, but $1.5 trillion in projected borrowing needs by 2028 is the kind of number that tends to concentrate minds.