Tech giants have quietly accumulated roughly $500 billion in off-balance-sheet debt to finance AI infrastructure and data centers, keeping these obligations largely out of public view under current accounting rules. Major companies, including Meta, Microsoft, Amazon, Google, and Oracle, are funding massive projects through private credit deals and long-term capacity agreements rather than traditional loans or bonds.
Institutional investors, especially insurance companies and pension funds, have poured around $450 billion into these private loans, drawn by yields of approximately 9%—far above the 4% available from conventional bonds. UBS estimates that tech firms are adding about $125 billion in new off-balance-sheet commitments every quarter.
Early signs of stress are already visible in credit markets, with Oracle’s credit default swap spreads rising 67 basis points in just two months. The viability of this debt depends on AI initiatives generating at least 12% annual returns, yet most large-scale AI projects remain unprofitable and heavily reliant on continued rapid adoption.
A slowdown could impair borrowers’ repayment capacity, leaving lenders holding collateral—partially built data centers—whose value is uncertain and often assessed by the borrowers themselves. Rising risk premiums, potential defaults, and heavy concentration in insurance and pension portfolios are amplifying concerns that this fast-growing shadow debt could pose broader risks to financial markets. Investors and regulators are now paying close attention to the opacity surrounding this $500 billion AI-fueled liability pile.