Governments used to have the bond market’s biggest appetite more or less to themselves. Now they have company, and it comes with a server farm attached.
Amazon, Alphabet, Meta, Microsoft and Oracle are borrowing heavily to fund artificial intelligence infrastructure. That puts them in direct competition with sovereign borrowers, especially the US government, for the same pool of investor cash.
The numbers behind the borrowing spree
Four of the largest US tech companies plan to spend about $650 billion on AI infrastructure in 2026. Widen the lens to all hyperscalers, and capex estimates for the year exceed $800 billion to $1.1 trillion.
AI-linked corporate debt issuance is reported to have reached approximately $450–500 billion in the first nine months of 2026. That is more than double the pace of previous years.
Analysts at Goldman Sachs and Morgan Stanley expect the trend to escalate, with Goldman Sachs projecting hyperscaler AI debt issuance could push toward or above $1 trillion in 2027.
Tech firms have moved more than $120 billion of AI data center spending off their books through special purpose vehicles, or SPVs, and leasing arrangements. Think of an SPV as a separate company created to own the asset and carry the debt. The tech giant uses the data center and pays for it over time, but the borrowing sits elsewhere.
Aggregate undisclosed lease and purchase commitments for major hyperscalers exceed $1.5 trillion.
From cash-funded growth to leveraged bets
Bond buyers have noticed. Even investment-grade names are paying up. Investors are demanding yields above 7% from Meta, a company that is not exactly short on revenue.
Oracle offers a sharper warning sign. Based on recent credit default swap spreads, its implied 5-year default risk has moved above 20%. CDS contracts work like insurance against a borrower failing to pay. When that insurance gets pricier, the market is signaling more worry.
All of this is happening against a heavy macro backdrop. Global debt hit a record $365.5 trillion in the first half of 2026. The US is still running high fiscal deficits, which means Treasury needs a steady stream of buyers for new government debt.
What this means for borrowers, investors and the US Treasury
Capital is finite, even in a market as deep as US fixed income. When hyperscalers issue hundreds of billions in new bonds, they are courting many of the same pension funds, insurers and asset managers that buy Treasurys and other sovereign debt.
The off-balance-sheet structures add a transparency problem. Investors pricing a hyperscaler’s bonds may not see the full scope of its commitments. The more than $1.5 trillion in undisclosed lease and purchase obligations is precisely the kind of figure that tends to matter most when sentiment turns.
For investors, a few signals are worth watching. Credit rating actions on the major hyperscalers would be one, since downgrades could move both bond prices and stock prices. CDS spreads, particularly for Oracle, offer a real-time read on how nervous lenders are getting.
Treasury auction demand is another. If government debt sales start drawing weaker interest while tech issuance stays strong, the competition for capital will move from theory to measurable fact.
Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.