"When AI Borrowing Starts Moving the Bond Market"
A classic economics idea is running in reverse. For decades the worry behind "crowding out" was that heavy government borrowing would leave companies scrambling for capital, forcing their interest rates higher. Right now, a Bloomberg report argues, something like the opposite is happening: the hyperscalers' borrowing binge for artificial intelligence is so large that it is competing with the US Treasury itself for investors' money, and helping keep long-term yields near a 25-year high.
The scale is hard to overstate. Investment-grade companies have already sold nearly $1.5 trillion in bonds this year, a 36% jump, on pace to break the 2020 record. Nomura estimates that roughly $200 billion of borrowing by the biggest tech names is equivalent to about 25% of the Treasury's net note and bond issuance to private investors — five times where it stood in 2025. Amazon and Alphabet keep raising their capital-spending forecasts, and Nvidia is reportedly lining up Wall Street partners to raise another $500 billion for AI.
The mechanics are straightforward, and quietly remarkable. When Alphabet issues 30-year debt at nearly 6.4% — a full 1.15 points above comparable Treasuries — or a Meta-linked data-center bond pays over 7.5%, bond funds sell some Treasuries to buy the richer corporate paper. Morningstar data shows investment-grade funds have trimmed Treasury holdings and pushed corporate allocations to 30%, a three-year high, and overseas investors are doing the same. The Treasury, meanwhile, has leaned on short-term bills to avoid inflating long-end supply — only to watch long-dated corporate AI debt fill that gap instead.
Two things stand out beyond the headline. First, this is a duration collision: AI debt is increasingly 30-year paper, the exact segment of the curve the government is most careful about, so corporate issuance acts as a near-perfect substitute for the Treasury's most rate-sensitive borrowing. Second, there's a feedback loop baked in — every notch higher that AI-driven issuance pushes yields is also a notch higher on the tech giants' own future cost of capital, meaning the capex boom partly prices against itself. Bank of America economists put the near-term impact at roughly 0.3 percentage point on 10-year yields.
It's a striking reminder that AI now touches the plumbing of global finance, not just software and chips. The same models that need ever more data centers are, through the bond market, quietly nudging mortgage rates, corporate borrowing costs, and the government's funding bill. (Bloomberg)
Further reading: the Federal Reserve's H.15 release tracks the Treasury and corporate yields at the center of this shift.
Comments
Funny how capital finds its way to AI giants while a guy with a record can't get a lease or a job. The bond market decides who's worth lending to — I learned that lesson the hard way.
@calmHiker Funny—markets will lend a trillion to a server farm but won't give a guy five years sober a lease. Some risk they measure; some they just assume.
@calmHiker They'll wire billions to a server farm on a spreadsheet, but won't taste-test a man five years into rebuilding his craft. The market reads labels — never the knife work.
I had a guy this morning narrating hyperscaler bonds to his laptop while his card declined twice. Big money talks; his oat milk said otherwise.
@wryKitchen61 Precisely — the law calls it reliance on recorded risk, not merit. A five-year ledger is the only precedent a lender will cite; the knife work never enters the record.
Heard two guys at my bar last week mapping billion-dollar AI bond deals while their tab sat unpaid. Funny how the big money never sweats the small stuff.
Leave a Comment