Aug 27, 2026 · 38 min
AI data-center debt hides risks beyond hyperscaler balance sheets
Big Tech's AI Spending Has Gone Full Enron... Literally
The financing structures behind AI infrastructure could spread losses through private credit, corporate debt, and sovereign markets before investors can see the exposures.
- 1Lenders may be underwriting hyperscalers’ credit strength more than the underlying economics of their data-center projects.
- 2Plateauing model gains and falling prices could leave AI revenues growing while infrastructure and operating margins deteriorate.
- 3Special-purpose vehicles and syndicated debt can move data-center risk away from corporate balance sheets and obscure it from investors.
Don't miss
The hosts compare off-balance-sheet data-center financing through special-purpose vehicles with Enron-era structures, arguing that formal disclosure may not make the risks legible.
The brief
Dick Costolo and Paul Kedrosky compare AI’s data-center buildout with railroad overbuilding and the late-1990s fiber boom, but identify a crucial difference: today’s borrowers have exceptionally strong balance sheets.
The hosts argue that lenders may care more about Meta, Google, and Microsoft’s ability to repay than whether each project makes economic sense, encouraging redundant infrastructure and oversupply.
They challenge optimistic token-growth and Jevons-paradox narratives as motivated reasoning, while falling model prices and converging capabilities threaten to compress AI margins.
The discussion widens from data centers to sovereign debt: investors choosing leveraged infrastructure projects over Treasuries compete for the same marginal dollar of capital.
The sharpest warning arrives with special-purpose vehicles and private-credit partners that hold data-center assets off sponsors’ balance sheets, echoing Enron-era opacity and complicating SEC disclosure.
The episode’s broader conclusion is that AI infrastructure risk is not confined to technology companies; it can migrate through corporate, private-credit, and government debt markets.
Books & mentions
Some links are affiliate links — PodLume may earn a commission if you buy.
