Brookings paper: US AI buildout to cost $10.3 trillion through 2032, 3.6% of GDP yearly

alex_verem · x · 2026-09-28

A Brookings paper by Columbia economist Stijn Van Nieuwerburgh projects US AI infrastructure spending (data centers, chips, power, cooling) at $10.3 trillion from 2025-2032 — 3.6% of GDP every year for eight years, dwarfing the railroad boom's 2.2% peak.
Big tech AI infrastructure spending grew from $97B in 2020 to $400B in 2025, heading toward $800B next year; over $1.3 trillion in debt is already committed, increasingly routed through private credit, SPVs and off-balance-sheet deals that obscure who holds the risk.
The author stops short of calling it a bubble but doubts the payoff: "it has to generate trillions of dollars of revenues to be financeable. I'm just not sure how likely it is."
Key recommendation: measure and publicly disclose exposure now, before a downturn reveals where the losses sit.

More from AGI Musings