Concrete, Silicon, & Leverage: The $4T Debt Behind AI Data Centers

The article analyzes the financial scale of the planned US data center buildout from 25 GW to 70 GW over five years, costing roughly $5 trillion globally.

The central question is whether credit markets can supply the $4 trillion in new debt needed (assuming 70%+ leverage).

Comparing this debt to existing credit markets shows it represents a 34% expansion of the US corporate bond market, triples the outstanding commercial paper market, exceeds the global private credit market, and equals 91% of the $4.

4 trillion municipal bond market. The author notes that municipalities may use municipal bonds to finance data centers, similar to power plants.

Servicing this debt requires annual AI revenue to reach $1.2–$1.5 trillion by 2030, up from an estimated $100–$200 billion today.

This implies a 55% CAGR over five years, compared to current hyperscaler growth rates of 37–82% (AWS 37%, Azure 43%, Google Cloud 82%), with growth accelerating.

For context, the global enterprise software market is ~$1.4 trillion today, out of an estimated $9 trillion in worldwide IT spending by 2030.

The article concludes that financing AI infrastructure is no longer a venture capital or corporate earnings story but a macroeconomic credit event rivaling the largest debt expansions in financial history.

Concrete, Silicon, & Leverage

View Original