Particle.news

Study Says U.S. AI Data‑Center Buildout Will Cost About $10.3 Trillion

The paper warns the scale and opaque financing of the boom could create broad financial strain and raise the odds of an oversupply correction

Overview

  • A Brookings/Columbia analysis estimates roughly $10.3 trillion in investment will be needed from 2025 to 2032 to add about 183 gigawatts of new compute capacity in the United States.
  • The report finds at least $1.3 trillion in debt has already been committed and says developers are using complex, hard‑to‑track funding vehicles that spread risk across banks, private credit, insurers and pension funds.
  • Hyperscaler capital spending has surged from about $97 billion in 2020 to more than $400 billion in 2025 and is projected to approach $800 billion in 2026, forcing many projects to seek outside financing.
  • Researchers warn history suggests five to eight years of rapid growth could be followed by speculative overbuilding, falling prices and a market correction if demand or revenue falls short of the returns investors expect.
  • Execution risks include single‑tenant concentration, fast hardware obsolescence, long development timelines and power procurement bottlenecks, and growing local resistance and executive calls for slower development could temper the pace of construction.