AI Data Centers Are Now the Biggest Infrastructure Bet in US History — Bigger Than Railroads, by Share of GDP
A Brookings Papers on Economic Activity study by Columbia Business School's Stijn Van Nieuwerburgh, published Sept 23, 2026, projects $10.3 trillion in US AI data-center investment through 2032 — an average of 3.63% of US GDP every year. That tops every prior US infrastructure boom on record: the railroad boom of 1870-1890 (2.24% of GDP, the previous record), the interstate highway system of 1956-1973 (1.13%), and the 1990s telecom/fiber buildout (1.10%). The paper's own thesis isn't just the scale — it warns that financing is migrating from transparent, on-balance-sheet spending into opaque joint ventures, private credit, securitization and special-purpose vehicles, all leveraged to uncertain AI demand. The chart has been circulating virally since publication: investor Andrew Chen first posted it Sept 23, Elon Musk quote-tweeted a reaction Sept 24 ('An Earth economy is less than a trillionth the size of a K2 economy'), and it is still being actively reposted as of Sept 26 (e.g., InvestAnswers/The Kobeissi Letter).
Receipts
- primaryhttps://www.brookings.edu/articles/financing-the-ai-buildout/
- alsohttps://qz.com/us-ai-infrastructure-buildout-biggest-economic-bet-history-092426
- alsohttps://money.usnews.com/investing/news/articles/2026-09-24/financing-of-historic-ai-buildout-raises-systemic-risks-in-us-researcher-says
- alsohttps://www.cryptopolitan.com/musk-ai-spending-3-63-gdp-kardashev-ii/
- alsohttps://techblog.comsoc.org/2026/09/24/the-ai-infrastructure-build-out-a-10-trillion-bet-on-compute-power-and-networks/
Flagged:Primary is Brookings' own Sept 23, 2026 publication of Van Nieuwerburgh's paper, which states the $10.3T/3.63% figures and the full historical comparison table (canals 0.66%, railroads 2.24%, electrification 0.50%, highways 1.13%, telecom/fiber 1.10%, AI 3.63%) directly. Quartz, US News (via Reuters), IEEE ComSoc and Cryptopolitan independently corroborate the same figures and the systemic-financing-risk framing. This card's freshness call: the paper itself published Sept 23 (3 days before this card was built), which is outside the strict 24-48h window this channel otherwise enforces — but unlike a stale recap, this story has an actively escalating viral cascade with new high-profile amplification each day (Andrew Chen's chart post Sept 23, Musk's quote-tweet Sept 24, continued heavy reposting through Sept 26, including the InvestAnswers/Kobeissi Letter repost that prompted this card), so it was treated as a live moment rather than old news being repackaged. The '3.63%' figure is written as filed by Brookings; some secondary outlets round it to '3.6%' — both refer to the same underlying number, not a discrepancy.