What happened
A study presented at a Brookings Institution conference by Columbia Business School professor Stijn Van Nieuwerburgh estimates that the U.S. AI infrastructure build‑out will require roughly 183 gigawatts of new data‑center capacity by 2032, costing about $10 trillion – roughly 3.6 % of annual GDP. The financing model has shifted from cash‑rich tech firms (Amazon, Meta, Google) to a web of special‑purpose vehicles, banks, private‑credit lenders and real‑estate firms, creating leverage and opacity reminiscent of the 2007‑09 subprime mortgage crisis. Van Nieuwerburgh notes that to justify the investment, the AI sector would need to generate about $3.7 trillion in annual revenue by 2032, implying an 80 % yearly growth rate from current combined revenues of OpenAI and Anthropic.
The study, authored by Stijn Van Nieuwerburgh, was presented at a Brookings Institution conference on September 26, 2026 and covered in a Reuters‑by‑Crossroads Today article. It quantifies the AI build‑out as a $10 trillion investment that will consume 3.6 % of U.S. GDP annually through 2032, surpassing the historical cost share of railroads, interstate highways, and telecom expansions.
Financing has moved from internal cash reserves of major tech firms to a complex network of banks, private‑credit lenders, real‑estate firms, and special‑purpose vehicles. Van Nieuwerburgh likens this opacity to the subprime mortgage crisis, warning that similar leverage could amplify systemic risk if AI demand falters.
The paper estimates that to achieve a reasonable return on the $10 trillion spend, the AI sector must generate roughly $3.7 trillion in annual revenue by 2032 – an 80 % compound annual growth rate from today’s combined $100 billion revenue of OpenAI and Anthropic. The study acknowledges that such growth is uncertain and that execution bottlenecks, rapid technological change, and high leverage create “meaningful downside risk.”
Source details: crossroadstoday.com
Why it matters
The analysis highlights a potential macro‑economic vulnerability: if AI demand stalls or financing conditions tighten, the high‑leverage structure could trigger broader financial distress. Policymakers, regulators, and investors may need to monitor AI‑related credit exposure, especially as local governments grapple with resource strain from data‑center construction. The comparison to historic infrastructure rollouts underscores the unprecedented scale of AI investment and the need for oversight to avoid a bubble‑burst scenario that could affect the wider economy.
The projected scale of AI infrastructure spending dwarfs previous technology rollouts, meaning any financial shock could have outsized effects on the broader economy. The reliance on external financing spreads risk across banks and credit markets, potentially exposing the financial system to AI‑specific downturns.
Policymakers and regulators may need to consider new oversight mechanisms for AI‑related financing, similar to those applied to mortgage‑backed securities after the 2008 crisis. Without such oversight, the sector could experience a bubble‑burst that reverberates through credit markets and inflation dynamics.
Local communities are already expressing concerns about resource strain from data‑center construction, indicating that social and environmental factors could further complicate financing and deployment timelines.
Interactive Mechanism: How It Actually Works
Explore the underlying technology behind this development interactively.
Why can ethical evaluation not be reduced to one model score?
What to watch next
Future reports on AI revenue growth, especially from leading model providers, will test the study’s revenue assumptions. Watch for regulatory actions on AI‑related financing, such as disclosures of special‑purpose vehicle structures or central‑bank assessments of inflationary pressure from data‑center construction. Additionally, monitor local opposition to new data‑center sites, which could slow deployment and affect financing timelines.
Revenue growth from leading AI model providers will be a key metric to validate the study’s assumptions. Quarterly earnings reports from OpenAI, Anthropic, and other major players should be closely examined.
Regulatory bodies, including the Federal Reserve and the Securities and Exchange Commission, may issue guidance or reporting requirements for AI‑related financing structures, especially if credit exposure rises.
Local opposition to new data‑center sites could delay construction, affecting the timeline and cost assumptions of the build‑out. Monitoring zoning decisions and community responses will be important.