AI Infrastructure Buildout to Consume $10 Trillion, 3.6% of US GDP

The artificial intelligence infrastructure expansion will demand approximately 3.6% of US gross domestic product annually through 2032—totaling more than $10 trillion—making it the most capital-intensive technology rollout in American history, according to new research from Columbia Business School.

Stijn Van Nieuwerburgh, a finance and real estate professor at Columbia, presented findings at a Brookings Institution conference this week showing the AI buildout will exceed the relative scale of railroads in the late 1800s (2.2% of annual GDP), the interstate highway system in the 1950s (roughly 1% annually), and the telecommunications expansion of the mid-1990s (also about 1% annually).

Complex financing structures emerge

What began as projects funded from the cash reserves of tech giants like Amazon, Meta, and Alphabet has evolved into an intricate web of financing arrangements involving AI firms, cloud hyperscalers, banks, private credit lenders, and real estate companies. Van Nieuwerburgh estimates the industry needs to build 183 gigawatts of new data center capacity over the next seven years, compared to 57 gigawatts currently installed.

"This is freaking complicated," Van Nieuwerburgh told reporters, describing the emerging arrangements between multiple players in the AI infrastructure ecosystem.

The shift to external financing has increased leverage across the sector and redistributed risks throughout the economy. Van Nieuwerburgh noted the opacity of special purpose vehicles used in these arrangements resembles the complex mortgage financing structures that preceded the 2007-2009 financial crisis.

Revenue projections face steep climb

For the investment to generate expected returns, the AI industry will need to reach approximately $3.7 trillion in annual revenue by 2032. With current combined revenues of OpenAI and Anthropic estimated around $100 billion, this would require roughly 80% annual growth—a pace that remains unproven.

"These developments do not imply that financial distress is imminent," Van Nieuwerburgh wrote in his paper. "Strong growth in AI applications, high utilization, and continued improvements in model capability could support the projected infrastructure and generate stable cash flows."

However, he cautioned that "the combination of uncertain demand, rapid technological change, execution bottlenecks, and high leverage creates meaningful downside risk if expectations are revised."

Why it matters

The scale and financing structure of AI infrastructure development has moved beyond a technology sector concern to a potential systemic economic risk. Federal Reserve officials are already considering whether the construction boom contributes to inflation, while local communities increasingly resist hosting data centers due to resource strains. The dependence on untested revenue streams and complex financing vehicles—combined with the sheer magnitude of capital deployment—creates vulnerability across the financial system if AI adoption fails to meet aggressive projections.

The research was first reported by Reuters and will be formally presented at the Brookings Institution conference.