A Brookings paper estimates U.S. artificial intelligence infrastructure spending could average 3.63% of gross domestic product (GDP) through 2032, a larger annual share of output than two earlier U.S. infrastructure booms. Columbia Business School professor Stijn Van Nieuwerburgh's central scenario totals $10.3 trillion from 2025 through 2032. It covers data centers, power infrastructure and specialized computing equipment, rather than company valuations.
On the paper's annual investment-to-GDP measure, railroads accounted for 2.24% from 1870 through 1890, while telecom and fiber accounted for 1.10% from 1996 through 2003. These figures compare annual investment with output, not the total dollars invested across each period.
In an Oct. 8 interview with Brookings, Van Nieuwerburgh said the historical calculations were assembled on different bases. He also said the railroad estimate may be too low because it excludes locomotives and rolling stock.
The paper's capacity model lays out a possible path based on its assumptions: 182.8 gigawatts of additional U.S. data-center capacity would become operational from 2025 through 2032, another 117.2 gigawatts would be completed after 2032, and 226.9 gigawatts would never be completed.
Oracle, Microsoft, Amazon, Meta and Alphabet spent more than $400 billion combined on capital projects in 2025, compared with about $97 billion in 2020. Van Nieuwerburgh's paper projected their total would exceed $800 billion in 2026, more than their combined operating cash flow. It describes a larger role for external funding through leases, joint ventures, project finance, private credit and structured-finance vehicles.
For investors and lenders, the paper flags reliance on a small group of tenants, equipment that may become outdated, power constraints, construction delays and uncertain asset values. NDTV Profit separately reported, citing SBI Funds Management, that uncertainty about AI revenue could make it harder for companies to sustain spending at current financing costs, with possible effects on U.S. equities, economic growth and capital allocation.
