Brookings Institution Forecasts $10.3 Trillion Investment in AI Infrastructure by 2032

The Brookings Institution has released a study projecting that the United States will invest a staggering $10.3 trillion in AI-related infrastructure between 2025 and 2032. This investment is expected to average around 3.63% of the U.S. GDP annually during this period, significantly exceeding past infrastructure spending benchmarks.

Historically, the peak of the railroad boom represented about 2.24% of GDP, while this new AI investment is set to outpace that by a considerable margin. The report, authored by Stijn Van Nieuwerburgh from Columbia University, breaks down the projected spending across various categories, including data centers, power systems, network infrastructure, and specialized chips.

Notably, hyperscaler capital expenditures are anticipated to rise dramatically, from $97 billion in 2020 to over $400 billion by 2025, with projections for 2026 alone reaching approximately $800 billion. Major tech companies such as Amazon, Alphabet, Microsoft, Meta, and Oracle are identified as the primary drivers of this spending, with a combined capital expenditure forecast of $4.2 trillion through 2029.

The study also highlights a significant increase in compute power, estimating an additional 183 gigawatts by 2032, with some estimates suggesting this could rise to as high as 509 gigawatts depending on demand and build-out speed.

However, the financing mechanisms for this massive investment are evolving, moving towards more complex off-balance-sheet structures such as joint ventures and private credit facilities. This shift raises concerns about future AI demand and operational risks, as reliance on these financing methods introduces uncertainties into the investment landscape.