Real Estate
Home›Real Estate›Industry›AI infrastructure build-out could cost about $10.3T by…
AI infrastructure build-out could cost about $10.3T by 2032
A Brookings and Columbia analysis estimates the plan would add 183 gigawatts of computing capacity by 2032, alongside uncertainty over whether part of the pipeline will ever be built.
AI infrastructure spending could total about $10.3T in investment capital between 2025 and 2032, a new paper from Columbia University economist Stijn Van Nieuwerburgh and the Brookings Institution finds, with the annual cost equivalent to about 3.6% of U.S. GDP per year. According to the research, the scale of the build-out is expected to be the most expensive in American history, surpassing past infrastructure booms such as the 1870 to 1890 railway expansion, which averaged 2.2% of GDP toward rail investment. The paper also contrasts the forecast with other large projects, estimating AI build-out costs exceed the highway system total and far outpace electrification spending around the turn of the 20th century, according to Bisnow. The authors estimate the spending would fund additional computing power of 183 gigawatts by 2032, while acknowledging that pipeline estimates vary. The analysis cites a larger total pipeline of 509 GW, assumes 227 GW of proposed capacity will never be built, and expects another 117 GW to come online after 2032, Bisnow reported. The paper also estimates the cost at roughly $8.2B for every 200 megawatts of compute power built.
Breaking down financing risk, Bisnow said the study points to increasingly complex and opaque financing mechanisms used to get projects built. It adds that Big Tech financing has risen sharply, with major hyperscalers’ capex climbing from $97B in 2020 to more than $400B in 2025, and projecting them to clear $800B in 2026, the outlet reported.