Brookings estimates U.S. AI infrastructure will cost $10.3 trillion from 2025 through 2032, averaging 3.63% of U.S. GDP a year—a larger share than past major infrastructure booms. It warns investment risks are shifting toward opaque off-balance-sheet financing.
Examples of the less-visible structures include joint ventures, private credit, securitization, special-purpose vehicles, lease commitments and loan guarantees. The paper says they depend on AI companies’ cash flows and collateral values, which are exposed to uncertain demand, rapid technological change, timely access to power and hardware, and the credit quality of a few data-center tenants. Study author Stijn Van Nieuwerburgh said it is premature to conclude the buildout already poses systemic risk comparable to earlier credit booms; he urged better measurement and transparency as the industry’s financing evolves.
Amazon, Microsoft and Google canceled large-scale projects in Arizona, Wisconsin and Indiana, respectively, after sustained opposition; residents cited concerns about power bills, noise, environmental effects and security. The Wall Street Journal reported that last year's largest data-center sales-tax exemptions were $1.94 billion in Virginia, $1.9 billion in Georgia, $1.57 billion in Ohio and $1.02 billion in Texas. Ohio Democratic Rep. Tristan Rader wants to repeal the state's exemption and renegotiate past deals. Amazon says it has invested nearly $40 billion in Ohio data centers since 2015; Meta says it has invested more than $2.3 billion there since 2018.
