Building out US artificial-intelligence infrastructure could require more than $10 trillion in investment through 2032, while increasingly complex financing structures may create systemic financial risks, according to a new study by Columbia Business School professor Stijn Van Nieuwerburgh.
Reuters reported that the paper, prepared for a Brookings Institution conference, estimates annual AI infrastructure investment could reach around 3.6% of US gross domestic product through 2032. The researcher said that would exceed the annual share of output absorbed by earlier large infrastructure buildouts such as railroads, highways and telecommunications.
The estimate is academic research rather than an official government forecast. The study argues that financing is moving beyond the cash holdings of major technology companies toward banks, private credit, real-estate vehicles and other structures, increasing leverage and spreading risk.
The researcher warned that untested revenue streams and complicated financing could magnify losses if the investment boom suffers a significant downturn.