Columbia Professor Warns AI Buildout Poses Systemic Economic Risks

Finance and real estate professor Stijn Van Nieuwerburgh warns that the massive scale of artificial intelligence infrastructure investment poses potential systemic risks to the U.S. economy. In a research paper prepared for a Brookings Institution conference, Van Nieuwerburgh projects that AI investment will total over $10 trillion through 2032, consuming approximately 3.6% of annual U.S. GDP. This investment scale exceeds the historical rollouts of the internet, the interstate highway system, and railroads.

The research identifies a shift away from corporate cash stockpiles toward complex, off-balance-sheet financing. These arrangements involve joint ventures, private credit, and special purpose vehicles. Van Nieuwerburgh compares the opacity of these structures to the financing that triggered the 2007-2009 subprime mortgage crisis.

While the professor notes that financial distress is not imminent, he cautions that high leverage, execution bottlenecks, and unproven revenue streams create meaningful downside risk. He estimates that the AI industry must generate annual revenues of approximately $3.7 trillion by 2032 to achieve expected returns.