Columbia professor warns US artificial intelligence buildout may cost over $10 trillion and create systemic financial risks

A new study by Columbia Business School professor Stijn Van Nieuwerburgh warns that the US artificial intelligence buildout will require over $10 trillion in investment through 2032, consuming 3.6% of GDP annually. This is more than previous major infrastructure projects like railroads or highways. The study, prepared for a Brookings Institution conference, highlights that financing has shifted from company cash reserves to complex external arrangements, increasing leverage and systemic risk. It draws parallels to the subprime mortgage crisis and notes that AI revenues would need to grow 80% annually to justify the investment.