In a Financial Times opinion column, chief economics commentator Martin Wolf draws on "Financing the AI Buildout," a paper by Columbia Business School professor Stijn Van Nieuwerburgh, to ask whether markets are ready for an AI bust. The paper estimates that U.S. spending on the AI buildout will average 3.6% of GDP from 2025 to 2032. That's more than the 2.2% spent on railways from 1870 to 1890 and the 1.1% spent on highways from 1956 to 1973.
Van Nieuwerburgh finds that capital spending by Alphabet, Amazon, Meta, Microsoft, and Oracle rose from $97 billion in 2020 to $416 billion in 2025, and is forecast to reach $801 billion in 2026. At that point, spending would exceed the companies' operating cash flow by $93 billion. Data center financing is shifting toward developers, REITs, infrastructure funds, and private equity, and much of it runs through bank loans and private credit with little public disclosure. As the paper puts it, "the sector's financing structure can transmit and amplify those risks." Wolf argues that the longer the boom and its financial engineering continue, the bigger the economic impact is likely to be when it ends.
