US AI infrastructure spending may surpass past mega-projects, study warns

A new study shows that US spending on AI infrastructure may consume more of the economy's output than past mega-projects like electricity, railways, interstate highways, and the internet. Stijn Van Nieuwerburgh, a professor at Columbia Business School, wrote in a paper that AI investment, once funded by cash from Amazon, Meta, and Google, has grown into a massive expansion plan.

By 2032, about 3.6% of US GDP is expected to go into this area each year, totaling over $10 trillion. That is higher than the railway boom of the late 1800s, which took about 2.2% of GDP annually. Data center construction and AI risks have become key topics in US politics and economics.

Van Nieuwerburgh noted that the huge scale, unproven revenue sources, and increasingly complex financing structures could bring major downside risks. He estimates that by 2032, the AI industry would need to generate about $3.7 trillion in annual revenue to deliver the expected returns.