Columbia Business School economist Stijn Van Nieuwerburgh warns that without hundreds of billions of dollars in AI data centre spending, equivalent to more than £200 billion at current exchange rates, the US economy "would basically be in a recession right now." His research, titled 'Financing the AI Buildout', estimates AI infrastructure investment could average roughly 2.8% of GDP, larger than the railroad boom at its peak, which peaked at around 2.4% of GDP between 1865 and 1890. A traditional data-centre server rack might consume 5 to 10 kilowatts of power, while a rack with cutting-edge AI chips can require more than 100 kilowatts. The price tag for a single state-of-the-art AI campus now runs to upward of $8 billion, increasingly funded through debt. Van Nieuwerburgh points to one Meta-linked data centre project financed with roughly $27 billion in debt through special-purpose vehicles, keeping borrowing off Meta's balance sheet. Bond investors have responded, with credit spreads for Google, Amazon and Meta widening, and Oracle's five-year credit default swap trading at a multi-year high. His central concern is that risk is becoming harder to see, as hyperscalers increasingly lease facilities, giving pension funds, private credit firms and structured-finance vehicles growing claims on AI infrastructure.