A 2026 study by Stijn van Nieuwerburgh from Columbia Business School estimates that US AI infrastructure investment will reach $10.3 trillion between 2025 and 2032. This annual expenditure represents approximately 3.63% of the US GDP.
Major hyperscalers, including Microsoft, Amazon, Meta, Alphabet, and Oracle, are expected to account for $4.2 trillion in capital expenditures through 2029. Research indicates that this spending will likely exceed their operating cash flow by 2026, forcing reliance on external financing.
The sector faces significant operational hurdles, including labor shortages, construction delays, and power supply constraints. Projections suggest a potential energy shortfall of up to 68 GW for data centers, with companies like Google reporting grid connection delays of five to ten years in regions such as PJM and ERCOT.
To justify the projected investment, the AI industry would need to generate roughly $3.7 trillion in annual revenue by 2032, compared to current estimates of $100 billion. The IMF has warned that a decline in AI profitability could lead to stock market corrections.
