AI Infrastructure Spending Surge Draws IMF Macro Warning

Hyperscaler spending on AI infrastructure has reached a scale that is drawing attention from economists and investors. Jensen Huang told CNBC that a 1-gigawatt NVIDIA AI factory costs $50-60B to build but can generate approximately $50B in annual rental revenue, implying a capital payback period of roughly one year. Combined hyperscaler capital expenditure from Alphabet, Microsoft, Amazon, Meta, and Oracle for 2026-2027 is now approximately 66% above where it stood at the start of 2026, with projections rising by roughly $750B since then. IMF Managing Director Kristalina Georgieva stated that AI investment as a share of GDP will likely exceed the historic railroad buildout, and identified AI infrastructure spending as a factor keeping global inflation above target. Separately, Sequoia's David Cahn calculated a roughly $600 billion annual gap between what hyperscalers are spending on AI infrastructure and what the AI ecosystem generates in actual sales revenue, a divergence that has widened into 2026.