The AI industry is steaming toward an inflection point: The moment when the industry, as a whole, goes cash flow negative and the hyperscalers start slowing down their capex spending on data centers. This chart from Stijn Van Nieuwerburgh of Columbia Business School shows “how quickly this financing constraint is tightening,” he wrote in a recent paper. “On current estimates, capex will surpass the five firms’ [Oracle, Microsoft, Amazon, Meta, and Alphabet's] combined operating cash flow for the first time in 2026.”

“Internal cash generation remains substantial, but it is no longer sufficient to finance the projected pace of investment without greater reliance on external capital or financing structures that shift assets and obligations away from the operating companies’ balance sheets,” he wrote.

Other analysts are already predicting that 2027 will be the year when AI capex starts to decelerate. “We expect the market will start thinking about a potential peak year in 2027 soon, likely after 3Q EPS and midterms,” Wells Fargo’s Ohsung Kwon told clients this week. “Most investors now agree that AI capex will likely slow meaningfully in 2028.”

The pace of spending simply can’t be sustained, according to Samuel Tombs and Oliver Allen at Pantheon Macroeconomics. “AI-linked investment probably will continue to grow very strongly, but hyperscaler capex growth looks set to moderate in 2027, while supply constraints and higher borrowing costs increasingly will bite,” they said in a recent email.

“Analysts’ forecasts for the five largest hyperscalers point to nominal capex rising at a materially slower pace in 2027 than in 2026,” they said. “Even so, the projected slowdown deserves attention. Supply-side constraints also are becoming a growing issue for the tech sector, with bottlenecks in the supply of advanced chips and electrical equipment, power grid capacity, and construction labor and materials increasingly biting.”

Higher borrowing costs driven by rising yields on long-dated bonds will crimp the plans of firms already holding leverage, they said. But “they are unlikely to derail the buildout while financing remains readily available and the largest tech firms continue to expect exceptionally high returns from AI.”