AI’s jobs risk goes beyond automation, says Krugman

AI could weaken demand for workers before it replaces their tasks. Paul Krugman argues that the data-centre boom diverts capital from businesses that employ far more people.

His chart contrasts the rise in hyperscaler spending since 2022 with a much smaller increase in total investment, both measured as shares of GDP. He interprets the gap as evidence of crowding out through borrowing costs and competition for electricity and equipment. The comparison alone does not establish causation.

Krugman’s illustrative calculation puts roughly 10 million fewer jobs supported by capital allocated to data centres rather than elsewhere, using a projection of more than $10 trillion in hyperscaler spending through 2032. This is a conditional comparison, not a forecast of 10 million additional unemployed people .

The question for the AI investment boom is whether downstream productivity gains outweigh displaced investment. Krugman acknowledges that real wages could still rise if AI makes the economy sufficiently more productive.

Source and original chart: Paul Krugman, “Will AI Crowd Out Jobs?”, 11 October 2026. Chart estimates are the author’s, not Dealroom data.