Sequoia warns: AI infrastructure needs $3 trillion in revenue to pay back

Sequoia Capital partner David Cahn published an updated analysis of AI infrastructure economics, concluding that the industry faces a widening payback gap.

He estimates 2026 AI infrastructure spending at approximately $1.5 trillion. After accounting for operating costs and operator margins, the industry would need to generate roughly $3 trillion in annual revenue to justify the investment — a figure Cahn himself notes may be conservative.

For context, the largest frontier labs are still far from that target. Anthropic is reported to be near $60 billion in annual recurring revenue, while OpenAI disclosed annualized revenue above $25 billion.

Apollo chief economist Torsten Slok warned that slower-than-expected payoffs from Google, Meta, Microsoft and Amazon could tip the economy into recession and trigger an S&P 500 correction. At the same time, token prices continue to fall and new models improve coding efficiency by 54%, tightening the commercial case for massive infrastructure buildouts.