Gary Marcus says AI infrastructure would need $2.5 trillion in annual revenue to pencil out
GaryMarcus · x · 2026-07-29
Gary Marcus says AI capex would require $2.5T in annual revenue to justify itself
Quoted post argues that big tech’s AI spending must at least cover depreciation and earn a small profit on each unit of capex.
- Under that assumption, the implied requirement is about $2.5 trillion of AI revenue every year, forever.
- The quote’s punchline is that current AI revenue is nowhere near that scale.
It’s a critique of the economics behind massive AI infrastructure spending, not a claim about a specific company’s fundraising or product launch.
More from Venture
- Creator marketing pro builds CreatorFlow app with AI coding agent EazoAI from one messy prompt — HeyToha · 2026-10-16
- 10 security vendors rated "cooked" by AI are up 73% in six months, beating S&P's 19% — RealGeneKim · 2026-10-16
- Anthropic IPO math: Amazon+Google set for ~$600B, FTX sold its stake 100x too early — simonada · 2026-10-16
- 7 ChatGPT prompts that turn your skills into side-income ideas, from a viral thread — aitrendz_xyz · 2026-10-16
- China's AI boom drives up to 660% profit surges for domestic chip firms; GPU maker MetaX turns profitable — pstAsiatech · 2026-10-16
- Dev adds Google Merchant Center to his MCP server — and explains how it beats Google's official one — DutchSEOnerd · 2026-10-16