Frontier labs captured $2.5tn in value vs $360bn for AI app companies
alexisgallagher · x · 2026-11-16
A thread by @fleetingbits quantifies where AI's economic value accrues: since 2023, frontier labs have created roughly $2.5tn in market value while application AI companies created only $360bn (including the likes of ElevenLabs). Revenue shows an even starker gap — OpenAI and Anthropic together have $130bn in revenue versus $13bn for all app companies, and over $115bn of frontier revenue never passes through the application layer. The takeaway: most AI value comes from model ownership, not productization. Quoter alexisgallagher notes this matches his own budget — he spends more on OAI/Anthropic usage than on all downstream apps — but flags the scenario that could flip it: stalled capability progress, open source catching up, commoditized models, and collapsing pricing power.
More from Venture
- Popular AI content company's organic visibility collapses to near zero — lilyraynyc · 2026-11-16
- Dev estimates AI-assisted enterprise SaaS still needs 18,000 hours vs 26,000 human-only — ucx98 · 2026-11-16
- Clanker Cloud hits 151 agent runs in a week, 700+ users, shifts to enterprise pilots — tekbog · 2026-11-16
- Jason Furman: AI bubble burst would mean 25% baseline stock market decline, 50% only tail risk — sebkrier · 2026-11-15
- Bindu Reddy: Nvidia's 80% margin unsustainable as Neocloud financing props up its own demand — bindureddy · 2026-11-15
- Polymarket opens next Nvidia CEO market: Rene Haas leads at 28% — Polymarket · 2026-11-15