Analyst: public markets misprice scaling and are underpricing AI lab domination
abhiadesai · x · 2026-10-25
Analyst Abhi Desai lays out the two key assumptions behind AI lab dominance: 1) scaling holds (more data and compute → better models) and 2) better models unlock more revenue use cases. Public market investors, he argues, discount #2 while not truly understanding #1.
- #1 makes model training the opposite of a commodity business and potentially a monopoly in the limit — an all-powerful AI (Bostrom's Singleton) is by nature a monopoly. If one lab races far enough ahead, its lead becomes insurmountable, which he ties to recent reports of "melting GPUs" and endgame planning at the labs.
- #2 is illustrated by the capability jump from tab autocomplete to "SWE in a box," which drove this year's massive coding revenue ramp; the same logic extends across enterprise verticals and eventually R&D tasks.
He concludes that outside data, compute, and vital physical inputs (critical minerals, factories, bio labs, energy), labs can compete in almost anything at the cost of compute, and both public and private markets are still underpricing lab domination across most verticals.
Related event: Analyst: Markets Underestimate Big AI Labs' Path to Dominance(2 posts)→
More from AGI Musings
- OpenAI cuts internet access for frontier model training after Hugging Face hack incident — emmanuelvivier · 2026-10-25
- Dan Faggella: work isn't a capitalist invention, it's life fighting entropy — danfaggella · 2026-10-25
- AI researcher: people are taking AI roughly the right amount of seriously — NathanpmYoung · 2026-10-25
- Dev complains frontier lab employees' bio-risk takes are a 'mind virus' ruining his sanity — owl_posting · 2026-10-25
- State of AI Report 2026: the frontier is now a three-lab race between Anthropic, OpenAI and Google — nathanbenaich · 2026-10-25
- Benchmarks should penalize misalignment, not ambiguity — environment fidelity over prompt over-specification — nirmal_dist · 2026-10-25