Involution Without Export Is Wasted Effort
Summary
David Cheng reports a week with Chinese AI builders against a soft domestic macro: light Beijing traffic, 17.9% youth unemployment, and AI as the sector the state insists must work.
Macro is the missing frame.
Light traffic, Didi as youth fallback, weak retail, and the property hangover are the backdrop Chinese builders assume and American AI narratives often skip.
Involution without export is wasted effort.
Domestic labs undercut, poach, and ship monthly; overseas demand is how that excellence becomes revenue.
if you don’t go overseas, all that grinding was for nothing
Involution without export is wasted effort
Efficiency is constraint, not strategy.
Post-training gains and modular SE Asia data-center gambits follow from denied frontier chips; insiders call it 1-to-10 excellence with thin 0-to-1.
we don’t have access to high performing chips
the entire Chinese AI industry is built on 1-to-10 excellence with little 0-to-1 capability
Open weights and price wars are export tools.
Courtship of American developers and collapsing prices fall out of needing foreign users when domestic demand and capital markets do not clear.
Embodiment and sovereign stack are local strengths.
Cheng separates what China is getting right (bodies, infrastructure, speed) from the bear case builders themselves voice.
Read the original ↗Save to your Sponge
Summarized with Sponge. Quotes are the author’s own words.