If the cheapest AI wins the market, does the country that subsidizes compute end up writing the rules?
American firms now route up to 46% of their AI workload through Chinese open-weight models because they’re 60-90% cheaper — even as US lawmakers open a probe. When the strategic resource shifts from oil to compute, should we treat ‘good enough and cheap’ as a national-security problem, or is trying to wall it off a losing game?
Commentaires (1)
In today’s episode of Minds, Bodies, and Terawatts (July 9th, 2026), we dug into the CNBC report that Chinese models like GLM 5.2 have quietly become the default for millions of routine American AI calls — not because they’re smarter, but because they’re a rounding error on the invoice. Our read, drawing on the book’s ‘When Compute Replaces Oil as Power’ framework: whoever supplies the cheapest reliable compute starts setting the terms, the same way the cheapest reliable oil once did. That’s why China weighing export restrictions on its own models is the tell — leverage flows to the supplier, not the smartest lab. The uncomfortable question is whether a probe can reverse an economic gravity this strong, or whether it just makes the dependency more expensive. Give the July 9th episode a listen and tell us where you land — is this a security emergency or just markets doing what markets do?
Related reading on unscarcity.ai:
Envie d'aller plus loin ?
Obtenez le plan complet dans <em>L'ère de la post-pénurie : Repenser la société à l'ère des machines</em>
