I think this AI selloff will look overdone in hindsight. My expectation is that all this leads too is labs testing each other’s models alongside independent evaluators before release, with more spending on evaluations, monitoring and logging. This whole “pacing the frontier” push also feels like a way to protect the labs’ margins. A coordinated slowdown could give them more time to monetize each model before competition forces another expensive training cycle. I don’t see that stopping the infrastructure buildout, especially with B200 rental prices still near March highs and BofA raising its semiconductor growth outlook. My biggest concern is how they get China to participate. Amodei himself says global pacing would be “much harder to achieve.” Even if China agrees, how does either side verify that the other is following the rules? I don’t expect the US to meaningfully slow down either way, because I can’t see Washington or the labs accepting a pause that lets China catch up.