AI infrastructure spending amid calls to slow model development

In a September 15, 2026 post, a user predicts the push to pace AI development will lead mainly to labs testing one another’s models alongside independent evaluators, plus more spending on evaluations, monitoring and logging. They argue a coordinated slowdown could give labs longer to monetize each model before another expensive training cycle. Citing B200 rental prices near March highs and BofA’s raised semiconductor growth outlook, they don’t expect infrastructure buildout to stop. Their biggest concern is whether China would participate and how either side could verify compliance; they don’t expect the U.S. to meaningfully slow down.