Build Data Centers to Slow AI

Rather than trying to choke off compute, policymakers could slow dangerous frontier model development by expanding regulated, transparent compute infrastructure — i.e., build more data centers in places and under rules that enable monitoring, white‑listing, and governance. The move reframes infrastructure expansion as a governance tool: more centralized, permitted capacity can be paired with oversight, procurement strings, and energy planning to shape who can train large models and how quickly. — This reframing makes local permitting, grid planning, and industrial policy central to AI governance debates and shifts attention from bans to who controls and sites the compute.