For the past two years, the conversation around AI infrastructure has focused on chips, models, and capital. In practice, that isn’t what’s slowing projects down. What matters is electricity–where it comes from, how much is available, and whether it’s already committed. A lot of proposed data center capacity looks good on paper, but only a fraction of it can actually be built on schedule. Grid access takes time. Permits take time. Locking in long-term power pricing takes time, and most projects are only starting that process…