The next phase of AI infrastructure development may be shaped less by the amount of electricity a site can secure than by what its operators can do when that electricity becomes constrained. A large power allocation offers an obvious advantage during project planning, but its commercial usefulness changes when grid conditions fluctuate, electricity prices become less favorable, or local supply arrangements impose operating restrictions. AI infrastructure cannot assume that every megawatt will remain equally available throughout its operating life, particularly as electricity networks accommodate competing industrial demand and increasingly concentrated computing loads. Power flexibility therefore deserves consideration as a planned infrastructure capability, with its technical contribution assessed through measurable demand adjustments, workload continuity, and the site’s ability to respond to changing electricity conditions.