The energy math is getting harder to ignore. Analysis published March 24, 2026, by NC State University’s FREEDM Systems Center, citing International Energy Agency data, projects that global AI data center energy consumption will grow approximately 95% between 2024 and 2030, rising from an estimated 485 terawatt-hours to roughly 945 terawatt-hours. That 945 TWh figure represents approximately 3% of projected global energy demand. To put it differently: AI’s data center footprint alone could consume more electricity than most countries generate today.
These are projections, not certainties. The IEA’s modeling has been revised before, and the pace of efficiency improvements in AI hardware could alter the curve. But the directional signal is consistent across multiple research institutions, and NC State’s workshop on AI data center energy needs, convened in March 2026, reflects serious institutional engagement with this question, not speculative commentary.
The policy pressure is building alongside the power demand. According to reporting from Cornell University’s news service, lawmakers have reportedly introduced a bill proposing a moratorium on new AI data center construction, citing community-level grid stress and resilience concerns. That bill claim could not be independently verified in this cycle: the Cornell source URL was inaccessible during the research and verification process. The hub reports it here with the appropriate qualifier. If confirmed, it would represent the first significant federal legislative push against AI data center expansion. If you’re tracking this story, watch the Cornell reporting for confirmation.
Foley and Lardner’s March 2026 webinar summary on AI, energy, and data centers provides a useful legal-sector signal: the law firms that advise infrastructure developers are already briefing clients on this question. That’s not leading-edge policy analysis, it’s a lagging indicator. By the time major law firms are running client webinars, the risk is already in the room.
For infrastructure investors and enterprise AI planners, the practical implication is simple. Energy cost and availability are no longer downstream considerations in data center siting decisions. They’re the lead variable. Communities and utilities that can offer reliable, cost-competitive power, ideally with a clean energy story, are going to see competition for that capacity. Communities that can’t are going to see the legislative backlash the Cornell report describes.
What to watch: the status of any federal moratorium legislation (bill number and sponsoring lawmakers not confirmed at time of publication); state-level utility commission proceedings in the Southeast and mid-Atlantic where data center concentration is highest; and IEA’s next modeling update, which will either reinforce or revise the 945 TWh projection.
TJS synthesis: The energy question isn’t a future problem for AI infrastructure. It’s a present constraint that hasn’t fully hit asset prices yet. Projects that baked in optimistic energy cost assumptions are going to face margin pressure as grid access tightens. The policy environment will follow the economics, which means the window to site and permit new capacity under the current regulatory regime may be shorter than the 2030 timeline suggests.