US Data-Center Power Forecast Points to a Potential Supply Gap for AI Growth
Редакция ЦОД:Главное
A TechNews review of Morgan Stanley and Citadel Securities estimates points to power availability, rather than compute equipment alone, becoming a binding constraint on US data-center and AI expansion. The cited forecasts place new US data-center electricity demand at 97 GW in 2026–2028, with a residual supply gap of about 33 GW even after alternative supply options are considered.
- Morgan Stanley’s cited forecast projects US data-center IT electricity demand rising from 9.19 GW in 2025 to 78.57 GW in 2029.
- New US data-center electricity demand is estimated at 97 GW for 2026–2028.
- The projected shortfall is 57 GW before, and about 33 GW after, specified alternative supply options are included.
- Grid interconnection, permitting and local opposition are identified as constraints on large-scale AI deployment.
- All figures are forward-looking estimates rather than operational capacity or actual consumption.
Power availability, grid interconnection capacity, permitting and local opposition are emerging as material constraints on US AI and data-center deployment, according to estimates cited by TechNews from Morgan Stanley and Citadel Securities.
Morgan Stanley’s cited forecast puts US data-center IT electricity demand at 9.19 GW in 2025, rising to 78.57 GW in 2029—a projected increase of 755% over four years. Separately, the article estimates that new US data-center electricity demand could reach 97 GW during 2026–2028.
After accounting for generation capacity under construction and available grid capacity, the initial projected shortfall is 57 GW. The estimate falls to approximately 33 GW after including alternative supply options: on-site natural-gas generation, fuel cells and nuclear-related projects. These are forecast supply-demand gaps, not commissioned capacity or measured electricity consumption.
The article also states that US data centers could consume up to 12% of national electricity demand by 2028. Citadel Securities identifies grid connection, permitting processes and local resistance as key obstacles to AI deployment at scale.
For developers, hyperscalers and colocation operators, the analysis suggests that securing deliverable power may increasingly determine whether AI capacity can be brought online on schedule. If the projections prove accurate, power-secured campuses, on-site supply and load-management approaches could receive greater emphasis, while grid constraints may defer portions of the announced data-center pipeline.
If these forecasts materialise, access to firm power and the speed of grid interconnection could become central determinants of US AI data-center competitiveness. This could shift development priorities toward power-secured sites, on-site supply options and load-management solutions, while delaying some proposed data-center projects.
- TechNews 科技新報 / КитайскийПервоисточник →
Материал подготовлен редакцией ЦОД: Главное на основе указанных первоисточников.