The U.S. government formally recognizes a grid rewrite
The constraint on AI data centers has shifted from semiconductors to power—an industry observation now supported by a U.S. government document. DOE’s Grid Deployment Office has published a draft 2026 National Transmission Needs Study and opened a 60-day public-comment period. The draft explicitly identifies data centers, manufacturing expansion, and large industrial loads as key causes of transmission-infrastructure constraints.
The numbers show demand accelerating
The draft’s figures are clear. U.S. data centers used 176TWh in 2023, or 4.4% of total electricity consumption. By 2028, that is projected to double or triple to 325–580TWh, reaching as much as 12% of total electricity use. Data-center electricity consumption has already tripled over the past decade, and AI demand is a major driver.
This acceleration is putting pressure on transmission. According to DOE, most congestion is concentrated in about 5% of all hours, becoming pronounced during large price spreads, high net load, and cold conditions. Constraints are already visible in regions such as NYISO, raising reliability concerns. Even if chip supply is available, AI infrastructure cannot run unless the transmission grid can deliver power to the rack—the government’s demand estimate now supports that reality.
At the core of the “three walls”
Longer transformer lead times, extended grid-interconnection queues, and regulation and permitting are the three walls repeatedly examined on this site. The DOE draft shows that these walls are a national planning issue, not merely an individual-company observation. DOE aims to establish a clear direction for transmission planning through cooperation with communities, industry, states, Tribes, and reliability organizations, placing transmission infrastructure at the center of policy.
The important point is that this defines investment and procurement timelines. Transmission expansion is a multi-year to decade-long project, and its gap with the accelerating pace of AI-driven data-center construction becomes a constraint on site selection and power availability.
Data centers lead demand
DOE identifies data centers, manufacturing expansion, and large industrial loads as key causes of transmission constraints.
Use doubles or triples
U.S. DC consumption rises from 176TWh (4.4%) in 2023 to 325–580TWh (up to 12%) in 2028.
Congestion is time-concentrated
Most congestion is concentrated in about 5% of hours, with constraints visible in NYISO and elsewhere.
Planning takes years to decades
Transmission expansion is long-cycle work. The gap with data-center construction constrains siting and power procurement.
Business implications and checkpoints
Those responsible for data-center power and site selection should check: (1) transmission constraints and interconnection queues in the relevant grid, including NYISO; (2) how concretely transmission expansion plans and schedules address the rapid growth in electricity use, since conservative power plans may be insufficient against the government’s upward demand estimates; and (3) the room for demand balancing through onsite generation, storage, and demand response, given that congestion is concentrated in particular hours. The view that the bottleneck is moving from chips to the transmission grid is now supported by an official government document.
Key grid constraints and transformer supply developments are covered continuously in related articles.