A draft federal study concludes that AI data centers are now the dominant driver of electricity demand growth, compelling planners to reconsider where and how the grid expands.
The Department of Energy’s draft 2026 National Transmission Needs Study states the US energy transmission system has entered a new planning era, driven by the rapid rise of AI data centers.
Historically, transmission projects were primarily driven by reliability concerns, aging infrastructure, and the need to connect new generation sources. Now, the DOE says rapid load growth has become the principal force behind new transmission investment, led by AI data centers, domestic manufacturing and other large industrial loads.
For developers pursuing AI campuses measured in hundreds of megawatts, the report validates a change already unfolding in the field. “At 500 MW or more, transmission is not a downstream interconnection issue,” said Ihab Osman, an independent strategist specializing in data center and other mission-critical infrastructure. “The campus itself becomes a transmission-planning event.”
The study identifies transmission needs across 20 regions and synthesizes more than 120 industry studies and planning documents. DOE says its findings will help inform future National Interest Electric Transmission Corridor designations.
AI Becomes the Planning Driver #
The DOE devotes an entire section of the report to AI-driven electricity demand. The agency says utilities, financial analysts and industry researchers now identify AI data centers as the primary driver of electricity demand growth.
The study cites projections from Lawrence Berkeley National Laboratory, EPRI, and McKinsey showing US data center electricity consumption could exceed 400 terawatt-hours by 2030 – more than 10% of total US electricity consumption recorded in 2023.
Lawrence Berkeley National Laboratory projects that data center electricity demand growing at a 13% to 27% compound annual rate through 2028, when data centers could consume 6.7% to 12% of all US electricity.
Those forecasts help explain why utilities are changing how they plan the grid.
“AI loads have become system-shaping infrastructure counterparties,” Osman said.
The DOE cites NERC forecasts showing total US electricity consumption rising 25%, from 4,281 TWh in 2024 to 5,353 TWh in 2034. EPRI projects overall electricity demand increasing 30% to 46% between 2020 and 2035, depending on how quickly data center development accelerates.
Transmission Becomes the Challenge #
The report argues that regions are responding by approving some of the largest transmission portfolios in their history.
DOE points to MISO’s $21.8 billion long-range transmission portfolio, SPP’s $7.7 billion investment program and Texas’ $33 billion commitment to its first 765-kV transmission backbone as evidence that planners are building for future load growth rather than today’s reliability needs.
NERC projects Texas Reliability Entity’s internal demand growing from nearly 89 GW in 2025 to more than 105 GW in 2028. The DOE also identifies Texas as the state expected to experience the largest percentage increase in electricity consumption over the next decade. Virginia and Texas already have the nation’s largest concentrations of data center electricity demand, while Arizona and Oregon are projected to post some of the fastest growth through 2030.
Generation alone, however, no longer guarantees a viable AI campus.
“Generation somewhere in a region is irrelevant to an AI campus if the grid cannot deliver it to the required node, under contingency, by the required energization date,” Osman said.
The findings also explain why new transmission has not eliminated congestion.
The study says the US energized 85,000 circuit-miles of new, upgraded and rebuilt transmission between 2016 and 2024. Despite these efforts, transmission congestion added an estimated $11 billion to wholesale electricity costs in 2023, with the majority of congestion occurring during only 5% of operating hours.
A New Site-Selection Problem #
For operators evaluating future campuses, DOE’s report is not a map of where power is available. It is a map of where transmission will be needed. “A transmission-needs map is not a transmission-availability map,” Osman said.
Developers must still assess whether planned transmission projects are approved, their expected service dates, and their ability to deliver hundreds of megawatts within hyperscale deployment timelines.
“The deliverable megawatt is becoming a geographic product,” Osman said.