EPRI Projects Data Centers to Hit 17% of U.S. Electricity by 2030

U.S. data centers consumed 177 to 192 terawatt-hours of electricity in 2024. EPRI projects that figure to reach 380 to 790 TWh by 2030, representing 9 to 17 percent of total U.S. electricity consumption. The February 2026 estimates are 60 percent higher than EPRI 2024 projections, driven by AI infrastructure construction over the past 18 months.

Scenario Range and Assumptions

EPRI built three scenarios based on how many announced projects reach commercial operation. The low scenario (380 TWh, 9 percent of U.S. electricity) assumes most projects under construction open but only 25 percent of those in advanced planning proceed. The high scenario (790 TWh, 17 percent) assumes all under-construction and advanced-planning projects open plus 30 percent of early-stage projects. The wide range reflects genuine uncertainty about which projects will secure grid connections, financing, and regulatory approvals on schedule. Data Center Knowledge noted that U.S. grid interconnection queues contain over 100 GW of data center projects seeking connection through 2030.

Grid and Infrastructure Implications

Adding 200 to 600 TWh of new data center load by 2030 requires new transmission, generation, and distribution infrastructure in concentrated geographic clusters. Northern Virginia, Hillsboro, Phoenix, and Dallas-Fort Worth host the highest densities of announced projects. Local utilities in these markets face transformer shortages, substation upgrade backlogs, and interconnection queue constraints. EPRI analysis calls for accelerated coordination between data center developers, utilities, and grid operators to avoid localized reliability problems as individual campuses grow past 1 GW of connected load.

Critical Analysis

Adding 380 to 790 TWh of data center load through 2030 introduces millions of switched-mode-power-supply rectifiers to the grid, each producing predominantly nonlinear current. Concentrated clustering (Northern Virginia, Phoenix, Dallas-Fort Worth) strains local distribution infrastructure: 138 kV and 230 kV substations serving single campuses now routinely exceed 500 MVA connected load, requiring dedicated step-down transformer banks and reactive-power compensation that the original substation designs did not anticipate.

Critical Perspective

EPRI 60% upward revision in 18 months illustrates the forecasting problem: the 790 TWh high-end scenario requires U.S. data center demand to grow 4.5x in six years, a rate with no precedent in any sector of American electricity consumption. Announced AI capacity converts to operating load at a rate the current interconnection queue directly constrains; PJM, MISO, and CAISO together hold 1,500 GW of queued generation needed to serve this load, with average queue processing times of five years. Lawrence Berkeley 2024 data center study projected 325 TWh by 2028 using more conservative utilization assumptions, a figure 2.4x lower than EPRI high-end for two years later. The question energy planners should be asking: which specific generation assets are under executed interconnection agreements today to supply the incremental 600 TWh between EPRI 2024 and 2026 forecasts?

Related Coverage

• IEA: U.S. Data Center Electricity Use to Surge 130% by 2030 — IEA projects similar surge in US data center electricity use
• U.S. Data Center Power Demand Projected to Reach 106 GW by 2035 — US data center power demand projected to reach 106 GW by 2035
• Data Center Power: How the Cloud Transition Is Driving 12GW — Cloud transition driving data center power demand growth
• PJM Projects 49 GW Shortfall by 2028 as AI Data Centers Surge — AI data centers surge, projecting grid shortfall in PJM