American households, manufacturers, and grid operators face a sharp increase in electricity demand over the next five years, with data centers accounting for roughly half the surge. Federal and international energy agencies project U.S. power consumption will grow by more than 420 terawatt-hours through 2030, a pace that would strain generation capacity, push wholesale prices higher in key regions, and force difficult choices about fuel mix and grid investment. The numbers are already moving fast: data-center electricity use tripled between 2014 and 2023, and government forecasts show it could triple again by 2028.
Why a 420 TWh demand spike changes the math for grid operators
The scale of the projected increase matters because U.S. electricity demand was essentially flat for more than a decade before 2020. That long plateau allowed utilities to retire aging coal plants and add renewables without major reliability concerns. A jump of more than 420 TWh, as the IEA demand outlook projects, would be equivalent to adding the annual electricity consumption of a mid-sized European country onto the American grid in roughly five years.
Data centers sit at the center of this shift. The U.S. Department of Energy found that data centers consumed 176 TWh in 2023, up from 58 TWh in 2014, and already represented about 4.4 percent of total national electricity use. The DOE’s projections place data-center consumption between 325 and 580 TWh by 2028, meaning the sector alone could account for 6.7 to 12 percent of all U.S. electricity within a few years, according to the agency’s data-center assessment.
The hypothesis that wholesale electricity prices in the PJM and ERCOT markets will rise at least 15 percent above current federal baselines by 2027 rests on a specific condition: data-center load hitting the DOE’s upper-bound trajectory of 580 TWh by 2028. If that scenario materializes, even announced natural gas plant additions would struggle to keep pace. The EIA has identified ERCOT, the Texas grid operator, as the region likely to see the fastest data-center load growth through 2027. PJM, which covers 13 states and the District of Columbia, has already seen data centers drive nearly all net demand growth since 2020. Both regions operate competitive wholesale markets where tight supply-demand balances translate directly into higher clearing prices for generators and, eventually, higher bills for end users.
No federal agency has published a point estimate confirming a 15 percent wholesale price increase tied specifically to data-center load. But the directional pressure is clear. The EIA’s analysis of fossil generation trends notes that faster-than-expected data-center growth could increase reliance on gas-fired power plants, which set the marginal price in most U.S. wholesale markets. When demand rises faster than new supply comes online, prices follow.
Federal data and IEA forecasts anchor the 30 percent growth claim
Three primary federal and international sources converge on the same conclusion: U.S. electricity demand is entering a period of growth not seen in decades, and data centers are the principal driver.
The IEA’s Electricity 2026 report projects that U.S. electricity use will add more than 420 TWh over five years, with data centers expected to contribute about 50 percent of that growth through 2030. That 420 TWh figure, applied against the roughly 4,000 TWh the United States consumed annually in recent years, implies an increase of more than 10 percent from data centers alone and well above 20 percent when industrial electrification, electric vehicles, and other loads are included.
The DOE’s own numbers reinforce the trajectory. Data-center electricity consumption grew more than threefold between 2014 and 2023, from 58 TWh to 176 TWh. The department’s upper-bound projection of 580 TWh by 2028 would represent a further tripling in just five years. Even the lower bound of 325 TWh would nearly double 2023 levels. Research from federal energy analysts ties this growth directly to hyperscale cloud computing, artificial intelligence training clusters, and cryptocurrency operations, all of which require dense, continuous power supplies.
The EIA’s Annual Energy Outlook 2026 separately models data-center server electricity use across the commercial building stock, treating it as a distinct and growing category within its long-term demand projections. That modeling choice reflects how significantly the federal statistical agency views data-center load as a structural shift rather than a temporary spike.
Grid reliability gaps and unanswered price questions
Several important questions remain unresolved despite the strength of the demand forecasts. First, the exact split between new gas-fired generation and renewable additions that will serve data-center load is still unclear. Many hyperscale operators, including Microsoft, Google, and Amazon, have signed large renewable energy procurement agreements, but the physical electricity their facilities consume often comes from the local grid mix, which in ERCOT and PJM still relies heavily on natural gas and, in some areas, coal. The timing mismatch between when new data centers connect and when new wind, solar, storage, or nuclear capacity actually comes online will be a key determinant of both emissions and prices.
Second, transmission constraints could magnify price increases in specific zones. Even if national generation capacity keeps pace on paper, bottlenecks in getting power from remote wind and solar resources to urban data-center clusters can force grid operators to dispatch more expensive local plants. That, in turn, raises locational marginal prices and can produce sharp regional differences in wholesale and retail rates. PJM has already flagged interconnection backlogs and local reliability concerns in Northern Virginia, one of the world’s largest data-center hubs.
Third, there is uncertainty around how much demand-side flexibility data centers will ultimately provide. Some operators are experimenting with shifting non-urgent computing tasks to off-peak hours or to regions with surplus renewable generation. Others are testing on-site backup generation, including gas turbines and fuel cells, as a way to reduce grid stress during peak periods. Yet most of the largest AI training and real-time cloud workloads require high uptime and low latency, limiting the degree to which they can be dialed back when the grid is tight.
These unresolved issues make it difficult to translate the DOE and IEA demand projections into precise price forecasts. A scenario in which data centers aggressively adopt flexible load practices, pair facilities with large-scale storage, and locate in regions with abundant renewables would look very different from one in which most new capacity is served by gas plants in already constrained markets. The 15 percent wholesale price increase hypothesis in PJM and ERCOT is therefore best understood as a stress-test of what could happen if high-end data-center demand materializes without a commensurate build-out of low-cost generation and transmission.
Policy responses and investment choices
Policymakers and regulators are beginning to grapple with how to manage this surge in electricity demand without sacrificing reliability or affordability. State utility commissions are weighing whether to require large data centers to contribute more directly to grid upgrades, either through higher interconnection fees or targeted infrastructure investments. Regional grid operators are revising their long-term planning processes to incorporate higher load forecasts and to encourage resources that can respond quickly to changing conditions.
At the federal level, agencies are using the new demand projections to stress-test existing policies. Higher-than-expected data-center growth could accelerate timelines for transmission planning, influence tax-credit allocation for clean energy projects, and shape reliability standards. The same projections are also informing debates over advanced nuclear, long-duration storage, and hydrogen as potential complements to wind and solar in serving around-the-clock digital loads.
For investors, the message embedded in the DOE, EIA, and IEA data is that the era of flat U.S. electricity demand is over, at least for the coming decade. That shift creates opportunities in generation, grid infrastructure, and efficiency technologies, but it also raises the stakes for getting the fuel mix and market design right. If supply fails to keep up, or if new capacity is concentrated in higher-cost resources, the result could be sustained upward pressure on wholesale prices in key markets.
Ultimately, the 420 TWh increase in projected U.S. electricity demand is not just a statistical footnote. It signals a structural transformation in how power is produced and consumed, driven in large part by the rapid expansion of data centers that underpins the digital economy. Whether that transformation leads to a more resilient, cleaner grid or to higher prices and reliability risks will depend on decisions being made now by regulators, utilities, technology companies, and grid operators in response to the accelerating load growth already visible in the federal and international data.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
