American electricity consumers and grid operators face a direct challenge from the fastest-growing single source of power demand in the country. Data centers now pull roughly 42 gigawatts across the United States, nearly double the load recorded three years ago. The sector already accounts for 4 to 5 percent of total U.S. electricity generation, and federal research warns that share could quadruple by the end of the decade if construction plans proceed as announced. The strain is not abstract: utilities in several regions are now weighing whether to extend the life of coal and natural gas plants that were slated for retirement, raising hard questions about emissions, electricity prices, and grid reliability for households and businesses that have nothing to do with cloud computing.
Why 42 gigawatts of data-center load changes the grid calculus
The speed of the increase matters as much as the size. Three years ago, data centers drew roughly 22 gigawatts. Reaching 42 gigawatts this quickly has outpaced the planning cycles that utilities and regional transmission organizations use to build new generation and upgrade power lines. That mismatch is already producing real friction. The Federal Energy Regulatory Commission’s 2025 State of the Markets assessment placed operating data-center capacity at approximately 50 gigawatts by the end of 2025, a figure that includes both the electrical draw and the overhead capacity built into each facility. The gap between the 42-gigawatt consumption figure and the 50-gigawatt capacity total reflects the fact that many facilities run below their maximum rating, but new hyperscale campuses are filling in fast.
The practical consequence is straightforward. When load grows faster than clean generation can be permitted, financed, and connected, the slack gets picked up by existing fossil plants. Analysis from the U.S. Energy Information Administration warns that fossil generation could rise in regions where data-center demand outpaces new renewable and nuclear capacity. That finding points toward measurable upward revisions to regional coal and gas capacity factors in upcoming EIA Short-Term Energy Outlook updates, particularly in NERC sub-regions that host the densest clusters of hyperscale construction, such as PJM Interconnection territory in Virginia and parts of ERCOT in Texas.
For residential and commercial ratepayers, the risk is concrete. Utilities that must keep aging gas or coal units online longer than planned pass those costs through rate cases. Industrial customers competing for the same transmission capacity face longer interconnection queues. And state clean-energy targets become harder to hit when the denominator, total electricity demand, keeps climbing. The same servers that power artificial intelligence models, streaming platforms, and enterprise software can therefore shape electricity bills and climate trajectories far from the tech campuses where expansion decisions are made.
Federal research quantifies the demand surge
Two independent federal-aligned analyses anchor the numbers behind this trend. The Electric Power Research Institute released its Powering Intelligence 2026 study in February, estimating that data centers currently consume 4 to 5 percent of U.S. electricity generation and could reach up to 17 percent by 2030 under a scenario that includes operational capacity, facilities under construction, and publicly announced expansion plans. That 17-percent figure represents a high-end scenario, not a baseline forecast, and EPRI’s methodology accounts for the difference between projects that have broken ground and those that exist only as corporate announcements.
Separately, the Department of Energy released its own evaluation of data-center electricity demand, drawing on research conducted by Lawrence Berkeley National Laboratory. The LBNL 2024 report provided historical consumption data in terawatt-hours and a range of forward projections. Together, the EPRI and LBNL analyses establish a two-source confirmation that the sector’s electricity appetite is growing at a pace not seen from any single demand category since the post-war industrial expansion.
The EIA’s modeling work adds a third dimension. Rather than focusing solely on how much power data centers will need, the agency examined what happens to the generation mix when that demand arrives faster than expected. Its analysis connected accelerating regional load growth, explicitly including data centers, to outcomes for coal and gas plant utilization. The Short-Term Energy Outlook framework uses scenario-based approaches to test how different demand trajectories change the dispatch order of power plants across NERC regions. When load rises sharply, plants that were running at low capacity factors, or scheduled for decommissioning, get called back into heavier service. That dynamic has already begun to surface in capacity-market auctions and integrated resource plans, where utilities now model higher baseload requirements than they did just a few years ago.
Gaps in the data and what to watch through 2027
Several important questions remain open. No primary federal source in the current public record supplies a detailed methodology behind the specific 42-gigawatt consumption figure or breaks it down by region with granular facility-level data. FERC’s State of the Markets materials reference approximately 50 gigawatts of operating capacity, but the distinction between nameplate capacity and actual consumption at any given hour depends on utilization rates that vary by facility type, workload, and cooling efficiency. Until regulators publish facility-level demand data, policymakers and communities must rely on a patchwork of estimates, industry disclosures, and regional planning documents.
That lack of granularity matters because the impacts are highly localized. A single hyperscale campus can draw more power than a mid-sized city, but the consequences for prices and emissions depend on which generators sit on the margin in that particular grid pocket. In regions where new solar, wind, and battery projects can be built and interconnected quickly, data-center growth may be met largely with clean resources. In areas with constrained transmission or slow permitting, the same growth can lock in higher fossil utilization for a decade or more. Without clear, standardized reporting on data-center load by region and time of day, regulators will struggle to distinguish between these outcomes.
Through 2027, several indicators will show whether the United States is bending this curve toward cleaner outcomes or drifting into a higher-emissions path. One is the pace at which new renewable and storage capacity actually connects to the grid, not just how much is announced. Interconnection backlogs have already delayed many projects that could otherwise serve rising digital demand. Another is the trajectory of capacity factors for coal and older gas plants in regions with heavy data-center construction. If those factors rise persistently, it will signal that new load is being met by legacy fossil assets rather than by incremental clean generation.
Policy responses will also be critical to watch. Some states are exploring requirements that large data centers procure a share of their electricity from new renewable projects, rather than relying solely on unbundled renewable energy credits. Others are considering location-based incentives that steer new facilities toward grid nodes with abundant clean power and available transmission capacity. At the federal level, more detailed reporting on large-load interconnections and standardized metrics for data-center efficiency could give planners the tools they need to align digital expansion with decarbonization goals.
Ultimately, the 42-gigawatt figure is less a destination than a waypoint. Data centers are poised to remain one of the dominant forces shaping U.S. electricity demand for the rest of the decade. Whether that growth deepens reliance on fossil generation or accelerates investment in cleaner grids will depend on decisions made now by utilities, regulators, and the companies building the next wave of computing infrastructure. For consumers, the stakes will show up not just in monthly bills, but in the resilience and carbon intensity of the power system that keeps the digital economy running.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
