The U.S. Energy Information Administration EIA is forecasting a dramatic surge in electricity consumed by data center servers over the coming decades, with the agency's Annual Energy Outlook 2026 projecting that servers alone could account for between 22% and 33% of all commercial building electricity use by 2050, up from an estimated 7% in 2025.

Projected Growth Across Two Scenarios

The AEO2026 report, released, presents two primary cases for assessing the trajectory of data center energy demand. In the Counterfactual Baseline case, server electricity consumption reaches 446 billion kilowatthours by 2050.

In the High Electricity Demand case, that figure climbs to 818 billion kilowatthours, nearly double the baseline projection.

The High Electricity Demand case reflects faster growth in server power draw and a larger installed stock of servers over the projection period.

Standalone data centers, which are represented within the "other buildings" category in EIA's Commercial Demand Model, are expected to drive the majority of that growth.

In the High Electricity Demand case alone, servers in standalone data centers are projected to consume 581 billion kilowatthours in 2050.

This figure exceeds the total server consumption projected in the Counterfactual Baseline case across all building types, underscoring how significantly assumptions about AI server adoption and efficiency can shape long-range forecasts.

AI Servers and Efficiency Assumptions Drive the Gap Between Cases

The divergence between the two scenarios is explained in part by differing assumptions about artificial intelligence infrastructure and hardware efficiency.

In the High Electricity Demand case, EIA assumes that AI servers will account for a larger share of the installed stock of servers over time compared to the Counterfactual Baseline case.

The agency also does not incorporate any efficiency improvement assumptions beyond historical trends in the High Electricity Demand scenario.

In contrast, the Counterfactual Baseline case assumes that after 2040, servers will become increasingly efficient, with a 10% reduction in average annual operational power draw occurring every three years, on top of historical efficiency trends.

Despite this assumption, continued growth in server installations still drives overall consumption upward in the baseline case, meaning efficiency gains are not projected to reverse the growth trajectory, only moderate it.

Commercial Electricity Intensity Set to Surpass Historical Record

The surge in data center activity is also expected to push a key metric of the broader commercial sector past a longstanding benchmark.

The commercial sector's electricity intensity, measured in kilowatthours consumed per square foot, is projected to exceed the 2003 historical high of 14.9 kilowatthours per square foot for the first time in 2031 or 2032, depending on the scenario.

EIA notes that in the 2003 Commercial Buildings Energy Consumption Survey, the newest commercial buildings, those constructed from 1990 onward, already consumed as much as twice the electricity per square foot as buildings constructed prior to 1959.

Data center servers and the associated end uses that support their operation, including space cooling and ventilation, are identified as the primary drivers pushing commercial energy intensity above that earlier record in the AEO2026 projections.

Cooling Requirements Add Substantially to Total Energy Load

Beyond the electricity consumed by servers themselves, the thermal management requirements of data center operations represent a significant additional burden on the grid.

EIA assumes that space cooling requirements in data center floorspace are as much as 2.9 times more energy intensive than cooling requirements in non-data center floorspace, on average.

The agency quantifies this gap between scenarios in concrete terms: electricity consumption for space cooling in the High Electricity Demand case is 84 billion kilowatthours higher than in the Counterfactual Baseline case in 2050, driven by the need to support more intensive data center operations.

The projections for space cooling are also described as sensitive to assumptions about population migration and weather patterns, with EIA noting that the same population and weather assumptions were applied across both cases when making that comparison.

Model Updates and Load Shape Characteristics

For AEO2026, EIA updated its Commercial Demand Model to report data center server electricity use separately from the broader category of commercial computing.

This methodological change allows for more granular tracking of how server demand evolves relative to other commercial computing end uses.

One distinctive characteristic of data center server demand highlighted in the report is its load shape.

EIA assumes data center servers operate with an essentially flat load profile, meaning demand for electricity to power servers remains consistent across all hours of the day.

This stands in contrast to many other commercial end uses that follow more variable daily consumption patterns, and it has implications for how utilities and grid operators must plan for and manage the incremental load.

Scope and Policy Cutoff

The AEO2026 is EIA's primary long-term energy outlook and was released in April 2026. As with prior editions, the analysis in most cases reflects only laws and regulations in effect as of December 2025.

EIA noted that any legislation, regulations, executive actions, or court rulings issued after that date are not incorporated into the projections. The principal contributor to the data center analysis was identified as Courtney Sourmehi.