Introduction

In April 2025, following a two-year hiatus, the U.S. Energy Information Administration (EIA) released its 2025 Annual Energy Outlook (AEO) which forecasts higher natural gas consumption over the next decade compared to prior AEO forecasts.^[1] More notably, it doubles the long-term growth rate in power consumption. The growth of natural gas consumption and electrical power generation is driven, in part, by anticipated construction of additional data centers and liquified natural gas (LNG) facilities that require external sources of electrical power.

The 2025 AEO presents the most bullish power consumption forecast published by the EIA in the past two decades. Despite this optimistic outlook, the EIA projections remain conservative when compared to other published forecasts-particularly those published by Independent System Operators (ISOs) and Regional Transmission Operators (RTOs)-which anticipate significantly higher growth in power consumption, primarily driven by the expansion of data centers.

This series of articles explores the sources of rising demand for power, including data centers and new LNG plants, and assesses their impact on total gas consumption.

The EIA Has Historically Underestimated Gas Consumption

The EIA’s AEO has consistently underestimated U.S. natural gas consumption. According to the EIA itself, in its 2022 retrospective review, 16 out of the then prior 17 AEOs underestimated domestic gas consumption.^[2]

This tendency to underestimate natural gas consumption is illustrated in the table below that compares AEO forecasts for 2024 to the year’s actual consumption.

The penultimate forecast, published in the 2023 AEO, underestimated 2024 natural gas consumption by 3.5 trillion cubic feet (Tcf), an error of approximately 10.5%. To further illustrate this pattern, the chart below compares actual U.S. gas consumption (dark gray line) with all historical AEO forecasts (green lines) since 2005.

Most outlooks fell well below actual reported gas consumption. A notable example of this trend is the 2010 AEO (dark green line), exemplifying an overly bearish forecast that incorrectly projected negative or minimal growth over the following12 years.

As shown in the chart below, the EIA increased its forecast of U.S. natural gas consumption in its 2025 AEO versus the previous 2023 publication. Nonetheless, the projected growth in the 2025 forecast is significantly lower than actual consumption growth from 2005 to 2024. It is important to note that domestic gas consumption does not include natural gas that is liquified and exported as LNG.

The EIA Has Historically Overestimated Power Consumption

While the EIA has consistently underestimated natural gas consumption, it has often overestimated total power consumption. The chart below compares actual U.S. power consumption (dark gray line) against historical AEO forecasts in trillion watt hours (TWh) per year. Most AEO forecasts exceeded actual consumption.

A notable example is the 2002 AEO (yellow line), which represented an overly bullish forecast which incorrectly assumed that prior 10-year growth trends would continue. Instead, power consumption stagnated after 2007, primarily due to deindustrialization and improvements in energy efficiency.^[3] This overly bullish 2022 forecast serves as a warning-forecasts can get ahead of themselves predicated on growth narratives that fail to materialize.

The chart below compares historical U.S. Power Consumption along with the 2023 and 2025 AEO forecasts.

The latest 2025 AEO projects solid power growth of 1.60% per year over the next 15 years. This represents a stronger outlook than the near-zero growth observed from 2007 to 2020 but remains weaker than the 2.4% growth observed from 1993 to 2007.

Importantly, the 2025 AEO forecast more than doubles the 15-year growth rate of 0.70% underpinning the 2023 AEO, indicating that the EIA-like many within the industry-anticipates increased levels of demand, largely the due to increased demand for further data center development and other industrial build-outs.

The EIA Has Historically Underestimated Gas Consumption for Power Generation

Despite overestimating power consumption, the EIA has consistently underestimated natural gas consumption for electricity generation. This cause of this discrepancy is likely the result of discounting natural gas’s projected market share of generation, often in favor of a combination of coal, nuclear power, and more recently, renewables. Primary reason include:

• Failure to anticipate the shale gas revolution, which resulted increased productivity allowing for increased gas supply to low prices.
• Underestimating utilization rates of natural gas plants, which are inherently more flexible than intermittent sources.
• Overlooking efficiency losses when additional natural gas-powered plants designed for peak loads come online, which negates efficiency gains from newer, highly efficient power plants designed for baseload capacity.

As renewable electricity generation expands, natural gas generation facilities remain crucial for grid stability and load balancing, due to the intermittent availability of renewable generation.

System Operators Forecast Major Increases in Power Consumption

Independent Service Operators (ISOs) and Regional Transmission Organizations (RTOs) manage most of the power generated in the U.S.^[4] These operators regularly update consumption forecasts for planning purposes. Two key system operators-ERCOT and PJM-have significantly increased their outlooks in recent years.

ERCOT 2025 Forecasts

ERCOT’s latest long-term forecast anticipates electricity consumption rising from 462 TWh in 2024 to 1,119 TWh by 2034, more than doubling within a decade. As shown in the chart below, a major shift in the forecasts occurs in 2024.

Unlike previous forecasts, that emphasized electric vehicle adoption driving demand, the latest outlook attributes demand growth primarily to industrial consumption—especially from data centers, LNG facilities, and hydrogen production.

ERCOT’s forecast incorporates input from transmission service providers, including formal requests and signed contracts from end users. To reflect a more realistic outlook, ERCOT applied a substantial haircut of more than 50% based on historical realization rates. The forecast shown in the chart above represents these adjusted values.

While there is debate over how much adjustment is appropriate, even with adjustments the forecast implies significant infrastructure expansion, raising questions about feasibility and timing.

What is not in debate is that there is strong demand for data centers, which if realized will add to ERCOT’s power generation requirements in the foreseeable future.

PJM 2025 Forecast

PJM Interconnection, which covers the Mid-Atlantic and parts of the Midwest, has revised its forecasts upward for four consecutive years. The 2025 forecast is its most aggressive forecast to date, projecting 68% growth in consumption over 16 years as demonstrated in the figure below.

The PJM area is divided into multiple zones as shown in the figure below.^[5]

The PJM area is divided into multiple zones as shown in the figure below.^[5]

The latest PJM forecast show significant increases outside the high-growth Dominion Zone, which had previously been responsible for most of the projected growth. Located near Washington, D.C., the Dominion Zone proved to be an early indicator of power demand growth when the Dominion Zone's forecast increased significantly in the 2022 and 2023 forecasts.

In contrast, the PJM 2024 forecast projected 35% growth over 16 years, with 58% of that growth attributed to the Dominion Zone. PJM's 2025 forecast, however, projects 68% growth over 16 years, with only 35% of demand growth from the Dominion Zone, as represented in the figure below.

The publication of the 2025 forecast highlights the shift to demand growth from zones outside Dominion, in particular the AEP Zone (Ohio and West Virginia) and PPL (Eastern Pennsylvania).

The change in growth drivers suggest that data center growth is expanding into regions such as Pennsylvania and West Virginia, that offer easier land acquisition, permitting, and proximity to major gas supply zones (e.g., the Marcellus Shale) while remaining close to centers of demand and population. There are similar trends emerging in the Midwest given the proximity to Utica Shale and numerous gas pipelines.

Combining ERCOT and PJM

Considering only the latest consumption growth forecasts from ERCOT and PJM—and not the rest of the country—U.S. power consumption would increase by 2.5% annually over the next decade. That rate of growth is significantly higher than the EIA’s forecasted 1.6% annual growth.

Data Center Demand For Electricity Will Be Concentrated, Baseload and Mostly Filled by Natural Gas Generation

Data centers exhibit two primary characteristics that shape electricity demand:

1. Geographic Concentration

   Data centers favor locations with robust broadband infrastructure, while also weighing power costs, permitting feasibility, and available land. Texas continues to attract significant data center investment due to these factors. On the East Coast, Pennsylvania and West Virginia are emerging as alternatives to Northern Virginia, where costs continue to rise. The Midwest is also experiencing growth for similar reasons.

2. Steady Baseload Demand

   Unlike residential or commercial loads, data center electricity usage is stable and continuous. Most consumption comes from computing hardware and cooling systems, resulting in minimal daily load variability—except for modest increases to adjust for extreme temperatures.

To meet this type of constant demand, natural gas-fired power plants are well-suited. They operate efficiently in a baseload configuration, providing a reliable, efficient, and steady generation. While renewables like wind and solar can support data center power requirements, they must be paired with dispatchable power generation, typically natural gas powered generation.

Nuclear energy is often proposed due to its low-carbon profile and steady baseload output. However, long development timelines, siting challenges, cost overruns, and delays make nuclear an impractical near-term solution. Emerging technologies like small modular reactors (SMRs) are promising but remain unproven as an economic solution.^[6] Over the next decade, natural gas is the most likely candidate to meet the need for additional power generation.

Data Centers Outside ERCOT and PJM

Beyond Texas and PJM, data center projects are being planned throughout the U.S. The West Coast—home to major population hubs and commercial centers like Silicon Valley—has significant demand. However, high energy costs and expensive site development pose challenges for locating data centers near these high-demand areas. Exceptions are smaller “edge” data centers that prioritize location over cost.

Parts of the Midwest outside PJM territories are also expected to see increased data center development. The development of additional data centers in the South and Mountain regions are be sporadic and influenced by the availability of supporting infrastructure and favorable permitting regulations.

Natural Gas Consumption to Soar

Rising electricity demand from data centers is expected to increase the need for natural gas-fired generation significantly. The chart below compares the EIA’s current forecast for natural gas generation with an alternative forecast that incorporates the expected load growth from ERCOT and PJM. For other U.S. regions, this forecast conservatively assumes their total nominal growth will match the combined nominal growth from ERCOT and PJM.

This alternative scenario includes the continued expansion of renewable generation, while assuming natural gas maintains its existing current market share of approximately 40% of total U.S. power generation.

As a result of the growth of power requirements, the consumption of natural gas in power generation must rise to meet demand.

Concluding Thoughts

The outlook for data centers points to sustained growth in both power-sector natural gas use and overall natural gas consumption. While the EIA has acknowledged this trend by adjusting its most recent forecast upward, its projections remain more conservative than those from the two prominent regional power operators.

In the following article in this series, we will examine another major driver of future demand—LNG export growth. This will not only require more natural gas for direct export once converted to LNG, but also increase electricity demand from power operators, particularly for LNG facilities that rely on external supply of power rather than self-generation.

[1] In 2024, for the first time since 1998, the EIA did not publish its AEO.

[2] U.S. Energy Information Administration - EIA - Independent Statistics and Analysis – Averaging forecasts compared to actuals for each available year from 2006 through 2021.

[3] U.S. Energy Information Administration - EIA - After more than a decade of little change, U.S. electricity consumption is rising again.

[4] The ISO/RTO Council (IRC) states that ISOs and RTOs manage about two thirds of power generated in North America.

[5] The ComEd Zone in Illinois is not shown.

[6] The collapse of NuScale’s project should spell the end for small modular nuclear reactors, M.V. Ramana, Utility Drive, January 31, 2024.