LEARN / REINDUSTRIALIZATION

Why Electricity Demand Is Growing Again

After roughly two decades of flat American electricity demand, forecasts have turned sharply upward and have been revised upward repeatedly. How much of that reflects load that will actually energise is genuinely disputed, and one utility produced a measurement rather than an opinion.

Key points

The turn

American electricity demand was close to flat from the mid-2000s through the early 2020s, as efficiency gains offset economic growth. Planning organisations built forecasts and reserve margins around that assumption, and both generation investment and transmission planning adjusted to a system that was not expected to grow much.

That assumption has been abandoned. The scale of the revision is visible most clearly in NERC’s own successive assessments, read in sequence rather than one at a time.

Data55 → 80 → 132 → 224 GW
Ten-year summer peak demand growth as forecast in NERC Long-Term Reliability Assessments for 2022, 2023, 2024 and 2025. The 2025 figure is a 69 percent increase over the prior year. Source: NERC, 2025 Long-Term Reliability Assessment.

Series and provenance
Agency
North American Electric Reliability Corporation
Program
Long-Term Reliability Assessment
Series
2025 LTRA, published January 2026
Measure
Ten-year summer peak demand growth forecast
Units
Gigawatts of peak demand growth over ten years
Adjustment
Committed projects only
Period
Assessments of 2022, 2023, 2024 and 2025
Latest
224 GW in the 2025 assessment, against 132 GW in 2024, 80 GW in 2023 and 55 GW in 2022
Source tier
Primary
Retrieved
August 9, 2026

NERC counts a large load only once a project advances from speculative and exploratory stages into development commitments, so the series lags real but uncommitted load.

Open the source

Utility filings moved the same way. Grid Strategies found that the total of five-year forward summer peak growth forecasts published by utilities rose from 38 GW in 2023 to 128 GW in 2024. NERC attributes most of the increase to data centers, with industrial electrification and manufacturing contributing.

The measurement problem

Three different numbers circulate and are frequently treated as one. A queue figure counts interconnection requests submitted. A utility forecast counts load a planner expects to serve. Committed load counts projects with a signed agreement and financial exposure behind them. Each is larger than the next.

The gap arises from how large load developers behave. A firm evaluating a site files requests with several utilities across several states for what is functionally one project, negotiates, builds in one location, and frequently leaves the other applications sitting in their queues as live megawatts. The result is the same campus counted in several places at once.

Data30 GW → 5.6 GW
AEP Ohio’s large load interconnection queue after the utility introduced a tariff requiring firm financial commitments to hold a position. More than 80 percent of submitted load did not survive the requirement. Reported by Carbon Direct.

Series and provenance
Agency
American Electric Power Ohio
Program
Large load interconnection tariff
Series
Large load queue before and after a financial commitment requirement
Measure
Effect of a commitment requirement on queue volume
Units
Gigawatts of requested load
Period
Before and after the tariff took effect
Latest
About 30 GW falling to 5.6 GW; more than 80 percent of submitted load did not survive the requirement
Source tier
Utility tariff outcome, retrieved through a secondary compilation rather than the utility filing
Retrieved
August 9, 2026

The Institute has not read the underlying tariff filing or the queue reports directly, and this figure is marked at lower confidence than an agency data release. It measures commitment under one tariff at one utility. A queue that collapses under a financial requirement does not establish that the demand was never real, only that it was not committed there.

Open the source

Measurement noteThe AEP Ohio result is the closest thing available to a controlled test, because the only variable that changed was the cost of holding a queue position. It should not be generalised without care: it measures one utility, at one moment, under one tariff design, and a queue that collapses under a financial commitment does not establish that the underlying demand was never real, only that it was not committed there.

The scale question is visible in Texas. ERCOT reported monitoring roughly 226 GW of large load interconnection requests in December 2025, up from about 63 GW a year earlier, with roughly three-quarters from data centers. The ERCOT system’s historic peak is near 85 GW. A queue more than twice the size of the entire system is not a forecast of anything.

Institute analysisThe useful discipline is to ask which of the three quantities a figure represents, every time one is cited, and to prefer committed load with financial exposure behind it. Where a source will not say which it is using, the figure describes how many applications were filed rather than how much load will arrive, and those two are separated by everything at issue.

Common misconceptions

That queue volume forecasts demand. A queue counts requests. Requests are cheap to file, are filed in parallel across territories, and are rarely withdrawn promptly when a project sites elsewhere.

That phantom load means the growth is not real. Duplicate counting inflates the gross figure and says nothing about the firm component underneath it. Both a large phantom share and substantial real growth can be true at once.

That flat demand for two decades predicts flat demand now. The efficiency gains that offset growth were largely one-time conversions, and the new load is a category that did not previously exist at this scale.

What the evidence says, and where it is contested

ContestedWhether current forecasts overstate is genuinely unsettled and both sides have real evidence. One position points to the AEP Ohio collapse, to ERCOT queue volume exceeding system peak, to utility executives describing double and triple counting, and to a Berkeley Lab study finding that most of twelve Western utilities overestimated demand in the mid-2000s. A second position holds that the queue-realisation record comes from generation interconnection, filed largely by thinly capitalised developers chasing tax credits, and that large load requests from firms with committed capital are a different category with a higher realisation rate. NERC counts only committed projects and has revised upward four times running, which is more consistent with a forecast lagging reality than leading it. The question will be settled by which loads energise.

Related Institute research

How Electricity Demand Is Measured
Why peak, energy and load factor are not interchangeable.

Why Reindustrialization Requires More Electricity
What this means for industrial siting.

Powering the Buildout
The Institute’s demand projections across three sources of new load.

Sources


Reference entry maintained by the Institute for American Manufacturing & Technology. Figures are drawn from primary sources and cited above. Where the Institute states a position rather than a finding, it is marked as such.

Private Professional Network

Institute for American Manufacturing & Technology

Join the IAMT Network. Connect with researchers, policymakers, and industry leaders working on American manufacturing, energy, and technology policy.

Apply for Membership →