Bernstein expects U.S. power demand to enter a roughly 3% CAGR phase by 2030

Overview

This report is Bernstein's medium-term outlook on power and the energy transition in the Americas, with the core centered on its proprietary U.S. power supply-demand model and forecasts extended through 2030. The report argues that after many years of low power demand growth, the U.S. is entering a new upcycle in electricity demand jointly driven by AI data centers, manufacturing reshoring, EVs, heat pumps, and macro GDP growth.

Core views

The report's core view is that U.S. power demand growth has already become a market consensus, and the key debate has shifted from 'whether demand will grow' to 'who can capture supply-side margins.' Bernstein expects total demand to rise from about 4,400TWh to about 5,100TWh by 2030, implying a CAGR of about 3%, significantly above the historical growth rate of about 0.35% from 2000 to 2024. Demand growth will appear across commercial, industrial, and residential segments, with data center commercial electricity use, manufacturing reshoring industrial electricity use, and EV and heat pump residential electricity use as the main increments.

Analysis framework

The report uses a proprietary supply-demand model: on the demand side, it forecasts by residential, commercial, industrial, and transportation sectors, and conducts scenario sensitivity analysis on drivers such as data centers, EVs, heat pumps, and manufacturing reshoring; on the supply side, it is based on the EIA late-stage planned projects database, subtracts retired capacity, adds solar capacity adjustments and nuclear restart assumptions, and then estimates generation by capacity factor. If supply is insufficient to match demand, the model assumes the gap is filled by natural gas generation.

Methodology notes

• sector_modelSector-based power supply-demand model

  Forecast U.S. power demand by residential, commercial, industrial, and transportation sectors

  The model uses the 2010-2019 baseline growth rate as the underlying trend and layers on incremental drivers such as data centers, manufacturing reshoring, EVs, and heat pumps to estimate future monthly demand.

• scenario_analysisData center PUE sensitivity analysis

  Estimate incremental data center electricity demand using PUE and capacity growth assumptions

  The base case uses 40GW of added capacity, 1.25x PUE, and about 50% utilization, resulting in about 230TWh of additional demand; the low-end and high-end scenarios are about 200TWh and 270TWh, respectively.

• supply_modelEIA planned project supply model

  Forecast supply based on planned power projects, retirements, capacity factors, and technology adjustments

  The supply forecast uses the EIA late-stage planned projects database and adjusts for solar forward underestimation, nuclear restarts, and natural gas filling the gap.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• GEV, FRVO, NEE, VST, CEG, LNG

  The report assigns an Outperform rating, viewing them as the main group of beneficiaries from U.S. power demand growth and supply constraints.

  Strengths

  They cover opportunities related to future power supply, power equipment, nuclear power, natural gas, or energy infrastructure, and can participate in the profit pool created by rising power demand.

  Weaknesses

  Company-specific differences in valuation, costs, project execution, and regulation require further review of individual company disclosures.

  Comparison

  Compared with the Market-Perform and Underperform lists, this group is considered by the report to have better relative performance potential.

  Risks

  If power demand materializes more slowly than expected, project interconnection is delayed, or fuel prices or policy changes are unfavorable, investment returns may come under pressure.

• BE, ENPH, TE, VG, CQP

  The report assigns a Market-Perform rating, indicating that they are related to the theme but with a more neutral relative return view.

  Strengths

  Some companies may benefit from demand for distributed energy, energy equipment, natural gas, or power infrastructure.

  Weaknesses

  The report does not provide enough evidence in the summary materials to demonstrate a clear relative excess return advantage.

  Comparison

  Compared with the Outperform portfolio, the report has lower confidence in this group's relative market performance.

  Risks

  Changes in demand growth, policy subsidies, equipment prices, financing costs, and project pacing may affect earnings delivery.

• FSLR, ORA

  The report assigns an Underperform rating, indicating weaker expected relative performance under the report's coverage framework.

  Strengths

  They are still related to the solar or renewable generation theme, and long-term industry demand is not entirely negative.

  Weaknesses

  The report holds a cautious or negative view on their relative market performance.

  Comparison

  Within the same power and energy transition theme, they rank behind the Outperform and Market-Perform lists.

  Risks

  If policy, costs, orders, or power price trends are better than assumed in the report, the negative view may be weakened.

• Natural gas generation and the LNG chain

  As the marginal source filling the supply-demand gap, natural gas generation plays the role of balancing power demand in the model.

  Strengths

  When power demand is rising and renewable intermittency and interconnection constraints exist, natural gas has peaking and reliability value.

  Weaknesses

  Over the long term, it still faces pressure from decarbonization policies and renewable substitution.

  Comparison

  Compared with coal generation, natural gas still grows in the report's forecast; compared with solar, its growth rate is lower but reliability is stronger.

  Risks

  Gas price volatility, pipeline constraints, emissions policy, and changes in power market rules may affect profitability.

Key data

• U.S. total power demand forecastFrom about 4,400TWh to about 5,100TWh by 2030, with a 2025-2030 CAGR of about 3%The report believes this growth rate, while lower than some external forecast ranges, is still very high relative to the long-term historical rate.
• Historical power demand growth rate2000-2024 CAGR of about 0.35%; CAGR of about 1.2% since 2019Used to compare the scale of the current demand upcycle.
• Incremental data center demandBase case of about +230TWh by 2030, with a scenario range of about +200TWh to +270TWhBased on 40GW of added data center capacity, a PUE range of 1.1x to 1.5x, and about 50% utilization.
• Commercial electricity demandFrom about 1,500TWh to about 1,800TWh by 2030, with a CAGR of about 3.6% over the next five yearsMainly driven by data centers and AI electricity demand.
• Industrial electricity demandFrom about 1,000TWh to about 1,100TWh by 2030, with a CAGR of about 2% over the next five yearsManufacturing reshoring is expected to contribute about 100TWh of incremental demand.
• Residential electricity demandFrom about 1,500TWh to about 1,700TWh by 2030, with a CAGR of about 2% over the next five yearsEVs and heat pumps each contribute about 70TWh of additional demand.
• Natural gas generation demandUnder the base case, it increases by about 1bcfd per year by 2030If the high-end data center demand scenario materializes, the incremental gas demand for power generation could rise to about 2bcfd/year.
• Supply mix forecast2025-2030 solar CAGR about 21%, wind about 3.6%, renewables overall about 9%, natural gas generation about 2.9%, coal about -6%Solar remains the fastest-growing power source, but policy headwinds lead the report to adopt a relatively cautious upward adjustment assumption.

Impact & implications

The investment implication is that demand growth and grid interconnection constraints are emerging at the same time, so utilities, independent power producers, the natural gas chain, companies related to nuclear restarts, renewable developers, and grid equipment and solutions providers all have opportunities to benefit. The report emphasizes that hyperscale cloud providers need new power within three years, while interconnection queues are longer, which will increase the value of behind-the-meter supply, fast-deployment generation, and solutions to grid bottlenecks.

Risks

• Data center projects are speculative, and interconnection applications may include many projects that will not be realized, causing incremental demand to fall below model assumptions.
• Changes in AI training and inference workloads, PUE improvement, and server utilization will significantly affect actual data center electricity consumption.
• Power demand may reach the grid before supply does, creating bottlenecks in interconnection, transmission, and infrastructure.
• Added solar capacity is affected by policy changes and macro headwinds, and actual supply growth may deviate from the model.
• As the marginal source filling the gap, natural gas is affected by fuel prices, pipeline capacity, environmental policy, and power market mechanisms.
• The adoption pace of EVs and heat pumps may be affected by consumer demand, policy subsidies, equipment costs, and the macro environment.

What to watch

• Whether added U.S. data center capacity, PUE, and utilization approach the report's base case.
• Whether EIA monthly power demand and generation data continue to validate the roughly 3% CAGR path.
• Whether manufacturing reshoring drives U.S. industrial GDO and industrial electricity demand upward in tandem.
• Whether EV sales mix, annual heat pump shipments, and incremental residential electricity demand materialize.
• Whether solar grid-connected capacity continues to exceed the initial estimates in the EIA forward planning database.
• Progress of nuclear restart projects, including Palisades, Three Mile Island, and Duane Arnold.
• Whether gas demand for power generation moves toward 1bcfd per year or the high-end scenario of 2bcfd per year.
• Whether interconnection queues, transmission investment, and grid equipment orders become constraints or investment opportunities.