Menu
Crude oil, gasoline, heating oil, diesel, propane, and other liquids including biofuels and natural gas liquids.
Exploration and reserves, storage, imports and exports, production, prices, sales.
Sales, revenue and prices, power plants, fuel use, stocks, generation, trade, demand & emissions.
Energy use in homes, commercial buildings, manufacturing, and transportation.
Reserves, production, prices, employment and productivity, distribution, stocks, imports and exports.
Includes hydropower, solar, wind, geothermal, biomass and ethanol.
Uranium fuel, nuclear reactors, generation, spent fuel.
Comprehensive data summaries, comparisons, analysis, and projections integrated across all energy sources.
Monthly and yearly energy forecasts, analysis of energy topics, financial analysis, congressional reports.
Financial market analysis and financial data for major energy companies.
Greenhouse gas data, voluntary reporting, electric power plant emissions.
Maps, tools, and resources related to energy disruptions and infrastructure.
State energy information, including overviews, rankings, data, and analyses.
Maps by energy source and topic, includes forecast maps.
International energy information, including overviews, rankings, data, and analyses.
Regional energy information including dashboards, maps, data, and analyses.
Tools to customize searches, view specific data sets, study detailed documentation, and access time-series data.
EIA's free and open data available as API, Excel add-in, bulk files, and widgets
Come test out some of the products still in development and let us know what you think!
EIA's open source code, available on GitHub.
Forms EIA uses to collect energy data including descriptions, links to survey instructions, and additional information.
Sign up for email subscriptions to receive messages about specific EIA products
Subscribe to feeds for updates on EIA products including Today in Energy and What's New.
Short, timely articles with graphics on energy, facts, issues, and trends.
Lesson plans, science fair experiments, field trips, teacher guide, and career corner.
EIA is continuing normal publication schedules and data collection until further notice.
After a 23-year hiatus, I am reintroducing the Administrator’s Foreword as part of the Annual Energy Outlook (AEO). The Foreword affords me an opportunity to provide context and outline future directions for one of our flagship products.
The U.S. energy system is rapidly changing. In recent years, technology innovation has accelerated the deployment of renewable energy, expanded markets for electric vehicles, and established record-high levels of petroleum and natural gas production. Heightened geopolitical risks have also influenced the energy system. And this year, recent federal legislation authorizes historic levels of investment in clean energy technology.
Ideally, we would model these dynamics to produce precise numerical forecasts that demonstrate how energy prices, technology deployment, and emissions will shift over time. Unfortunately, such precise forecasts are not possible. The 30-year decision landscape we model is too complex and uncertain. Thus, our objective must be to identify robust insights rather than precise numbers—think ranges and trends, not predictions and point estimates.
The AEO includes a series of projections—which we refer to as cases—each with different input assumptions that represent alternative views of how uncertainty may be resolved in the future. The Reference case represents our best guess under nominal conditions, which presumes no new policy or laws over the modeled time horizon. It’s best to think of the Reference case as the experimental control: a baseline against which we can judge the other cases. Although the Reference case serves as an important benchmark, judgments about energy futures should never be based on a single projection. The AEO side cases represent plausible variations in key input assumptions that tend to drive the largest changes in projected outputs from the Reference case. This year’s AEO narrative focuses on the full set of modeled cases in order to derive insights about our collective energy future.
Among the uncertainties we must confront, the timing, structure, and targets associated with yet-to-be-developed policy are the most uncertain. We only consider current laws and regulations across all modeled cases in this AEO. For some readers, this approach may be unsatisfying because policy rarely remains static for long periods. But this AEO should be considered part of an iterative policymaking process rather than apart from it; it gives decision-makers an opportunity to peer into a future without new policy. If the projected outcomes are undesirable from their viewpoint, they can effect change.
This year’s edition of the Annual Energy Outlook includes three enhancements that improve the characterization of future uncertainty and provide more technical details on the model results.
Although the AEO core side cases address key uncertainties, each case represents a one-factor change to the Reference case. But, real energy markets often surprise us in more ways than one, particularly over the decades-long timeframes modeled in the AEO. In this year’s edition, we include cases that combine assumptions from our macroeconomic and zero-carbon technology cost cases. These new cases reflect a combination of demand-side changes (macroeconomic growth affecting energy demand) and supply-side changes (renewables costs affecting generating capacity deployment) to expand the range of projections in the Annual Energy Outlook.
Running a set of cases is not enough: how we present and discuss them within the report affects the insights that our readers draw from the analysis. Although the Reference case is an important benchmark, each case represents a possible alternative. So, in each of the figures in this report, you will see shaded areas that represent the range of results obtained across the modeled cases. Uncertainty can be characterized in many other ways beyond the analysis of multiple cases. One way, presented in the discussion section, uses deviations between realized and projected values drawn from previous AEO editions to derive a cone of uncertainty for future energy-related CO2 emissions. Looking ahead, you should expect to see more innovation in how we treat uncertainty.
The narrative tends to focus on model-based results. We recognize that some readers want a deeper technical explanation around key issues. Although we describe our modeling approach elsewhere, how the model formulation and input assumptions influence the results is not always clear. To better explain key results, we included a series of technical notes in the narrative that focus on heat pump deployment, cost projections for renewables, electric vehicle deployment, and crude oil trade.
At EIA, we are also pursuing broader changes to our long-term modeling efforts. I would like to highlight three such efforts.
One of our priorities at EIA is to make our data and model-based analysis as transparent and accessible as possible. We are working to make the National Energy Modeling System (NEMS) publicly available in GitHub under a permissive, open-source license, and we hope to complete this effort later this year. Making our models open source allows users to examine, reuse, and redistribute our code under clear legal guidelines. Giving you this kind of access is important to the learning and discovery process associated with energy modeling.
Building on the combination cases in this year’s AEO, we are expanding our capability to model a wider range of future scenarios using NEMS. In particular, decision makers need objective and rigorous assessments of net zero emissions pathways to inform ongoing policy discussions. We are working to incorporate novel fuel and technology pathways into NEMS and to appropriately treat uncertainty around technologies with limited commercial deployment.
Although regularly updated, we have been using NEMS to produce the AEO since 1994—a span of nearly three decades that has born witness to significant changes in the real energy system, energy modeling methods, and software development practices. Moving forward, we need a flexible, next-generation modeling framework that can rapidly assess the cost, emissions, reliability, security, and community-level impacts associated with a number of contemporary energy issues. Some of these issues include pathways to a net-zero energy system, supply chain risks, rapid technology innovation, and shifting trade patterns. This modeling system will also be open source to promote transparency and encourage innovation within the modeling community. We’ve begun discussing this new framework, and I look forward to sharing our progress throughout my tenure as EIA Administrator.
I’d like to thank our long-term modelers for their willingness to take on new directions and their tremendous effort to produce this year’s AEO. I am very excited by the future work outlined above, and I feel privileged to help lead such a talented team of energy modelers.
Our Annual Energy Outlook 2023 (AEO2023) explores long-term energy trends in the United States. Since last year’s AEO, much has changed, most notably the passage of the Inflation Reduction Act (IRA), Public Law 117-169, which altered the policy landscape we use to develop our projections.
Energy-related CO2 emissions fall across all AEO2023 cases because of increased electrification, higher equipment efficiencies, and more zero-carbon electricity generation.
We project that U.S. energy-related CO2 emissions drop 25% to 38% below the 2005 level by 2030. For reference, the United States’ nationally determined contribution (NDC), submitted as part of the Paris Agreement, calls for a target of 50% to 52% of net greenhouse gas emissions below the 2005 level by 2030.1 We only consider energy-related CO2 emissions, which does not cover the full NDC scope. Total energy-related CO2 emissions in 2050 declined by 17% in this year’s Reference case compared with last year’s. Some of the primary factors that contributed to the change in our base case include the IRA, updates to technology costs and performance across the energy system, and changes in the macroeconomic outlook. All AEO2023 cases assume current laws and regulations, and compared with last year’s AEO, there is a significant shift toward lower future emissions. The IRA represents a complex piece of legislation, and we could not model all provisions given model structure and uncertainty over select implementation details. The appendix includes a detailed accounting of IRA provisions and how we addressed them. To further explore possible emissions reductions, we also derive a cone of uncertainty based on an empirical analysis of our past projections and find that the energy-related CO2 emissions reduction can be as high as 38% below 2005 levels in 2030.
Overall, our lower projected U.S. energy-related CO2 emissions is driven by increased electrification, equipment efficiency, and renewable technologies for electricity generation. However, emissions reductions are limited by longer-term growth in U.S. transportation and industrial activity. As a result, these projected emissions reductions are most sensitive to our assumptions regarding economic growth and the cost of zero-carbon generation technology.
Renewable generating capacity grows in all regions of the United States in all AEO2023 cases, supported by growth in installed battery capacity.
In AEO2023, we see stable growth in U.S. electric power demand through 2050 in all cases we considered because of increasing electrification and ongoing economic growth. The combination of declining capital costs and government subsidies, including IRA initiatives, drive rising renewable technologies for electricity generation, such as solar and wind. Once built and when the resource is available, wind and solar are the least cost resources to operate to meet electricity demand because they have zero fuel costs. Over time, the combined investment and operating cost advantage increases the share of zero-carbon electricity generation. As a result, in AEO2023, we see renewable generating capacity growing in all regions of the United States in all cases. Across all cases, compared with 2022, solar generating capacity grows by about 325% to 1019% by 2050, and wind generating capacity grows by about 138% to 235%. We see growth in installed battery capacity in all cases to support this growth in renewables. Across the span of AEO cases, relative to 2022, natural gas generating capacity ranges from an increase of between 20% to 87% through 2050.
Technological advancements and electrification drive projected decreases in demand-side energy intensity.
Not only is the U.S. electric power sector’s composition changing, but we see increased electrification in the end-use sectors. We project more heat pumps and electric vehicles, as well as electric arc furnaces increasingly deployed in the iron and steel industry. In the residential and commercial sectors, higher equipment efficiencies and stricter building codes extend ongoing declines in energy intensity. Despite the growth in adopting heat pumps, natural gas-fired heating equipment, including furnaces and boilers, continue to account for the largest share of energy consumption for space heating in U.S. residential and commercial buildings across all cases through 2050. In the transportation sector, light-duty vehicle energy demand declines through 2045 as more electric vehicles are deployed and stricter Corporate Average Fuel Economy (CAFE) standards largely offset the continued growth in travel demand. The energy demand then increases as rising travel overcomes increasing efficiency. Across all cases, light-duty vehicle energy demand decreases by 3% to 28% in 2050 relative to 2022.
High international demand leads to continued growth in U.S. production, and combined with relatively little growth in domestic consumption, allows the United States to remain a net exporter of petroleum products and natural gas through 2050 in all AEO2023 cases.
Despite no significant change in domestic petroleum and other liquids consumption through 2040 across most AEO2023 cases, we expect U.S. production to remain historically high as exports of finished products grow in response to growing international demand. Despite the shift toward renewable sources and batteries in electricity generation, domestic natural gas consumption remains relatively stable—ending recent growth in most cases. Natural gas production, however, in some cases continues to grow in response to international demand for liquefied natural gas, supported by associated natural gas produced along with crude oil. Given the combination of relatively little growth in domestic consumption and continued growth in production, we project that the United States will remain a net exporter of petroleum products and natural gas through 2050 in all AEO2023 cases.
The Annual Energy Outlook 2023 (AEO2023) explores long-term energy trends in the United States. Since we released the last AEO in early 2022, passage of the Inflation Reduction Act (IRA), Public Law 117-169, altered the policy landscape we use to develop our projections. The Appendix in this report explains our assumptions around IRA implementation and how we implemented the IRA in our AEO2023 cases. We are also releasing a separate Issues in Focus paper that explores how these assumptions affect our model-based projections. We have seen significant national and international short-term market volatility associated with economic growth as the world reemerges from the COVID-19 pandemic and political instability associated with Russia’s full-scale invasion of Ukraine. We continuously monitor such developments and consider how they may affect our long-term projections.
AEO2023 includes a Reference case and 12 side cases that explore key areas of uncertainty about how energy markets will develop. We retooled our graphs to emphasize the range of results, denoted by shaded areas, across the full suite of modeled cases. We derive our key analytical insights by assessing the results across cases and examining how overall trends may vary under different assumptions. This year, we’ve added a discussion section focused on sources of uncertainty in the AEO2023 projections. We also now derive a cone of uncertainty associated with future energy-related CO2 emissions using deviations between past projections and realized values.
Under the Paris Agreement, the United States set a goal to reduce economy-wide greenhouse gas emissions by 50% to 52% of 2005 levels by 2030. We only consider energy-related CO2 emissions, which does not cover the full NDC scope. We project lower U.S. energy-related CO2 emissions in 2030 relative to 2005 in the AEO2023 Reference case and all side cases (Figure 1). CO2 emissions are most sensitive to economic growth and assumptions related to the cost of zero-carbon generation technology. Combinations of these two sets of assumptions form the upper and lower bounds of projected CO2 emissions. Emissions decrease by 25% in 2030 relative to 2005 under the combined high economic growth and high zero-carbon technology cost assumptions and by as much as 38% under low economic growth and low zero-carbon technology cost assumptions. Both of these cases hinge on specific assumptions regarding the relationship between economic growth and zero-carbon technology development.2 In the High Economic Growth case, emissions fall initially and then begin to increase again in 2040 as industrial activity and travel (measured in vehicle miles traveled) increase, surpassing emissions reductions from the electric power sector.
The largest variations in projected U.S. CO2 emissions across cases occur in the electric power, transportation, and industrial sectors. Although economic growth assumptions affect consumption and, in turn, projected CO2 emissions in all sectors, different case-specific assumptions affect sectors differently. For example, emissions from the electric power sector are particularly responsive to assumptions about the cost of zero-carbon technologies, and transportation sector emissions are sensitive to assumptions about fossil fuel supply and cost, particularly oil and petroleum products.
In this section, we discuss renewables displacing fossil fuels in the electric power sector.
Economic growth paired with increasing electrification in end-use sectors results in stable growth in U.S. electric power demand through 2050 in all cases. Declining capital costs for solar panels, wind turbines, and battery storage, as well as government subsidies such as those included in the IRA, result in renewables becoming increasingly cost effective compared with the alternatives when building new power capacity.
Power demand is increasingly met by renewables throughout the projection period.
Renewables are increasingly meeting power demand throughout the projection period (Figure 2). Natural gas, coal, and nuclear generation shares decline. Renewable power outcompetes nuclear power, even in the Low Zero-Carbon Technology Cost (ZTC) case, which evaluates the impact of more aggressive cost declines for nuclear and renewables than the Reference case. Most natural gas-fired generation comes from combined-cycled power plants as opposed to simple-cycle combustion turbines. Uncertainty in natural gas prices across cases leads to various projections for combined-cycle units in the short term, but in the long term, natural gas demand from the electric power sector stabilizes across all cases.
In order to meet increasing demand for electric power throughout the projection, total installed power capacity close to doubles across most cases, even in the Low Economic Growth case (Figure 3). Cases with a higher share of renewables in the generation mix have higher total grid capacity due to the inherently lower capacity factors of solar and wind compared with coal, nuclear, and combined-cycle plants.
We project that renewable power capacity will increase in all regions of the United States in all AEO2023 cases, although regional resource availability results in varying renewable resource mixes across regions (Figure 4). Across all cases, between 40%-60% of the renewable power capacity in the Mid-Continent region in 2050 comes from wind, and the Southeast and the region managed by the California Independent System Operator (CAISO) have large shares of solar and a small amount of wind power capacity in all cases.
Once built and when the resource is available, wind and solar generation outcompete other technologies for system dispatch because they have zero fuel costs. Across all AEO2023 cases, some renewable generation is left unused and curtailed, typically midday when solar generation can exceed demand in some regions and seasons. Battery capacity is built in all cases to store and dispatch some of this otherwise unused generation in later hours, decreasing reliance on fossil fuel capacity, such as natural gas-fired peaking units or load-following combined-cycle units (Figure 5). Battery storage is also used to replace natural gas-fired capacity to provide reserve capacity. In the Reference case in 2050, 160 gigawatts (GW) of standalone battery storage capacity will be deployed, and deployment varies between 40 GW and 260 GW in the other cases.
Technical Note 1: Renewable costs and deployment Continued decline of renewable technology costs relative to natural gas-fired generating technologies contributes to the change in generation mix in our projection period. In addition to recent policies that also favor renewables in the generation mix, we use learning factors to represent learning by doing, which reduces capital costs. Learning factors are calculated independently for each of the major design components of the technology, and they increase based on how much new capacity is deployed. For details on renewable costs in the ZTC cases, please see the Appendix. New, untested components decrease at a more rapid initial rate than mature components or conventional designs. More details can be found in the NEMS Electricity Market Module documentation.
Technical Note 1: Renewable costs and deployment
Continued decline of renewable technology costs relative to natural gas-fired generating technologies contributes to the change in generation mix in our projection period. In addition to recent policies that also favor renewables in the generation mix, we use learning factors to represent learning by doing, which reduces capital costs. Learning factors are calculated independently for each of the major design components of the technology, and they increase based on how much new capacity is deployed. For details on renewable costs in the ZTC cases, please see the Appendix. New, untested components decrease at a more rapid initial rate than mature components or conventional designs. More details can be found in the NEMS Electricity Market Module documentation.
As a result of renewables growth, we project that U.S. coal-fired generation capacity will decline sharply by 2030 to about 50% of current levels (about 200 GW) with a more gradual decline thereafter. We project between 23 GW and 103 GW of coal-fired capacity operating in 2050 (Figure 6). The IRA provides additional incentives to wind and solar power generation, which accelerates the near-term decline of electric power sector coal-fired generating capacity and hastens the timeline for retirement in the U.S. coal fleet.
Coal consumption in the U.S. electric power sector in the Reference case drops to 189 million short tons (MMst) in 2030 and to 131 MMst in 2050 from 458 MMst in 2022. Coal disposition, which includes exports and consumption by the electric power sector and other end-use sectors, declines to a low of 170 MMst in 2050 in the Low ZTC case. In a high natural gas price environment, such as in the Low Oil and Gas Supply case, coal disposition could remain as high as 350 MMst in 2050. In all cases, annual coal exports average about 110 MMst in 2050, and end-use coal demand averages about 36 MMst. The ratio of coal exports to domestic coal consumption generally increases through the projection period in all cases. The majority of domestically produced coal is exported by 2050 in the Low ZTC case, 45% is exported in the High Oil and Gas supply case, which are the two cases that have the least coal demand from the electric power sector. Even in cases with more aggressive retirement of coal-fired power plants, such as the Low ZTC case, some of the relatively newer and more efficient coal power plants remain online across the United States because they can provide cheap dispatchable power to the grid.
The electric power sector is also decreasing its reliance on natural gas in favor of renewables, which we will discuss in Section 3.
Moderate growth in U.S. energy consumption is the result of economic growth, population growth, and increased travel offsetting continued energy efficiency improvements. Demand-side energy intensity—the measure of energy consumed per household or per square foot of commercial floorspace—decreases as a result of changes in technology, policy, consumer behavior, demographics, and fuel mix. In this section, we quantify the decreases in CO2 emissions intensity and demand-side energy intensity, and we discuss equipment changes in the buildings sector, electrification in iron and steel production, and technological advancements and government standards in the transportation sector.
Total energy consumption, including electricity use and electricity-related losses, increases by as much as 15% from 2022 to 2050 across the AEO2023 Reference case and side cases.
U.S. energy consumption increases in many end-use sectors across all AEO2023 cases. Total energy consumption, including electricity use and electricity-related losses, increases by as much as 15% from 2022 to 2050 across the AEO2023 Reference case and side cases (Figure 7). The largest increases, in percentage terms, are in the industrial sector where energy consumption increases as much as 32% and in the transportation sector where energy consumption increases as much as 8%. Energy consumption in the residential and commercial sectors are the least sensitive to changes in assumptions across cases.
In addition to macroeconomic growth assumptions, which affect energy consumption in all sectors, many sectors are also highly responsive to zero-carbon technology cost assumptions. Increasing energy consumption, improving end-use and electric power sector technology and efficiency, and declining costs for zero-carbon generation technologies, which, in turn, leads to cheaper electricity, all lead to increased electrification in the end-use sectors. The share of electricity in the residential and transportation sectors increase the most as demand for space cooling increases and electric vehicles gain a larger market share. The residential sector purchased 5.1 quadrillion British thermal units (quads) of electricity in 2022, and residential consumption of purchased electricity increases between about 14% and 22% from 2022 to 2050 across all cases, reaching between 5.9 and 6.3 quads. Electricity purchased for transportation reaches between about 0.6 quads and 1.3 quads in 2050, from 0.1 quads of purchased electricity in 2022, an increase of between 892% and 2,038% across all cases. Electricity purchased in the industrial sector is most influenced by economic growth assumptions, increasing by about 3%, from 3.5 quads in 2022 to about 3.6 quads in 2050 in low economic growth cases and by about 36% to 38%, to about 4.7 quads, in high economic growth cases.
Despite modest growth in total energy consumption in the residential and commercial sectors, due to a growing number of households and expanding commercial floorspace, average energy intensity declines through 2050 across all cases (Figure 7 and Figure 8). Building envelope efficiency improves as states and localities adopt newer building energy codes and some existing households and commercial spaces receive additional insulation, air sealing, and other weatherization upgrades.
An established trend toward warmer winters and population shifts toward warmer and drier areas of the United States reduce energy consumption for space heating in all cases. At the same time, the established trend toward warmer summers leads to increasing electricity consumption for space cooling.
Technical Note 2: Modeling growth in residential heat pump installations Our projections of residential heat pumps, like all major end-use equipment, begin with the census division-level stocks from our 2015 Residential Energy Consumption Survey (RECS). We use data from the most recent U.S. Census Bureau’s Survey of Construction to further align recent space heating equipment shares in newly built housing units. Equipment purchase decisions account for federal and non-federal subsidies that further reduce installed costs of high-efficiency equipment such as air- and ground-source heat pumps or high-efficiency natural gas furnaces. These subsidies include national tax credits extended by the Inflation Reduction Act (IRA) and utility rebates to end users at the census-division level. Technology performance and energy prices are considered as well. Some provisions of the IRA—such as those targeting low-income households—are not explicitly included in our modeling. Refer to the Appendix for additional details.
Technical Note 2: Modeling growth in residential heat pump installations
Our projections of residential heat pumps, like all major end-use equipment, begin with the census division-level stocks from our 2015 Residential Energy Consumption Survey (RECS). We use data from the most recent U.S. Census Bureau’s Survey of Construction to further align recent space heating equipment shares in newly built housing units. Equipment purchase decisions account for federal and non-federal subsidies that further reduce installed costs of high-efficiency equipment such as air- and ground-source heat pumps or high-efficiency natural gas furnaces. These subsidies include national tax credits extended by the Inflation Reduction Act (IRA) and utility rebates to end users at the census-division level. Technology performance and energy prices are considered as well. Some provisions of the IRA—such as those targeting low-income households—are not explicitly included in our modeling. Refer to the Appendix for additional details.
Natural gas-fired heating equipment, including furnaces and boilers, continue to account for the largest share of energy consumption for space heating in U.S. residential and commercial buildings across all cases throughout the projection period. Over time, older heating equipment is replaced by newer, more efficient equipment that meets updated federal minimum energy efficiency standards.
Federal and non-federal subsidies both encourage homes and businesses to adopt high-efficiency natural gas and electric equipment, including heat pumps. Electric heat pumps, including ground-source heat pumps, gain market share over the projection period, increasing from 11% of households in 2022 to between 14% and 15% of households in 2050 across all cases; however, their growth is limited by:
Steel production is an energy-intensive industry, but the choice of production technology significantly affects its energy and emission intensity. U.S. manufacturers continue to transition away from the combustion-powered, integrated steel mill process to steel produced by electric-arc furnaces, which have a lower energy intensity and make up about 68% of U.S. steel produced in 2022. Over the projection period, the share of U.S. steel produced by the electric-arc furnace process increases by 4% to 7% across the range of cases (Figure 9). The energy intensity of U.S. steel production continues to fall across all cases, declining between 12% and 21% across all cases (Figure 9). In 2018, steel production accounted for 1.3% of U.S. energy demand, and we project total energy demand for iron and steel production to fall relative to total U.S. energy demand, after peaking in 2027.
In the transportation sector, light-duty vehicle (LDV) fuel economy increases due to rising Corporate Average Fuel Economy (CAFE) standards and electric vehicle (EV) sales. In addition to required fuel economy increases, consumer purchase decisions are also influenced by fuel prices. Consumer interest in EVs, which are significantly more efficient than internal combustion engine vehicles, and the impact EV adoption has on average light-duty vehicle fuel economy are sensitive to the price of gasoline (Figure 10).
The updated CAFE standard, which applies to model years 2024 through 2026, significantly increases average new vehicle fuel economy requirements. By 2026, the updated CAFE standard is 28% higher than the Safer Affordable Fuel-Efficient (SAFE) standard for new vehicles it replaced, resulting in an increase from 37 miles per gallon (mpg) to 47 mpg by 2026. In addition to improved conventional vehicle fuel efficiency, EV sales increase through 2050, increasing EVs on the road. The CAFE standard, which offers credits to EV manufacturers, and decreasing battery prices help drive this increase in EV sales.
We project LDV energy consumption to fall through the early 2040s as a result of fuel economy improvements but then to rise due to increasing vehicle miles traveled for the remainder of the projection period.
U.S. passenger vehicle-miles traveled increases steadily with population and income throughout the projection period, growing between 12% and 33% across all cases. In the Reference case, 23% more vehicle miles are traveled in 2050 than in 2022 (Figure 11). We project LDV energy consumption to fall through the early 2040s as a result of fuel economy improvements but then to rise due to increasing vehicle miles traveled for the remainder of the projection period (Figure 11).
Technical Note 3: Electric vehicle (EV) deployments Projected declines in EV vehicle component costs, along with federal and state policies that provide incentives for EV purchases or require minimum sales, drive EV sales growth in our model projection. We derive cost declines for EV powertrain components and batteries by using learning rates based on cumulative production, resulting in increased projected driving range and a continual decline in EV prices over the projection period. We derive EV sales shares at the census-division level using a consumer choice model based on preference data and calibrated to align with historical sales data. In addition to other vehicle attributes, our consumer choice model captures the impact of vehicle price, cost to drive, access to refueling, and the effect that availability of vehicle propulsion options has on consumer purchase decisions. All of these factors contribute to the attractiveness of EVs to consumers and increases EV deployment relative to internal combustion engine vehicles. We assume Corporate Average Fuel Economy standards result in technological improvements and increased EV adoption because of declining cost and favorable fuel economy credits. The clean vehicle credit in the 2022 Inflation Reduction Act, which varies from $3,750 to $7,500 per vehicle, drives additional EV sales. Our Transportation Demand Model also ensures legally enforceable state minimum EV sales requirements are met in each census division and EV prices adjust to account for tax credits at purchase. In addition, other factors, such as a rising number of charging stations, affect our EV sales projections. We base our model projections on these data, assumptions, and current enforceable laws and regulations and do not assume that state and federal stated goals for EV sales are met. These details are available in our model documentation.
Technical Note 3: Electric vehicle (EV) deployments
Projected declines in EV vehicle component costs, along with federal and state policies that provide incentives for EV purchases or require minimum sales, drive EV sales growth in our model projection. We derive cost declines for EV powertrain components and batteries by using learning rates based on cumulative production, resulting in increased projected driving range and a continual decline in EV prices over the projection period. We derive EV sales shares at the census-division level using a consumer choice model based on preference data and calibrated to align with historical sales data. In addition to other vehicle attributes, our consumer choice model captures the impact of vehicle price, cost to drive, access to refueling, and the effect that availability of vehicle propulsion options has on consumer purchase decisions. All of these factors contribute to the attractiveness of EVs to consumers and increases EV deployment relative to internal combustion engine vehicles. We assume Corporate Average Fuel Economy standards result in technological improvements and increased EV adoption because of declining cost and favorable fuel economy credits. The clean vehicle credit in the 2022 Inflation Reduction Act, which varies from $3,750 to $7,500 per vehicle, drives additional EV sales. Our Transportation Demand Model also ensures legally enforceable state minimum EV sales requirements are met in each census division and EV prices adjust to account for tax credits at purchase. In addition, other factors, such as a rising number of charging stations, affect our EV sales projections. We base our model projections on these data, assumptions, and current enforceable laws and regulations and do not assume that state and federal stated goals for EV sales are met. These details are available in our model documentation.
Although U.S. consumption of petroleum products remains relatively flat, international demand supports U.S. exports of petroleum and other liquids (Figure 12). The dynamics of international trade affect domestic production of natural gas and of petroleum and other liquids.
In all cases, we project that the United States will remain a net exporter of petroleum products through 2050.
Although domestic consumption of petroleum and other liquids does not increase through 2040 across most cases, U.S. petroleum and other liquids production remains high because of increased exports of finished products in response to growing international demand (Figure 13). In all cases, we project that the United States will remain a net exporter of petroleum products through 2050. In the High Oil Price case, increased production leads to the most exports among all cases over the projection period. The Low Oil Price case shows the opposite: decreased production along with the lowest export volumes.
Crude oil imports remain relatively flat in the Reference case but vary widely in the side cases (Figure 13); the Low Oil and Gas Supply case leads to the greatest level of imports throughout the forecast period while the High Oil and Gas Supply case leads to the lowest imports (Figure 13). This wide range in imports is mainly due to the tradeoff between domestic production and imports. In the Low Oil and Gas Supply case, crude oil imports increase significantly, partially to account for falling domestic crude oil production. The opposite occurs in the High Oil and Gas Supply case, in which increased domestic production balances lower crude oil imports.
In the High Oil Price case, U.S. crude oil imports initially decline but begin to increase starting in 2030 because of changing trends in domestic crude oil production. In the early years, domestic crude oil production increases rapidly due to high prices. However, crude oil production begins to fall after 2030 because wells are drilled increasingly close to one another, resulting in well productivity declines. As wells are drilled closer together, they produce less crude oil and eventually become unprofitable, at which point new drilling stops. Crude oil imports decline early in the High Oil Price case as crude oil production increases; imports increase after 2030 as well productivity and crude oil production declines.
Because international demand for finished petroleum products keeps exports high, U.S. refinery runs remain strong as the U.S. refinery sector remains competitive in the global market through 2050. Refinery capacity remains relatively constant through 2050, and refinery capacity utilization remains high, at around 90% or higher, under favorable economic conditions.
Technical Note 4: Crude oil trade dynamics Crude oil is a global commodity, and the United States participates in the global market as both an importer and exporter of crude oil and its associated products. Because of logistical, regulatory, and quality considerations, both exporting and importing petroleum often makes economic sense. For example, a refiner in the Gulf Coast may find it more profitable to export motor gasoline to Mexico rather than shipping it to the East Coast because cheaper gasoline imports from Europe may be available to the East Coast. The chemical makeup of the imported or exported product also affects crude oil trade. The type of crude oil—light or heavy, low-sulfur (sweet) or high-sulfur (sour)—helps determine the processes that refine it into a petroleum product such as distillate fuel oil or propane. In short, the United States imports different types of crude oil to optimize production across its various refineries. The World Oil Price path is an exogenous assumption to the model and affects domestic crude oil production and international trade. For each year of the projection period, NEMS computes the Brent crude oil price, provides a supply curve of world crude oil-like liquids, and provides supply curves for each foreign crude oil type considered. NEMS also provides, for each year of the projection period, exogenous supply and demand curves for U.S. import and export of petroleum products.
Technical Note 4: Crude oil trade dynamics
Crude oil is a global commodity, and the United States participates in the global market as both an importer and exporter of crude oil and its associated products. Because of logistical, regulatory, and quality considerations, both exporting and importing petroleum often makes economic sense. For example, a refiner in the Gulf Coast may find it more profitable to export motor gasoline to Mexico rather than shipping it to the East Coast because cheaper gasoline imports from Europe may be available to the East Coast.
The chemical makeup of the imported or exported product also affects crude oil trade. The type of crude oil—light or heavy, low-sulfur (sweet) or high-sulfur (sour)—helps determine the processes that refine it into a petroleum product such as distillate fuel oil or propane. In short, the United States imports different types of crude oil to optimize production across its various refineries.
The World Oil Price path is an exogenous assumption to the model and affects domestic crude oil production and international trade. For each year of the projection period, NEMS computes the Brent crude oil price, provides a supply curve of world crude oil-like liquids, and provides supply curves for each foreign crude oil type considered. NEMS also provides, for each year of the projection period, exogenous supply and demand curves for U.S. import and export of petroleum products.
In response to Russia’s full-scale invasion of Ukraine in early 2022, the United States banned petroleum imports from Russia. AEO2023 projections reflect this policy change. However, we assume that equivalent imports from other countries substitute for U.S. crude oil imports (especially unfinished oil imports) from Russia, minimizing effects on domestic markets.
In the United States, electrification is displacing combustion fuels in the demand sectors. As electricity generation shifts to using more renewable and battery sources, domestic natural gas consumption for electricity generation is likely to decrease by 2050 relative to 2022, which contrasts with relatively stable growth over the past decade (Figure 14).
More natural gas is consumed in the industrial or electric power sectors than in any other sectors of the U.S. economy. Projected consumption in both sectors is very sensitive to changes in our side case assumptions, particularly in the Oil and Gas Supply cases (Figure 15). These cases, which result in the most and the least natural gas consumption in the industrial sector, vary widely due to differences in resource extraction assumptions. By 2050, natural gas consumption in the industrial sector diverges from the Reference case by 14% in the Low Oil and Gas Supply case and 18% in the High Oil and Gas Supply case.
In the electric power sector, our projections for natural gas consumption generally fall but range widely, with consumption in 2050 diverging from the Reference case by over 50% in the bounding cases.
In the electric power sector, our projections for natural gas consumption generally fall but range widely, with consumption in 2050 diverging from the Reference case by over 50% in the bounding cases (Figure 15). Natural gas consumption remains below the peak in 2022, at nearly 12 trillion cubic feet, through 2050 across all side cases except the High Economic Growth and High ZTC case. In the Low ZTC case, lower costs for renewables makes natural gas less competitive, resulting in a larger decrease in natural gas consumption compared with the Reference case. In the High Economic Growth and High ZTC case, increased economic activity drives increased end-use demand, which results in more natural gas consumption. Higher costs for renewables make natural gas a more competitive option in that case, further increasing natural gas consumption in the electric power sector.
In the Reference case, U.S. natural gas production increases by 15% from 2022 to 2050, and consumption decreases by 6% from its peak in 2022 (Figure 16). Across all cases, domestic production outpaces domestic consumption; production increases across all side cases except in the Low Oil and Gas Supply case and the Low Oil Price case.
In some cases, exports to satisfy growing international demand for natural gas encourage growth in domestic natural gas production. A significant portion of production growth is due to liquefied natural gas (LNG) export demand, which drives the overall increase in natural gas exports (Figure 17).
Historically, most LNG was traded under long-term, oil price-linked contracts because a global LNG price benchmark did not exist and because oil could substitute for natural gas in industry and for power generation, which was especially common in Asia. These factors supported highly correlated international natural gas and oil prices. With growth in more market based-LNG, the strength of the relationship between international natural gas prices and oil prices has eroded. However, we expect that future oil prices will still affect additional LNG export capacity and overall export levels. When the Brent price is high relative to the U.S. Henry Hub price, like in the High Oil Price case, building more LNG export capacity and exporting LNG are more economical than when the Brent price is lower relative to Henry Hub. In the Low Oil Price case, the Brent price is lower, and the Henry Hub price is higher, which curtails LNG exports to below current volumes in the near term and causes LNG capacity to be underutilized near the end of the projection period.
International demand for LNG exports results in rising natural gas production, favoring areas that have better access to terminals. In AEO2023, dry natural gas production grows in the Southwest, which has easy pipeline transport to the Gulf Coast, where LNG is exported. Production in the Gulf Coast also generally increases across the projection period, due to its proximity to LNG export terminals, in all cases except the Low Oil and Gas Supply case (Figure 18).
In AEO2023, shale gas and associated dissolved natural gas from oil formations are the primary sources of long-term growth of domestic natural gas production through 2050. Increased production wells in the Permian Basin (Southwest region) is the primary driver behind associated dissolved natural gas growth. Increases in shale gas production mainly comes from the Texas-Louisiana Salt Basin (Gulf Coast Region) and the Appalachian Basin (East Region).
In the High Oil Price case and High Oil and Gas Supply case, oil production growth leads to increased associated dissolved natural gas and shale production (Figure 19). The opposite occurs in the Low Oil Price case and Low Oil and Gas Supply case.
Energy market projections are inherently uncertain because many of the events that will shape future energy markets—including developments in policy, technology, demographics, and resources— are not known. To illustrate the role of uncertainty, AEO2023 includes a baseline Reference case and several side cases that systematically vary important underlying assumptions. Many sources of uncertainty exist beyond the ones we test explicitly, including new policy, unforeseen geopolitical events, and rapid technology innovation, particularly around technologies that are in the earliest stages of development.
Our key assumptions in the Reference case provide a baseline for exploring long-term trends, based on current laws and regulations as of November 2022. These assumptions include provisions of the IRA; however, we were unable to model all provisions, as indicated in the appendix. Any future legislation would further affect technology trajectories and emissions pathways. We publish the current laws and regulations considered in the AEO2023 on the AEO website.
We account for current events that affect the energy markets with the information available at the time we prepare this publication. However, we cannot foresee future events such as wars, supply disruptions, pandemics, or other such issues that could having lasting impacts on the U.S. energy system.
The technologies considered in the AEO2023 include only well-documented trends in energy innovation. Additional breakthroughs not considered here might occur. Examples of the kind of breakthrough we don’t consider include early-developmental-stage technologies such as hydrogen, enhanced geothermal, and fusion, as well as other technologies currently unknown or not well characterized.
The sources of uncertainty described above, among others, result in observed values that are different than our projected values. To explore some key uncertainties, we include a number of side cases that incorporate plausible alternatives to assumptions in the Reference case. We can also develop a sense of future uncertainty by calculating the differences (or statistical errors3) between realized values and our Reference case projections from previous AEO editions. Given constantly evolving modeling techniques and a dynamic energy landscape replete with non-linear relationships, statistical errors from past projections will not necessarily provide an accurate basis for estimating uncertainty in current projections. Nevertheless, using statistical errors from past projections implicitly captures real world factors that were difficult to anticipate in past AEO editions.
In the 2022 edition of our Annual Energy Outlook Retrospective Review (AEO Retrospective), we quantified the statistical errors associated with past Reference cases over several key output metrics and projection timeframes, ranging from 1 to 15 years. Building on the work from the AEO Retrospective and following the NP2 method (a methodology detailed in a 2017 paper by researchers Lynn H. Kaack, Jay Apt, M. Granger Morgan, and Patrick McSharry), we can use statistical errors from past projections to develop a cone of uncertainty—similar to those used by the National Oceanic and Atmospheric Administration (NOAA) to produce hurricane path cones—which we can apply to the AEO2023 projections. Employing statistical errors from past projections along with side case projections can help us better assess the possible range of uncertainty in the AEO results.
For example, we can use the statistical errors from past AEO Reference case projections to project future uncertainty in U.S. energy-related CO2 emissions produced by the AEO2023 Reference case until 2030 (Figure 20). Figure 20 includes two uncertainty cones for total energy-related CO2 emissions, representing statistical error estimations where the narrower and wider cones capture 50% and 95% of the historical projection errors around the Reference case, respectively. Our Reference case projection has a wide range of future uncertainty in total energy-related CO2 emissions, which in some instances goes beyond the bounds explored in our most extreme side cases.
Because the AEO Reference case includes only laws and regulations current at the time it is developed, past Reference case projections of total energy-related CO2 emissions tend to be higher than actual because they don't include subsequent public policies that further reduce emissions. The uncertainty cones capture these over-projected statistical errors in total energy-related CO2 emissions and show the possibility for lower emissions compared with our most extreme side case with the lowest total energy-related CO2 emissions. We see the possibility of total energy-related CO2 emissions increasing in the short term before falling in the long term, both in our High Economic Growth and High Zero-Carbon Technology Cost case and the upper 95th percentile of statistical errors from past projections.
We will continue to explore our use of this technique, as well as other ways to quantify uncertainty, in future analyses.
In the AEO2023 Reference case, we assess how U.S. and world energy markets would operate through 2050 under current laws and regulations as of November 2022 under evolutionary technological growth assumptions. Our key assumptions in the Reference case provide a baseline, or experimental control, for exploring long-term trends. An overview of the laws and regulations included in AEO2023 is available on the AEO website. This Appendix addresses the Inflation Reduction Act and how we incorporated it into the Reference case and side cases.
Global market balances, primarily international supply and demand factors, will drive future crude oil prices. To account for these factors, oil prices are an exogenous assumption in our analysis. In the AEO2023 High Oil Price case, the price of Brent crude oil, in 2022 dollars, reaches $190 per barrel (b) by 2050, compared with $101/b in the Reference case and $51/b in the Low Oil Price case.
Compared with the Reference case, the High Oil and Gas Supply case assumes that the estimated ultimate recovery per well for tight oil, tight gas, or shale gas in the United States is 50% higher. Similarly, this case assumes that undiscovered resources in Alaska and the offshore Lower 48 states are 50% greater than assumed in the Reference case. Technological improvement rates that reduce costs and increase productivity of oil and natural gas production in the United States are also 50% higher than assumed in the Reference case. Conversely, the Low Oil and Gas Supply case assumes that the estimated ultimate recovery per well for tight oil, tight gas, or shale gas in the United States; the undiscovered resources in Alaska and the offshore Lower 48 states; and rates of technological improvement are all 50% lower than assumed in the Reference case.
The High Zero-Carbon Technology Cost case and the Low Zero-Carbon Technology Cost case examine the sensitivities around capital costs for electricity-generating technologies that produce zero emissions, which include renewables, nuclear (a zero-carbon technology included in these cases for the first time in this AEO), and diurnal storage technologies. We assume capital costs decline over time from learning by doing as commercialization expands and construction and manufacturing experience accelerates. The High Zero-Carbon Technology Cost case assumes no cost reductions from learning by doing. The Low Zero-Carbon Technology Cost case assumes faster, exogenously determined technology cost declines through 2050, resulting in about a 40% cost reduction by 2050 compared with the Reference case for each zero-carbon technology. In addition, we assume fixed operating and maintenance costs will decline along with the capital cost from technology improvement. These cases replace the Low and High Renewable Cost cases from prior AEOs and now reflect the zero-carbon technology suite, as represented in our models, which is targeted by incentives in the Inflation Reduction Act.
The High Economic Growth case and Low Economic Growth case address the effects of economic assumptions on energy consumption modeled in the AEO2023. From 2022 to 2050, the High Economic Growth case assumes the compound annual growth rate for U.S. GDP is 2.3%, and the Low Economic Growth case assumes a 1.4% rate. By contrast, the Reference case assumes the U.S. GDP annual growth rate is 1.9% over the projection period.
In addition to our eight standard core cases, we have added four combination cases for AEO2023. These cases simultaneously vary economic growth and zero-carbon technology cost assumptions. The four combinations are:
The Inflation Reduction Act (IRA), or Public Law 117-169, that took effect on August 16, 2022, includes energy- and climate-related provisions.4 We reviewed the law for directives that would influence energy consumption, production, and trade in the U.S. economy as modeled in NEMS. We incorporated provisions of this law into the Reference case and side cases, unless otherwise noted. Case assumptions were frozen in mid-November 2022, and AEO2023 does not include regulatory guidance or provisions issued after that time.
This document summarizes the energy- and climate-related provisions from the IRA as incorporated into the AEO2023 Reference and side cases. This document does not cover details about provisions included in the Low and High Uptake cases, which vary the number of bonus tax credits and incentives applied to eligible technologies. Those case definitions and results are discussed in the Issues in Focus: Inflation Reduction Act Cases in the AEO2023. Further details about how we modeled the IRA provisions are also in the NEMS Assumptions documents. The No IRA case assumes the same economic outlook as the Reference case but excludes the IRA provisions.
All cases use the macroeconomic outlook from S&P Global IHS Markit as of September 2022.
Although all provisions of the IRA are current law, some are not explicitly included in the NEMS version used for AEO2023. We did not include these provisions for one of three reasons:
Some provisions are not listed in the following table because they are not relevant to this analysis or had minor impacts to the energy system. In the Excluded column of Table 1, the numbers in bold correspond to one or more of the reasons listed above. The use of -- indicates that a certain column is not applicable.
