From Initial Assumptions to Integrated Energy-System Responses
The U.S. Energy Information Administration’s Annual Energy Outlook 2026 includes a Counterfactual Baseline and ten side cases that test selected changes in economic, technological, resource, demand, and policy conditions. EIA has described the principal assumptions and many of the resulting sectoral outcomes. The broader analytical value of the suite lies in treating the cases as structured experiments: identifying what changes first, tracing how the response propagates through the integrated energy system, and determining which results warrant further investigation. This article maps those experiments, examines the mechanisms connecting their effects across sectors and markets, and identifies both the research opportunities and the unresolved questions that emerge from the published cases.
Side cases are starting points for analysis
EIA produces the AEO using the National Energy Modeling System, or NEMS, an integrated model of U.S. energy supply, demand, conversion, prices, and related economic activity. A side case changes one assumption or, more often, a structured package of assumptions. NEMS then determines how the represented energy system responds. The resulting projection is conditional: if the specified assumptions were in place and the modeled relationships operated as represented, this is the pathway the model would produce. It is not a prediction or a probability-weighted forecast. Nor is the modeled pathway the end of the analysis. It is the starting point for understanding how the effects of an initial change propagate through the energy system.
A useful way to interpret a side case is to trace three stages within the model. The first is the initial modeling change. The second is the direct response in the module, sector, or market where the changed assumptions first enter NEMS. The third is the propagated response across connected sectors and markets. The analyst must then take a fourth step: identify the next questions raised by the results. This distinction matters because the most informative outcome may occur well beyond the part of the model where the original assumptions were changed.
Figure 1. How to interpret an AEO side case
The Counterfactual Baseline provides the common experimental control for these comparisons. EIA formerly called this the Reference case, but the new terminology reinforces that the baseline is neither EIA’s preferred future nor necessarily the most likely one. The side cases likewise should not be interpreted collectively as a statistical confidence interval. They are differently structured experiments designed to illuminate selected uncertainties and trace their effects through the modeled energy system.
Mapping the AEO 2026 experiments
For this analysis, the ten side cases fall into seven analytical categories. Some are paired high and low cases. Others examine a one-directional stress or the absence of specified regulations. The Combination case is different again: by applying the electricity and transportation policy changes together, it allows analysts to examine how those changes interact within the same modeled system.
Table 1. NEMS Connect’s analytical map of the AEO 2026 side cases
| Experimental category | Cases | Initial source of change | Principal analytical question |
| Macroeconomic growth | High and Low Economic Growth | Alternative assumptions for population growth and nonfarm labor productivity | How does faster or slower economic growth propagate through sectoral activity, energy demand, prices, and energy-system investment? |
| Domestic oil and natural gas supply | High and Low Oil and Gas Supply | Estimated ultimate recovery per well, undiscovered resources, and rates of production technology improvement | How do changes in domestic production economics propagate through fuel prices, electricity markets, industrial demand, trade, and infrastructure? |
| Zero-carbon technology costs | High and Low Zero-carbon Technology Cost (ZTC) | Alternative overnight capital-cost trajectories for specified technologies; fixed O&M costs also decline faster in the Low ZTC case | How do alternative cost trajectories reshape competition among technologies for power generation, carbon capture, and hydrogen production, with resulting effects on fuel use, prices, and emissions? |
| Electricity-demand stress | High Electricity Demand | An exponential trend in the installed stock of AI servers and more intensive electricity use to support data centers | How does the energy system respond when data-center-related commercial electricity demand grows more rapidly than in the Counterfactual Baseline? |
| Power-sector regulation | Alternative Electricity | Assumed absence of the April 2024 Clean Air Act Section 111 greenhouse gas requirements for affected generating units | How does the absence of these requirements change power-plant operation, retirement, and investment, with resulting effects on fuel use, carbon capture, and emissions? |
| On-road vehicle regulation | Alternative Transportation | Assumed absence of specified MY2027 and later federal fuel-economy and tailpipe-emissions standards, together with revised assumptions for manufacturer product availability and charging-infrastructure deployment | How do the combined assumptions about vehicle standards, product availability, and charging infrastructure reshape technology adoption and fuel demand, with resulting effects on electricity markets, refining, hydrogen, trade, and emissions? |
| Policy interaction | Combination | The Alternative Electricity and Alternative Transportation assumption packages applied together | Where are the joint effects additive, and where do cross-market interactions offset or reinforce the stand-alone effects? |
Note: The experimental categories and principal analytical questions reflect NEMS Connect’s interpretation of the cases described by EIA, and are not EIA classifications.
The categories clarify why comparisons across cases require care. A paired resource case, a demand stress, and a regulatory counterfactual do not represent equivalent distances from the baseline. They differ in structure, magnitude, and purpose, and each answers a different kind of question. The range across the cases therefore should not be interpreted as ranking the relative likelihood or importance of the underlying uncertainties.
From an initial change to a system response
Economic scale and concentrated demand
The High and Low Economic Growth cases begin with alternative assumptions about population and productivity, producing different pathways for employment, income, and economic output. Those changes affect energy consumption across buildings, industry, and transportation. The High Electricity Demand case introduces a change concentrated within the commercial sector: rapid growth in electricity use by data-center servers.
These cases provide an instructive comparison because EIA projects comparable average annual growth in electricity consumption through 2050 of 1.6% in the High Electricity Demand case and 1.5% in the High Economic Growth case, even though that growth arises from different economic structures. In the High Economic Growth case, additional electricity use accompanies broader growth in household, commercial, industrial, and transportation activity. In the High Electricity Demand case, by contrast, the incremental growth is concentrated in commercial buildings, largely because of additional data-center server and cooling loads.
Rapid data-center development also raises economic questions extending beyond its representation as an electricity load. It concentrates capital investment, construction activity, electricity infrastructure requirements, and demand for skilled labor in a rapidly developing sector. At the same time, the economic value and productivity of the resulting computing services are difficult to observe in real time. The historical data needed to estimate relationships among investment, labor and material inputs, computing output, productivity, and energy use may not yet provide a stable basis for long-term modeling.
The published cases were constructed to examine different questions. Read together, however, they bring the next research question into focus: how should macroeconomic models represent an emerging sector when its investment requirements are already visible but the value and productivity of its output remain uncertain?
Figure 2. Illustrative pathways through the integrated energy system
Fuel supply, prices, and technology competition
The High and Low Oil and Gas Supply cases enter NEMS primarily through domestic resource, recovery, and assumptions about technological improvement. Their effects do not stop with crude oil and natural gas production. Changes in production economics alter natural gas prices, which affect power-sector dispatch and investment, the competitiveness of wind and solar, the economics of new nuclear capacity, industrial fuel use, and the competitiveness of U.S. liquefied natural gas (LNG) exports.
The Low Oil and Gas Supply case is especially revealing because higher natural gas prices make new nuclear generation economical in EIA’s results in the 2040s. An upstream resource assumption thus becomes a technology investment outcome in the electricity sector. The result also suggests a further analytical question: how sensitive is the modeled nuclear response to the duration of high natural gas prices, the assumed costs of new nuclear capacity, and other constraints on deployment?
The ZTC cases approach part of the same competitive system from the technology-cost side. NEMS Connect has already applied those assumptions, combined with higher data-center electricity demand. That work showed that increased load does not carry a fixed generation or emissions consequence. The outcome depends on the available technologies to serve the load and their associated costs.
Our subsequent analysis followed the response beyond the power sector, showing how changes in generation can affect natural gas demand, production, prices, and LNG exports. The analytical lesson is broader than either the load or technology-cost case: NEMS does not stop at the power plant.
Policy changes and market interaction
The Alternative Transportation case changes both the regulatory environment and related assumptions about vehicle offerings and charging-infrastructure deployment. The first-order results include lower vehicle electrification and higher gasoline and diesel consumption relative to the Counterfactual Baseline. The effects then extend into refinery throughput, petroleum imports and exports, electricity sales, hydrogen demand, crude oil prices, and LNG exports.
The scale of that propagation is notable. EIA projects 9% lower electricity sales and 21% higher LNG exports in the Alternative Transportation case than in the Counterfactual Baseline in 2050. Fewer electric vehicles reduce electricity demand and the associated use of natural gas for power generation. At the same time, greater liquid-fuel demand raises Brent crude oil prices, widening the modeled economic margin for U.S. LNG exports.
Transportation-sector hydrogen use also falls to almost zero in the Alternative Transportation and Combination cases. This result suggests that the modeled emergence of hydrogen in transportation, particularly freight trucks, depends heavily on the represented vehicle standards rather than on hydrogen becoming independently competitive across a broad transportation market.
The Alternative Electricity case changes power-plant operating and investment conditions. Assuming the represented Section 111 requirements are not in place affects coal retirements, natural-gas plant utilization, capacity additions, carbon capture, fuel consumption, and emissions. One result is initially counterintuitive: EIA projects approximately 365 GW of cumulative natural-gas capacity additions through 2050 in the Alternative Electricity case, compared with approximately 430 GW in the Counterfactual Baseline. Because the affected new units are not limited to the represented 40% annual capacity factor, they can produce more electricity per unit of installed capacity, so less capacity is needed to serve a given level of demand.
The Combination case is valuable because connected markets need not respond additively. The Alternative Transportation assumptions reduce electricity demand and can make more natural gas available for LNG exports. The Alternative Electricity assumptions change coal use and natural-gas generation. Comparing the Combination case with the two stand-alone policy cases can therefore show whether the electricity-sector response reinforces, offsets, or redirects the transportation case’s effects on natural gas use, prices, and LNG exports. That comparison turns the Combination case from a simple policy package into a test of interaction across connected markets.
From individual cases to a research program
NEMS Connect’s AEO 2026 work to date illustrates how side-case analysis can proceed as a sequence. Our initial high server-load sensitivity asked how the energy system responds to a larger electricity-demand stress. Combining that load with alternative ZTC assumptions asked what determines how the additional demand is served. Following the response into natural gas markets showed how power demand and generation choices affect production, prices, and LNG trade. A separate Lower Asian LNG Demand sensitivity reversed the direction of inquiry by changing an international gas-market assumption and tracing the response back into U.S. natural gas markets and power-sector capacity decisions.
The common method is more important than any single result. Begin with a transparent assumption, separate the direct response from the propagated response, identify the mechanisms connecting the sectors, and then determine whether another sensitivity would add useful information. Not every result needs another model run. The discipline lies partly in recognizing which uncertainties are material, which mechanisms can be examined within NEMS, and which questions require a different level of geographic, temporal, or technological resolution.
Research directions from the official cases
Economic growth and structural change
The macroeconomic cases merit deeper examination not simply because GDP is higher or lower, but because rapidly developing sectors may alter the relationships among investment, labor demand, output, productivity, and energy use. Data-center development provides a current example. The next analytical step is to examine whether the economic structure represented in the broad GDP pathways adequately captures the investment, input requirements, and output of emerging sectors before a stable historical record exists.
Oil and gas supply transmission
The paired supply cases provide a strong platform for tracing how upstream resources and productivity move through production costs, natural gas prices, electricity investment, nuclear competitiveness, LNG exports, refinery operations, and regional infrastructure. Analysis should focus on these connections rather than treating production volumes as the final result, while distinguishing infrastructure that NEMS builds in response to market conditions from assumptions included in the case design, such as the additional crude oil pipeline and export capacity in the High Oil and Gas Supply case.
Transportation-policy propagation
The Alternative Transportation case warrants closer examination of how vehicle standards, manufacturer product-availability assumptions, and charging deployment propagate through vehicle adoption, hydrogen demand, refinery utilization, petroleum trade, electricity load, natural gas consumption, and LNG exports. The near-elimination of transportation-sector hydrogen use is particularly notable because it suggests that the modeled emergence of that market, especially for freight trucks, depends heavily on the Alternative Transportation assumption package.
Policy interactions in the Combination case
The Combination case may be the clearest test of the value of an integrated model. A useful decomposition would first calculate the difference between each stand-alone policy case and the Counterfactual Baseline. The two stand-alone effects could then be added and compared with the difference between the Combination case and the baseline.
Figure 3. Decomposing the Combination case
If the combined result differs materially from the sum of the stand-alone effects, the difference represents an interaction among the modeled responses. Examining that interaction by sector and over time would help identify where the electricity and transportation changes reinforce, offset, or redirect one another. Discrete investment decisions and market thresholds may be especially important because even modest changes in demand or prices can alter the timing or selection of capacity additions.
What the published suite does not isolate
The side cases are intentionally limited. Several cases change packages of assumptions rather than a single variable, so their results cannot always be attributed to an individual component. National results can also conceal regional transmission, pipeline, interconnection, or resource constraints. Long-run technology costs are not equivalent to the amount of equipment, labor, financing, permitting, or supporting infrastructure that can be delivered in a particular year. A technology can appear economic in a long-run model and still face binding deployment constraints.
These boundaries do not diminish the value of the cases. They help define the next research design. Some questions call for additional NEMS sensitivities. Others require regional transmission planning, production cost, power flow, resource adequacy, or international market models. The objective is to connect those tools and assumptions coherently, rather than asking a single model to address every geographic, operational, financial, and institutional dimension of the problem.
A further research direction: alternative world oil prices
Alternative world oil prices remain an important uncertainty for long-term energy analysis. AEO 2025 included High and Low Oil Price cases, while the AEO 2026 case suite does not. The High and Low Oil and Gas Supply cases examine uncertainty in U.S. resources, well recovery, and production technology, but they do not represent alternative global oil-market environments.
Additional analysis using alternative world oil-market assumptions and associated price paths could examine effects on U.S. transportation, refining, petroleum trade, domestic production, LNG competitiveness, and the macroeconomy. One particularly useful question is how materially different oil prices might affect electric-vehicle adoption independently of the vehicle standards examined in the Alternative Transportation case. Such analysis would complement the published cases and extend the research platform they provide.
From case results to structured inquiry
The AEO 2026 side cases are most useful when treated as the starting point for structured inquiry. Understand the changed assumptions, trace the integrated response, identify the mechanisms producing the results, recognize the model’s boundaries, and frame the next question carefully.
Used that way, the published cases provide more than a range of projections. They provide a transparent platform for learning how the U.S. energy system responds as economic, technological, resource, demand, and policy conditions change, and for determining which questions to examine next.
Sources
Official EIA sources
U.S. Energy Information Administration, Annual Energy Outlook 2026 Narrative. Official description and interpretation of the Counterfactual Baseline and side-case results. https://www.eia.gov/outlooks/aeo/narrative
U.S. Energy Information Administration, AEO 2026 Case Descriptions. Formal case names, scenario identifiers, and detailed modeling assumptions. https://www.eia.gov/outlooks/aeo/assumptions/case_descriptions.php
U.S. Energy Information Administration, AEO 2026 Release Presentation. Summary figures comparing selected cases across the economy and energy sectors. https://www.eia.gov/outlooks/aeo/pdf/AEO2026_Release_Presentation.pdf
U.S. Energy Information Administration, Annual Energy Outlook 2025. Documents the High and Low Oil Price cases used for comparison with the AEO 2026 suite. https://www.eia.gov/outlooks/archive/aeo25
Related NEMS Connect analysis
NEMS Connect, Data Centers Are Different: What a High-Demand Stress Case Shows about AI Load Growth. Introduces the higher server-load sensitivity used in the subsequent analytical sequence.
NEMS Connect, How Technology Costs Shape the Emissions Impact of High Data-Center Load. Combines a high-load sensitivity with alternative zero-carbon technology-cost assumptions.
NEMS Connect, Why Integrated Modeling Matters: From Power Demand to Gas Markets. Traces changes in the power-sector response into natural gas production, prices, and LNG exports.
NEMS Connect, Lower Asian LNG Demand. Changes a long-run international gas-market assumption and follows the response into U.S. natural gas and power-sector capacity decisions.