How power-sector technology costs affect natural gas demand, prices, production, and LNG exports
The power-sector result is only the beginning
Data-center electricity demand is emerging as an important source of uncertainty in U.S. energy planning. In an earlier NEMS Connect analysis, we increased data-center server electricity use beyond the U.S. Energy Information Administration’s High Electricity Demand case to examine how the energy system might respond to even faster load growth. By 2050, server-purchased electricity reached approximately 1,086 billion kilowatt-hours in the NEMS Connect High Electricity Demand Stress sensitivity.
That analysis showed that higher load affects more than electricity sales. It changes generation, capacity additions, fuel consumption, prices, and emissions.
A second NEMS Connect analysis then held the high server-load assumption approximately constant while changing the cost assumptions for zero-carbon technologies. Lower zero-carbon technology costs shifted the power system toward substantially more renewable generation and storage, while higher costs increased reliance on natural gas.
This article uses those same high-load scenarios to follow the power-sector response into natural gas demand, prices, production, and LNG exports.
The central question is:
How does a change in the power-sector generation mix affect natural gas prices, production, and LNG exports?
The answer illustrates one of the main advantages of an integrated model such as the National Energy Modeling System. A change in the generation mix affects fuel consumption; fuel consumption affects prices and production; and those changes can affect trade. The consequences of a power-sector assumption therefore do not stop at electricity generation or carbon dioxide emissions.
Holding high data-center load approximately constant
The comparison carries forward the three NEMS Connect sensitivities introduced in the previous article:
- HED Stress: the higher data-center electricity-demand sensitivity
- HED + Low ZTC: the same high-load stress combined with EIA’s Low Zero-carbon Technology Cost assumptions
- HED + High ZTC: the same high-load stress combined with EIA’s High Zero-carbon Technology Cost assumptions as an illustrative high-cost comparison
Data-center server purchased electricity reaches approximately 1,086 to 1,089 BkWh in 2050 across the three sensitivities. The comparison therefore holds the high-load question approximately constant and isolates how different technology-cost assumptions change the system response. The previous article examined the resulting generation, capacity, electricity-price, and emissions outcomes. Here, the analysis follows those outcomes into the natural gas market.
From electricity generation to natural gas demand
The first effect is direct, in that changes in natural gas generation produce corresponding changes in power-sector natural gas consumption. In the HED Stress sensitivity, renewable generation reaches about 3,300 BkWh in 2050. Under the Low ZTC assumptions, it rises to about 4,272 BkWh, while natural gas generation falls from about 2,946 BkWh to 2,035 BkWh. Under the High ZTC assumptions, renewable generation falls to about 3,116 BkWh and natural gas generation rises to about 3,117 BkWh.
The increase or decrease in renewable generation helps determine how much gas-fired generation is needed. The resulting change in gas-fired generation then translates into a corresponding change in power-sector natural gas consumption.
Figure 1. Technology-cost effects propagate from power generation to gas markets
The Low ZTC sensitivity provides the clearest illustration of how the power-sector response propagates into natural gas markets.
Lower zero-carbon technology costs make renewable generation and storage more competitive. The model builds and uses a different power-sector capital stock, reducing the amount of natural gas required for electricity generation. Power-sector gas consumption falls from approximately 20.0 quadrillion British thermal units in the HED Stress sensitivity to about 13.9 quads in the Low ZTC sensitivity.
Lower domestic gas demand changes prices and production
The decline in power-sector gas consumption reduces overall demand in the domestic natural gas market.
In the Low ZTC sensitivity, the Henry Hub natural gas price falls from approximately $4.87 per million British thermal units in the HED Stress sensitivity to about $4.50/MMBtu in 2050, expressed in 2025 dollars. Dry natural gas production falls from approximately 54.9 trillion cubic feet to 50.7 Tcf.
This result is important because the reduction in power-sector gas demand does not simply leave an equivalent quantity of gas available for another use. The market adjusts.
Lower demand reduces prices. Lower prices, in turn, reduce the economic incentive for marginal production. As a result, part of the decline in power-sector consumption is absorbed through lower domestic production.
The High ZTC sensitivity moves in the opposite direction. Greater reliance on gas-fired generation raises power-sector gas use to approximately 21.1 quads. Henry Hub rises to about $5.07/MMBtu, and dry gas production increases to approximately 56.0 Tcf.
The power-sector technology assumption therefore affects both the demand side and the supply side of the natural gas balance.
Why LNG exports increase when domestic gas use falls
The LNG result extends the analysis beyond domestic power and gas markets, while also pointing to the next set of questions.
In the Low ZTC sensitivity, LNG exports increase from approximately 10.6 Tcf in the HED Stress sensitivity to about 12.0 Tcf in 2050, an increase of roughly 13 percent.
At first glance, it may appear counterintuitive that LNG exports rise while dry gas production falls. The explanation lies in the interaction among domestic demand, price, production, and international trade.
Lower power-sector gas consumption reduces pressure on the domestic gas market, and Henry Hub prices fall. Although producers respond by reducing output, the lower domestic price also supports a larger LNG-export volume. LNG exports consequently increase even though total dry gas production is lower. The result reflects the interaction of domestic prices, production economics, export capacity, and the international market conditions represented in NEMS.
This does not mean that every unit of gas displaced from electricity generation becomes an LNG export. Much of the adjustment occurs through reduced production. The result instead reflects a new equilibrium across domestic consumption, prices, production, and trade.
The High ZTC case is less symmetrical. Greater power-sector gas demand raises prices and production, but LNG exports remain approximately unchanged at the precision shown in the model results. Determining why the export response is limited would require examining the underlying LNG capacity, utilization, international demand, and other model constraints in greater detail.
The LNG response is therefore conditional. It depends not only on the amount of domestic gas available, but also on prices and on the broader market conditions governing U.S. exports.
Within NEMS, the increase in LNG exports is a downstream result of changes that begin in the power sector and move through domestic gas demand, prices, and production. It is not the end of the broader analytical chain. Evaluating how additional U.S. LNG exports affect global prices, suppliers, consuming regions, and emissions would require a complementary global gas-market framework, as used in DOE’s recent LNG analysis.
A power-sector assumption can become a trade result
No LNG-export assumption was changed to produce this result.
The sensitivity began with assumptions about zero-carbon technology costs under high data-center electricity demand. Those assumptions changed the generation mix. The generation mix changed power-sector natural gas consumption. Gas consumption affected Henry Hub prices and domestic production. Those changes, in turn, affected LNG exports.
That sequence is why integrated modeling matters.
A power-sector model might estimate generation and fuel burn. A separate gas-market model might estimate production or prices. NEMS links these outcomes within a common framework so that electricity demand, generation investment, fuel consumption, prices, production, and trade respond together.
The result also broadens how the energy implications of data-center growth should be considered. The issue is not limited to whether higher electricity demand raises power-sector emissions. Depending on the technologies used to serve that load, the consequences may extend into domestic natural gas markets and international LNG trade.
The modeling insight
High data-center load creates a need for additional electricity supply, but the fuel-market consequences depend on how the power sector meets that need. In these sensitivities, lower zero-carbon technology costs increase renewable generation, reduce power-sector natural gas use, lower Henry Hub prices and production, and support higher LNG exports. Higher technology costs move most of those effects in the opposite direction, although the LNG response is more limited.
These are not official EIA projections or probability-weighted forecasts. They are independent diagnostic sensitivities using the public AEO 2026 NEMS framework. Their value lies in showing the direction and approximate scale of linked energy-system responses under specified assumptions.
The broader lesson is straightforward:
NEMS does not stop at the power plant. It follows the consequences into fuel demand, prices, production, and trade.
From one result to the next question
This analysis has followed the response far enough to show how a power-sector technology assumption can become a natural gas-market and LNG-export result. The increase in LNG exports also raises broader questions about global gas markets, competing suppliers, destination regions, and emissions. Those questions would require complementary global analysis and are beyond the scope of this sensitivity.
Closer to the original high-load case, the results also suggest that the next questions are not simply about whether data-center electricity demand will be higher or lower. They concern how that demand behaves, where it appears, what resources are available to serve it, and which constraints matter most.
The next article will examine those research directions in greater detail.
More broadly, asking these questions in a structured sequence is a learned discipline in its own right. The objective is not to pursue every possible extension, but to identify which uncertainties are most consequential, frame sensitivities that can isolate them, interpret the resulting system responses, and recognize when the analysis has gone far enough for the decision at hand.
That process is central to the effective use of NEMS. NEMS provides an integrated framework, but the value of the analysis depends on the quality, sequencing, and interpretation of the questions posed within it.
Related NEMS Connect analysis
The first article develops the higher data-center electricity-demand stress case and examines its effects on generation, capacity, prices, and emissions.
The second article examines how alternative zero-carbon technology-cost assumptions change the power-sector response to approximately the same high server load.