This section presents the findings from analyzing 240 near cost-optimal energy system pathways designed to achieve net-zero CO2 emissions by 2050. These pathways were developed using MGA and exhibit variations in fossil fuel use, levels of electrification in end-use, as well as the incorporation of other net-zero enabling technologies, such as hydrogen production and direct air capture.
2.1 Near cost-optimal decarbonization pathways eliminate coal, reduce petroleum use, and offer a broad range of natural gas futures.
Figure 1 illustrates the range of primary fossil fuel use, specifically coal, natural gas, and petroleum, within the near cost-optimal net-zero CO2 pathways along with the least-cost (deterministic) net-zero pathway for the U.S. energy system. The figure also shows the least-cost (deterministic) current-policy pathway, which includes the Inflation Reduction Act (IRA) provisions but excludes any other carbon constraints after the IRA provisions expire. Figure 1a shows that by 2050, most near cost-optimal pathways result in the near elimination of coal use by 2050, consistent with previous work.9,42,43 Although achieving net-zero targets while maintaining higher levels of coal use is theoretically possible, nearly 99% of the decarbonization pathways rely on less than 0.1 EJ of coal in 2050, representing a 98% reduction from current levels. In 2020, the U.S. electric sector accounted for approximately 90% of the coal use in the energy system, with the remainder attributed to the industrial sector.44 Consequently, Fig. 1a highlights that the pursuit of net-zero futures necessitates a rapid reduction in coal use in the electric sector, with the median pathway achieving a 71% reduction by 2030 compared to 2020. While most near cost-optimal net-zero pathways align with the least-cost deterministic pathway, leading to the elimination of coal in the electric sector by 2040, some pathways extend coal phase-out until 2050. This extension can be attributed, in part, to the availability of the IRA tax credits that enable continued electric sector coal use when combined with carbon capture and sequestration (CCS) technologies. As observed in the least-cost current-policy pathway, the IRA succeeds in reducing coal use while the tax credits are active. Still, there is a rebound to nearly pre-existing levels of coal use once these credits expire (typically by 2033 for most provisions).
Figure 1b shows that the distribution of natural gas use exhibits significant variation across the near cost-optimal pathways, with consumption in 2050 ranging from less than 4 EJ to 21 EJ, the latter being comparable to 2020 levels. This diversity in natural gas use is primarily driven by the industrial sector, which encompasses various applications, including direct air capture (DAC), hydrogen production through steam methane reforming (with and without CCS), and industrial manufacturing and non-manufacturing demands. In the least-cost current-policy pathway, industrial natural gas consumption increases steadily through 2050, surpassing the levels seen in any of the net-zero pathways. The range of industrial natural gas used in the near cost-optimal net-zero pathways often exceeds the results in the current-policy case until the last decade of the study period (i.e., 2040–2050). While natural gas use in the current-policy pathway may be comparable in magnitude to the decarbonization pathways, the drivers of such consumption differ. A substantial portion of natural gas is allocated to DAC or hydrogen production in the decarbonization pathways. For instance, the least-cost net-zero pathway uses about 8 EJ of natural gas for DAC in 2050, accounting for almost half of total natural gas use. In the current-policy pathway, these technologies are not widely adopted, and natural gas is primarily used for process heat or conventional boilers in the industrial sector. Within the electric sector, natural gas use undergoes a rapid reduction, declining from approximately 9 EJ in 2020 to less than 2.5 EJ by 2030 in over half of the near cost-optimal pathways. By mid-century, natural gas use in the electricity sector approaches zero in all modeled net-zero pathways. In the commercial sector, natural gas use remains relatively constant until 2035 across the pathways, after which it decreases to approximately one-tenth of the current commercial natural gas use by 2050 (0.2–0.5 EJ, Interquartile Range, IQR). By contrast, the transition from natural gas in the residential sector is slower, with levels remaining relatively constant until 2040 before declining to a median of 0.4 EJ (0–0.9 EJ, IQR). Overall, reductions in natural gas use across all decarbonization pathways are a consequence of the increased electrification of end-use technologies in all sectors. However, despite substantial reductions, all decarbonization pathways retain some level of natural gas use. Compared to coal and petroleum, natural gas is a lower-carbon alternative, particularly for challenging-to-decarbonize processes in the industrial sector. It is important to acknowledge that factors not considered in Temoa, such as labor impacts and energy security perceptions, will likely influence natural gas use during a low-carbon transition.
Primary energy use from petroleum products, shown in Fig. 1c, consistently declines from 31 EJ in 2020 to 3.7 EJ (3.1–4.1, IQR) in 2050 across the decarbonization pathways. The transportation sector exhibits a steady decrease within a relatively narrow range of outcomes. By 2050, all pathways use more than 2.5 EJ of petroleum for transportation but less than 4.8 EJ, with the deterministic least-cost net-zero pathway falling on the higher end of this range. By contrast, petroleum use in the transportation sector reaches 10 EJ by 2050 in the least-cost current policy pathway (i.e., without a net-zero constraint). The adoption of electric vehicles (EVs), synthetic liquids, and hydrogen (discussed in Section 2.3) drives the reduction in petroleum use in the net-zero pathways. In the absence of additional decarbonization policies, substantial petroleum use continues through 2050.
2.2 Decarbonization pathways consistently rely on solar and wind resources with battery storage, enabling a movement toward increased end-use electrification.
Figure 2a illustrates that total electricity consumption within the near cost-optimal pathways consistently increases, reaching a median total use of 9,800 TWh (9,500 − 10,500 TWh, IQR) by 2050. This growth is a three-fold increase in electricity use compared to current levels. While the range of total electricity consumption varies, all decarbonization pathways necessitate substantial and rapid investments in clean electric generation capacity to fulfill the needs of end-use sectors. Figure S4 in the Supporting Information (SI) disaggregates the electricity consumption by sector. Electricity demand in the least-cost net-zero pathway reaches 9,200 TWh in 2050, placing it at the lower end of the distribution of all the near cost-optimal pathways. Although some pathways exhibit lower electricity consumption than the least-cost option, most pathways lean toward higher relative electricity use by 2050. By contrast, the least-cost current-policy pathway remains similar to net-zero trajectories until 2035 but diverges from the net-zero pathways after that year. Most provisions of the IRA expire by 2033. Without additional policy interventions after the IRA expires, electricity consumption would experience only modest increases to meet population and economic growth.
Figures 2b-f offer insights into the evolving generation sources within the power sector, and Figure S2 in the SI shows the capacities. Figure 2b shows that even the most conservative decarbonization pathways require a ten-fold increase in solar generation by 2050, compared to current levels. This level of solar, totaling 3,800 TWh (3,600–4,100 TWh, IQR), nearly matches the total power system generation from all sources in 2020. This trend highlights the magnitude of transformation required for deep decarbonization and emphasizes the pivotal role of solar power in a decarbonized power system. Across the modeled pathways, the greatest relative increase in solar generation occurs between 2025 and 2030, with a three-fold increase spurred by the federal Production and Investment Tax Credits (PTC and ITC) available through the IRA. Figure 2c shows that generation from wind also experiences substantial growth, reaching a median of 6,800 TWh (6,100–7,400 TWh, IQR) in the net-zero pathways. Most of the wind generation and capacity comes from onshore resources, with a smaller contribution from offshore resources that produce a median100 TWh (80–110 TWh, IQR) by 2050. Much like solar, the PTC and ITC incentives drive a two-fold or more increase in wind generation between 2025 and 2030. The rapid expansion of wind and solar power highlights the need for substantial and sustained investments in integrating these variable resources to achieve net-zero CO2 emissions effectively. The growth of wind and solar in the net-zero pathways is complemented by the expansion of battery storage, with installed battery capacity projected to reach 770 GW (740–790 GW, IQR) by 2050, compared to less than 1 GW in 2020 (Fig. 2f). Figure S3 in the SI further shows that 94% of power generation in the median pathway comes from renewables by 2050 (93–95%, IQR), consistent with previous research.42 Furthermore, variable renewable sources constitute an increasing share of all renewable generation, reaching 96% by 2050.
Notably, no new nuclear infrastructure is brought online across the decarbonization pathways. However, existing nuclear capacity remains available across all the near cost-optimal decarbonization pathways, contributing between 270 and 710 TWh of generation and providing approximately 100 GW of capacity in 2050 (Fig. 2d and Figure S2 in the SI). By contrast, most existing nuclear infrastructure retires by 2035 in the least-cost current policy pathway. This pattern suggests that while nuclear generation is an important component of a net-zero power sector capable of providing firm power, its continued use without an emissions constraint is not economical after the expiration of the IRA tax credits.45 When considering both renewable energy and nuclear power, the median contribution to power generation is 98.1% (97.9–98.5%, IQR) by 2050, underscoring the importance of a low-carbon power system (Figure S3).
Figure 2e shows that electricity generation from natural gas decreases from present levels (~ 1,300 TWh) to nearly zero for most near cost-optimal decarbonization pathways by 2050. This decrease is not strictly monotonic, as a temporary increase in natural gas electricity generation occurs between 2030 and 2035 when federal IRA tax credits expire. The rebound in natural gas consumption for power generation is most prominent in the least-cost current-policy scenario, exceeding 1,000 TWh in 2050. Consequently, additional policy measures beyond the IRA will be essential to fully decarbonize the power sector.
Substantial investments in electricity transmission capacity will be necessary to support the high levels of electrification in the net-zero pathways. Based on the regional representation of the U.S. energy system in Temoa (Figure S1 in the SI), the results suggest the need for 47 GW (46–49 GW, IQR) of new inter-regional transmission lines between California and the Southwest by 2050. Substantial transmission expansion also occurs between the Central and North-Central regions, totaling 50 GW (45–57 GW, IQR), with the highest pathway reaching 69 GW. Other pathways also indicate the need for transmission expansion in various regions, such as between California and the Northwest and between the Southwest and the Northwest. New transmission into and out of the Northeast, Mid-Atlantic, Southeast, and Texas is comparatively small. The consistent deployment of this transmission capacity in the near cost-optimal pathways highlights the benefits derived from inter-regional electricity transfer in these regions.
2.3 Hydrogen is consistently used to meet hard-to-decarbonize sectors' demands across near cost-optimal decarbonization pathways.
Hydrogen has the potential to play an important role in decarbonization efforts, particularly in the industrial and transportation sectors.46,47 Consistent with other studies that identify hydrogen as a key component of a net-zero future, Fig. 3a shows a median production of 7.7 EJ (5.6–10.2 EJ, IQR) by 2050.42 While the least-cost net-zero pathway shows 3.5 EJ of hydrogen production in 2050, this value is at the lower end of a broad distribution of hydrogen production across the near cost-optimal net-zero pathways.
Figures 3b-d show that the primary mechanisms for hydrogen production up until 2040 are from natural gas steam-methane reforming and bioenergy with carbon capture and storage (BECCS). In approximately 25% of the near cost-optimal pathways, steam methane reforming with CCS produces at least 1 EJ of hydrogen, with larger amounts produced in a limited number of pathways. However, in 2045 and 2050, electrolysis emerges as a cost-competitive alternative, supplanting hydrogen production from natural gas steam-methane reforming. The IRA tax credit, Internal Revenue Code section 45V, incentivizes green hydrogen production with a subsidy of up to $3/kg of H2. In the database used for this analysis, both electrolysis from new renewables and BECCS qualify for the full credit, while steam methane reforming with CCS qualifies for a $1/kg of H2 credit. These incentives result in an increase in total hydrogen production from 2030 onwards. While hydrogen proves to be an attractive energy carrier, the chosen pathway to production is sensitive to emissions and cost assumptions. The carbon dioxide removal benefits from BECCS prove to be an attractive way to take advantage of the federal tax credits and meet CO2 constraints.
Figures 3e-h illustrate that hydrogen is primarily used in the transportation sector, followed by industrial applications. In the transportation sector, 1.4 EJ (1.2–1.7 EJ, IQR) of hydrogen are used in 2050 for fueling fuel-cell vehicles, particularly heavy-duty vehicles. Additionally, 2.3 EJ (0.4–4.2 EJ, IQR) of hydrogen is used for synthetic fuel synthesis, serving several transportation demands. In the industrial sector, the primary role of hydrogen is to replace conventional boilers and meet the demand for process heat. Hydrogen use for industrial processes converges around 1.0 EJ in 2050 across all net-zero pathways. Hydrogen is used for synthetic natural gas production for heating in the residential and commercial sectors only in the final decade, with a median 2050 hydrogen use of 1.2 EJ (1.0–1.3 EJ, IQR). While hydrogen is a viable option for electricity generation in combined cycle power plants, it was seldom chosen in the near cost-optimal net-zero pathways. However, in a subset of these pathways, substantial amounts of hydrogen are used for electricity generation in 2025. The presence of two tax incentives in the IRA, one for hydrogen production and the other for clean electricity generation facilitates this choice. As the IRA provisions expire, the use of hydrogen-enabled power plants diminishes in future time periods, but the capacity remains to meet power sector reserve margins.
2.4 Carbon dioxide removal and management technologies span a wide range of deployment across near cost-optimal decarbonization pathways.
As detailed in Section 2.1, all of the net-zero CO2 emissions scenarios retain some fossil fuel use across in 2050. The resulting residual CO2 emissions are primarily from hard-to-decarbonize sectors like aviation or high-temperature processes in the manufacturing sector. Given these residual emissions, carbon management, and particularly carbon dioxide removal (CDR), is likely to play a pivotal role in enabling net-zero futures. Figure 4 shows the carbon management technologies represented in the net-zero pathways, including bioenergy with carbon capture and storage (BECCS), coal and natural gas electricity generation with carbon capture and storage (CCS), and direct air capture (DAC). The near cost-optimal pathways exhibit a wide range of potential deployment levels for these technologies, with some pathways more heavily reliant on carbon mitigation options.
Figure 4a shows that most near cost-optimal pathways have a notable reliance on BECCS. BECCS can be an attractive tool in decarbonization efforts, as it couples the production of energy carriers (electricity or hydrogen) that can then meet service demands across the energy system with carbon dioxide removal. While the contribution to electricity generation from BECCS remains minimal in the near cost-optimal pathways, many pathways incorporate hydrogen production (Section 2.3). Coal power with CCS and natural gas steam methane reforming with CCS are not deployed in the least-cost current-policy or the least-cost net-zero pathways. By contrast, Figs. 4b-c depict that these technologies are extensively deployed in a small subset of near cost-optimal net-zero pathways. Overall, total CCS, calculated as the sum of BECCS, coal CCS, and natural gas CCS, amounts to 680 Mt CO2/yr (590–860 Mt CO2/yr, IQR) in 2050 in these pathways (Fig. 4d).
Figure 4e shows that the median DAC deployment in 2030 (when the technology first becomes available in the model) is 0.52 Gt CO2/yr (0.14–1.29 Gt CO2/yr, IQR). Currently, DAC is a nascent technology and has not been deployed on a large scale. However, the IRA tax credit (Internal Revenue Code section 45Q) can incentivize the adoption of DAC. Even in the least-cost current-policy scenario (i.e., without a net-zero requirement), DAC is employed to take advantage of the available tax credits. Consequently, the model builds and uses DAC capacity while these tax credits remain in effect until 2033. The tax credits catalyze capital investments, and the infrastructure continues to be used beyond the expiration of the credits. In 2050, DAC use is expanded in the net-zero pathways to compensate for residual CO2 emissions from hard to decarbonize processes,48 reaching 1.3 Gt CO2/yr (1.0–1.5 Gt CO2 /yr, IQR). This wide range of outcomes suggests that DAC deployment is sensitive to cost shifts and incentives, at times serving as a backstop in achieving net-zero targets in response to changes in the rest of the energy system.
Figure 4f displays the range of total geologic sequestration, spanning 0.9 to 1.9 Gt CO2/yr in 2050, with a median result of 1.7 Gt CO2/yr. The least-cost net-zero pathway prioritizes the extensive sequestration of CO2 rather than using it for synthetic fuel production.49 However, near cost-optimal pathways indicate that net-zero futures are possible with lower levels of geologic sequestration of CO2.
2.5 Near cost-optimal net-zero pathways can differ substantially from a deterministic cost-optimal pathway.
To understand the characteristics of near cost-optimal decarbonization pathways in more detail, we used clustering approaches to identify “illustrative” pathways that represent groups of decarbonization pathways. These illustrative pathways offer insights into key differences between the least-cost solution and solutions obtained with a 1% slack on the total system cost. As described in Section S2 of the SI, the pathways are identified using k-means clustering on the near cost-optimal decarbonization pathways. Figure 5 shows results for six illustrative pathways that differ in the deployment of hydrogen, DAC, and energy system-wide electricity use. These groups were chosen as they represent important levers in decarbonization efforts.50–52
Figure 5a presents carbon dioxide emissions for the chosen illustrative pathways. While the timing and magnitude of mitigation measures vary across the selected pathways, there are some observable common trends. The CO2 constraints in this study apply a linear reduction to net-zero emissions by 2050. However, these limits are not binding in 2030 due to the decarbonization impacts of the IRA. Further, the power sector is generally the first to decarbonize, consistent with other studies.52 The pathways representing low hydrogen and high electricity (Low H2 and High Elec in Fig. 5) deploy coal power but mitigate these emissions with CCS. There are also commonalities in the primary energy consumption of illustrative pathways shown in Fig. 5b. Increased deployments of solar, wind, and biomass accompanied by reduced or eliminated coal and petroleum use are ubiquitous. All cases greatly reduce or eliminate power sector, transportation, and building emissions but must contend with residual CO2 emissions from hard-to-decarbonize industrial processes and upstream fuel emissions.
Carbon management is required in all illustrative pathways, but there is heterogeneity in the technologies chosen for this purpose. For example, the low hydrogen and high electricity pathways rely on carbon management from several technologies. In contrast, other pathways, such as the ones representing high hydrogen or DAC use, rely heavily on one carbon management option. The carbon removal in the high DAC case allows for higher emissions across the energy sector, resulting in higher petroleum consumption and lower biomass use compared to the low DAC pathway. More carbon management is required in the low electricity pathway, with increased DAC use driving higher natural gas consumption compared to the high electricity pathway. Hydrogen can play an important role as a low-carbon energy carrier. Its absence in the low hydrogen pathway results in higher transportation and industrial emissions. The pathway with high hydrogen use shows increased biomass consumption compared to the low hydrogen case, indicative of the hydrogen production process via BECCS.
2.6 Near cost-optimal pathways expose tradeoffs and complementarities between decarbonization technologies.
Different energy sources and end-use technologies tend to be used more frequently alongside or in the absence of other options. Figure 6 explores the relationships between select technology deployments across the near cost-optimal decarbonization pathways in 2050, showing correlations among energy sources and carriers (Fig. 6a), among end-use technologies (Fig. 6b), and between these two groups (Fig. 6c). A positive correlation indicates that the technologies are more often deployed together, while a negative correlation suggests potential competition between technologies. The results only show the strength of the correlation. Two technologies may have a positive correlation even if both have low deployment levels or overall use is decreasing.
In Fig. 6a, solar generation, wind generation, and battery capacity demonstrate a notable positive correlation. This association stems from the critical role batteries play in ensuring reliability as the adoption of variable renewable energy sources like solar and wind increases. Hydrogen is produced extensively via electrolysis in the near cost-optimal net-zero pathways in 2050. This pattern drives the positive correlations between hydrogen production with wind generation, solar generation, and battery capacity, as additional renewable generation provides a carbon-free energy source for hydrogen electrolyzers. Hydrogen is most negatively correlated with petroleum due to fuel switching in the transportation sector, particularly for heavy-duty vehicles. While hydrogen commonly replaces natural gas in the industrial sector, this negative relationship is dampened due to pathways in which hydrogen is produced via steam methane reforming. Additionally, biomass exhibits a positive correlation with hydrogen, given the prevalence of BECCS for hydrogen production. The positive correlation between biomass and synthetic liquids follows from the link between hydrogen and synthetic liquids.
When considering end-use technologies, EVs compete with hydrogen and internal combustion vehicles to meet transportation demand (Fig. 6b). Competition also exists between heat pumps and natural gas heaters in the buildings sector, as well as electric and hydrogen boilers in the industrial sector. Scenarios with more internal combustion engine (ICE) vehicle use are positively correlated with more natural gas heating of buildings. The persistence of these technologies is accompanied by increased DAC use, allowing net-zero emissions targets to be met. In pathways with more deployment of EVs or heat pumps, less DAC use is required. Moreover, electric industrial boilers and heat pumps exhibit a positive correlation with EVs, indicating a trend where the electrification of various end-uses is interconnected. This counters the notion of exclusion or competition among different electrification methods for end-uses.
Figure 6c illustrates the correlation between end-use technologies and energy sources/carriers. Within the transportation sector, EV adoption shows a clear connection to increased renewable electricity generation, displaying positive correlations with wind and solar energy sources. Conversely, EV adoption exhibits negative relationships with petroleum, synthetic fuels, natural gas, and biomass utilization. The integration of hydrogen vehicles also contributes to increased demand for renewable electricity and synthetic liquids but tends to reduce petroleum use. There is heightened reliance on synthetic liquids and natural gas in net-zero pathways where more ICE vehicles persist. A weak negative correlation exists between hydrogen and DAC, shaped by two competing patterns. When low- or zero-carbon hydrogen is introduced as an alternative energy carrier, there is a reduced dependency on DAC. However, when hydrogen facilitates the production of synthetic liquid fuels, DAC becomes necessary to capture resulting CO2 emissions. Notably, these synthetic fuels display positive correlations with wind and solar energy sources used in generating the hydrogen feedstock.