“If you think about energy in your own life, you might wonder about gasoline and electricity prices, or the cost of renewables, or whether to buy an electric vehicle,” said Joe DeCarolis, head of CMU’s Department of Engineering and Public Policy and co-director of the OEO along with Paulina Jaramillo, a professor of engineering and public policy at CMU, and Jeremiah Johnson, a professor in civil, construction and environmental engineering at NC State University. “We feed the model input data, and it optimizes the deployment and utilization of energy technologies across the system.”

Temoa is a mathematical model implemented in Python that operates on a large input dataset to produce projections about the US energy system years to decades in the future.

Since factors like public policy, resource availability and pricing, and new technologies are constantly evolving, Temoa’s database requires regular upkeep to ensure that its analyses are as relevant and accurate as possible.

One of the biggest challenges with a model like this is that it’s very data-intensive. It’s a continuous process to keep the database refreshed, and it requires a lot of resources.

Joe DeCarolis, Co-Director, Open Energy Outlook

With support from the Scott Institute’s 2025 hardware and software tool upgrade program, the OEO team updated Temoa’s framework to portray how growing energy demands — particularly from data centers and cryptocurrency mining — could affect electricity costs in the US. The OEO team gathered information about where data centers are located around the country and the rate at which electricity demand could grow over time, to be applied towards the analysis.

Enabled by this new information, the OEO published a paper in Environmental Research Letters in May 2026 which estimated that electricity prices could rise anywhere between 6 to 29 percent over the next several years in areas around the U.S. that have clusters of data centers.

The team models hourly demand to estimate the impact of data centers on electricity consumption, prices, and emissions. This report is the latest example of how Temoa, first created by DeCarolis at NC State in 2010, can be a resource for public understanding and policymaking.

The Open Energy Outlook was made possible through generous support by the Alfred P. Sloan Foundation.

A warning sign came in December 2024, when capacity market clearing prices in PJM Interconnection jumped from $30 - $270 dollars per MW-day. That ninefold increase sent shock waves through the Mid Atlantic and Midwest and will ultimately show up in bills for roughly 67 million customers across 13 states. State leaders reacted strongly. But the price spike was not a random event. It was a market signal.

The suspected driver is the rapid growth in electricity demand from data centers and cryptocurrency mining operations. Recent estimates suggest that electricity use from these sources could increase 350 percent between 2020 and 2030, growing from about 4% of total U.S. electricity consumption to 9 percent. That is an extraordinary shift in a single decade for a sector that already operates at scale.

Through the Open Energy Outlook Initiative, my colleagues and I have modeled what this growth means under current policies and market structures, including the provisions of the Inflation Reduction Act that may be at risk. Our findings are clear. If we stay on our current path, wholesale electricity costs increase by an average of 8% nationally. In some regions the impacts are far larger. In Central and Northern Virginia, where the concentration of data centers is highest, we project cost increases exceeding 25% by 2030.

These cost increases are not abstract modeling artifacts. They reflect real grid planning challenges. In parts of the PJM region, more than 25 gigawatts of aging coal capacity that might otherwise retire are kept online to meet incremental demand from data centers. In the short run, running these plants can be less expensive than building new gas, renewable, or nuclear capacity. But that choice has consequences for both consumers and emissions.

On the emissions side, the numbers are sobering. Under current policies, we estimate that additional data center and cryptocurrency mining load could increase power sector carbon dioxide emissions by 30 percent in 2030 relative to a case without that load growth. In absolute terms, that is about 275 million metric tonnes of carbon dioxide per year by 2030. To put that in perspective, that is roughly equivalent to the annual emissions of France.

The regional story also matters. In PJM, added demand is largely met by coal and natural gas, and some of the emissions are effectively shifted across state lines. This creates a form of carbon leakage where one state’s demand growth drives emissions in another state’s generators, complicating state level climate targets. By contrast, in Texas, targeted transmission investments and abundant wind resources make it more feasible to serve incremental load with lower emission generation. The same national trend can produce very different outcomes depending on regional resource availability and planning decisions.

What makes this moment particularly challenging is the speed of change. Traditional planning frameworks assume demand growth of one to two percent per year. In some localized areas, demand tied to data centers is growing at rates of 20% to 30% per year. Utility resource planning processes take years. Interconnection queues are already backlogged. Market mechanisms that are supposed to signal new investment are struggling to keep pace with sudden demand shocks.

The core issue is not whether data centers are good or bad. Digital infrastructure brings substantial benefits, including increased productivity from AI-based technologies. The question is how to integrate this growth into the power system without imposing disproportionate costs on consumers or undermining emissions goals.

There are policy tools available. Cost allocation and rate design reforms can ensure that large new loads contribute fairly to infrastructure investments. Strategic siting and stronger incentives for clean energy procurement can reduce pressure on carbon intensive generation. Transmission planning reform could unlock renewable rich regions and fundamentally alter projected cost and emission outcomes. Greater demand flexibility requirements, particularly for cryptocurrency mining and certain data center operations, may offer additional system benefits if implemented carefully.

None of these solutions is simple. They require coordination across federal, state, and regional authorities, as well as between public institutions and private firms. But the alternative is to let markets lurch from one price spike to the next while emissions climb.

When I look back at the December 2024 capacity auction in PJM, I see it as an early warning signal. The grid is telling us that something fundamental is changing. Our modeling suggests that higher costs and emissions are a likely outcome. But they are not inevitable. With thoughtful, proactive policy, we can accommodate digital economy growth while protecting affordability and continuing progress toward cleaner electricity.

Over the last two years, four leading research teams set out to harmonize their electric power system model assumptions to produce more robust policy support. Teams harmonized model results for two scenarios: one that reflects only current policies and one that constrains power sector emissions to an economy-wide net-zero emissions scenario. For these scenarios, models were configured with the same numerical inputs. For example, each team assumed U.S. electricity demands would increase by 72% between 2027 and 2050, consistent with strong electrification. They also made similar assumptions related to the numerical values of parameters — such as technology costs and the economic discount rate — as well as the alternative technologies available to meet growing electricity demands.

Teams estimated that building and operating a U.S. energy system under current policy would cost about $3.73 trillion (in net present value) and emit about 26.3 billion tonnes of CO2 between 2027 and 2050. In contrast, a net-zero energy system would cost $5.15 trillion and emit about 4.46 billion tonnes of CO2 over this period, declining to about 110 million tonnes per year in 2050. Why does a net-zero energy system still have positive emissions? In a net-zero scenario, emissions from some hard-to-abate activities, like those in heavy industry, are offset by negative emission technologies like carbon dioxide removal.

Once the models were harmonized, the teams were ready to experiment with different decarbonization policies and model configurations, evaluating alternatives against the costs and cumulative emissions associated with current policy over the planning horizon of 2027 to 2050.

This same logic applies when estimating the cost of reducing greenhouse gas emissions. While some analyses have quantified the cost of various abatement options in isolation to identify those that are least cost, this approach fails to reflect the interdependencies of our energy system. You can add solar power to displace coal generation and estimate the avoided emissions and costs, or deploy a fleet of electric vehicles to replace gasoline cars and calculate the same metrics. However, these actions are interconnected — adding solar affects the grid mix that powers EVs, and widespread EV adoption changes electricity demand. To fully understand the emissions reductions and cost impacts, you must consider the entire system’s response rather than evaluating each action in isolation.

Through an interdisciplinary collaboration including engineers and economists, we recently tackled this issue using a detailed technology-rich energy system optimization model to estimate the marginal abatement costs of reducing emissions in the United States. Through this systems-based approach, we estimated the emissions reductions we would achieve at increasing abatement costs, while identifying the technologies deployed at each stage. Figure 1 shows these emissions trajectories for the nation as a whole (left), as well as regional emissions (right). Each line represents one marginal abatement cost applied consistently over time, showing that higher costs on emissions lead to lower emissions, as expected.

A few things jump out. Some emissions cuts are basically free. Relative to 2020 emissions levels (shown as the dashed lines), our reference case shows a drop in CO2-eq emissions of about 19% by 2050. Why? Because cleaner technologies are cheap and cost competitive on their own. There are many solar and wind projects being built today to reduce costs. We also see that $40 per ton CO2-eq gets us pretty far, cutting emissions by 36% by 2050, while ramping up wind and solar generation and making some headway in electrifying space heating and transportation. At $100 per ton, we’re entering deep decarbonization territory, with a 63% drop in emissions, coal almost entirely out of the picture, substantial reductions in the industrial sector, and more widespread electrification. But getting to net-zero is tough and occurs at costs near $400 per ton. Displacing those last emissions from industry, aviation, and buildings requires higher-cost carbon management solutions like direct air capture.

We see that under these various emissions abatement futures, not all regions are created equal. Texas, the Southeast, and the Central United States (think: Great Plains states) are prime spots for either geological storage of carbon dioxide or bioenergy growth — options that may play important roles in achieving the final step to net-zero. Meanwhile, regions like the Mid-Atlantic and Northeast have a harder time fully decarbonizing, with persistent emissions in sectors including residential space heating.

Figure 2 shows a different view of these results, with cumulative 30-year emissions reductions under increasing marginal abatement costs (left) and annual snapshots (right). Again, we see that as emissions costs increase, there are diminishing returns in avoided emissions, illustrating that emissions abatement is increasingly costly as the energy system decarbonizes. However, the emissions avoided at a given cost varies substantially over time. The mitigation achieved under lower costs doesn’t drop much over time, with mature mitigation options dominating. However, higher abatement costs drive the adoption of emerging technologies like direct air capture and some industrial decarbonization. As these technologies improve, there is greater opportunity for cost reductions, resulting in greater emissions reductions at comparable costs in later years.

Our modeling approach did not assume fixed electrification rates for applications like heat pumps and EV adoption; instead, these choices were optimized within the model. We found that electrification plays a major role in decarbonization, and at a carbon price of $400 per ton — sufficient to achieve net-zero emissions by 2050 — U.S. electricity generation more than doubles compared to today (see Figure 3). This shift is largely driven by transportation and industry moving away from direct fossil fuel use, while in the baseline scenario without a carbon price, electricity demand grows much more slowly. Solar and wind generation expand at all abatement costs, but their growth is dramatic when net-zero emissions is achieved. Without a carbon price, natural gas generation actually increases over time, whereas higher emissions costs drive greater reliance on biomass as a dispatchable alternative. The sharp rise in electricity demand underscores the importance of modeling end-use technology adoption and system-wide interactions.

Decarbonizing our energy system reinforces the idea that “you can’t just do one thing” in two important ways. First, no single mitigation strategy is enough — we need a diverse portfolio of solutions, from renewables and electrification to carbon management and efficiency improvements, to achieve deep emissions cuts. Second, every action we take influences the broader system, triggering responses that shape costs, energy demand, and technology adoption. Ignoring these interdependencies risks overlooking unintended consequences or missing more effective pathways. A holistic, systems-based approach is essential to crafting strategies that are both impactful and resilient.

This work was recently published in Environmental Research: Energy as an open-access article with much more detailed results.

The research was conducted by Jeremiah X. Johnson, Michael Blackhurst, Aranya Venkatesh, Aditya Sinha, Katherine Jordan, Nicholas Z Muller, Cameron Wade, and Paulina Jaramillo.

In 1981, the Natural Resources Defense Council challenged a decision by the U.S. Environmental Protection Agency, led at the time by Anne Gorsuch Burford, to change its interpretation of the word “source” under the Clean Air Act. Though the lower courts decided against the Agency’s action, the decision was appealed by Chevron Corporation to the Supreme Court, which agreed to take up the case.

The resulting decision, Chevron U.S.A., Inc. v. Natural Resources Defense Council, Inc. (1984; 468 U.S. 837), was to set precedent for decades of regulatory and judicial decision-making. Siding with the EPA and reversing the lower court, the decision in Chevron held that agencies were entitled to a certain deference when making interpretations of otherwise-ambiguous statutory language. The two-part test that emerged would serve to guide judicial review of agency decisions for nearly four decades. What would come to be known as the Chevron doctrine required judicial review to first determine whether the statute in question was ambiguous in its direction. If not, the agency must implement the law as written, full stop, end of story. In situations where the authorizing statute was ambiguous, however, Chevron held that courts should defer to an agency’s interpretation of how to implement that law (subject of course to limits of reasonableness of that interpretation).

Ok, but what does that look like in practice? Well, in 2015, for example, the U.S. EPA argued in the final rule promulgating President Obama’s Clean Power Plan that “[i]n the absence of specific direction or enumerated criteria in the [Clean Air Act] concerning what pollutants from a given source category should be the subject of standards, it is appropriate for the EPA to exercise its authority to adopt a reasonable interpretation of this provision” (80 FR 64530; October 23, 2015). Specifically citing Chevron, the U.S. EPA concluded it had a “rational basis” for regulating CO2 emissions from fossil fuel power plants.

Fast forward a few more years. With Associate Justice Neil Gorsuch — Anne Gorsuch’s son — siding with a 6-3 majority, the Supreme Court held in Loper Bright Enterprises v. Raimondo that the central premise of the Chevron case was decided in error. Citing the Administrative Procedure Act, the Supreme Court in Loper Bright returned authority for determination of the appropriateness of an agency’s action squarely to the courts. No longer were agency interpretations to be given such strong deference in judicial review.

That’s my non-lawyer’s version of the legal history, but what does any of this have to do with energy in the here-and-now? Again, quite a lot. Looking back in time, Chevron stood for forty years, in which time a great deal of regulatory decision-making has occurred. These are decisions that were made (and occasionally later challenged in court and upheld) assuming strong deference to agency interpretations. A quick search of the Chevron citation in the Federal Register — the official record of U.S. regulatory activity — shows that 318 separate final actions citing the case have been taken since 1994 (as far back as we can go with a text search). These 318 actions are scattered across 18 separate agencies, but rules issued by the EPA make up nearly one-third. A majority of those issued by the EPA pertain to energy in some fashion, be it air pollution, vehicle fuel economy, or renewable fuels. Though the majority opinion in Loper Bright took pains to note that previous decisions relying on Chevron were not automatically overturned by simple virtue of the decision, the possibility exists that new challenges to past agency actions could soon follow.

But the net effect of the Loper Bright decision may not be as dramatic as it might first appear. As I teach in my environmental policy seminar, U.S. legislation has become increasingly specific in recent years, reducing the scope for agency interpretation. This trend is itself partly due to concerns that laws won’t be implemented pursuant to Congress’s wishes unless they are clear in intent and direction. For example, specific considerations for agencies to undertake when developing implementing regulations and the presence of so-called hammer clauses forcing decisions to be made under particular timelines or conditions have all become more common in recent decades. It’s not a perfect apples-to-apples comparison, but just compare 2022’s Inflation Reduction Act to 1973’s Endangered Species Act. Sure, the 274-page IRA has been hailed as the single most impactful climate law to date, but the ESA holds tremendous influence over a vast swath of private and federal actions and does so in just 41 pages of text, leaving a great deal open to interpretation.

Of course, that’s not to say that there won’t be challenges to programs. After all, much of the recent fights over climate and energy have taken place in the context of EPA’s interpretation and implementation of the Clean Air Act, a law that has been revised and updated over the years, but one that still retains a decades-old legislative architecture. At least under the current administration, less reliance has been placed on Chevron to defend agency determinations — possibly in recognition of the potential for it to be eventually overturned — something that is again visible in EPA’s flat-lined citation of Chevron in the figure above.

Going forward, the decision in Loper Bright will surely continue the trend toward increasing specificity in legislation. This increasing specificity will require analysis upfront to ensure, to the maximum extent possible, that specifically-worded legislative provisions will have the intended effect. Particularly given political polarization and the difficulty of passing major legislation of any sort, options to tinker with or fix provisions are likely to be few and far between, increasing the need to understand the effects of a law before it is ever even voted on. That’s where the modeling community comes in. Indeed, we saw this in the case of the IRA, a law that was seemingly developed and evaluated in real-time. Unfortunately, an increasing trend toward specificity could also tie the hands of future administrations and limit their ability to adapt to new, previously unforeseen situations, calling for an even wider array of approaches and perspectives to be deployed at the time of legislation development.

Christopher Galik is a Professor in the School of Public and International Affairs at NC State University.

In an effort to keep Temoa central to the Open Energy Outlook (OEO) Initiative’s success, the OEO team initiated a significant code modernization project to produce Temoa version 3.0. In late 2023, OEO engaged with Western Spark, a Veteran Owned Small Business (VOSB), to restructure some of the core model components, modernize the codebase, and increase model performance. Several of the core objectives of the modernization effort include:

• Up-to-date dependencies: Bring all of the core language and external dependencies up to date. In the course of active research, there is seldom time to make the numerous adjustments needed to pace the evolution of software packages key to model maintenance and performance. This effort brings all of the supporting Python and external library dependencies up to date for increased performance, project security, and to help onboard newer users.
• Development best practices: Introduce several best practices to the development cycle. The Temoa codebase had reached a size that necessitated a better, dedicated approach to testing code and event logging needed to ensure performance and facilitate future development. This update cycle has introduced diverse code-testing elements and many modernizations in the codebase to make the model more performant and approachable by developers interested in modifying the core model.
• Refined execution modes: Refine the execution of different operating modes within the model. Over the course of its use, there have been several efforts to build novel extensions to the codebase to handle differing types of uncertainty analysis or explore alternative solutions. Those approaches were somewhat fragile extensions to the main body of the project. This restructuring effort commonizes much of the execution path and refines some of the internal elements of modules, wholly refactors others, and is intended to provide more rapid and accurate answers to questions on uncertainty.

So, where are we? We have passed the halfway point and are finishing refinements and testing of Temoa Version 3. This delivery should be more approachable to new users, performant on large models, and more familiar to developers who wish to dive in on model modifications to answer new questions on evolving perspectives of future energy policies and options. Testing on the new codebase shows it to be faster by a factor of 10x in many cases for model creation. Additionally, a rich log output details model execution and highlights possible data errors to guide users in the development of more accurate models. An energy network evaluation and visualization tool has also been incorporated to help visualize network complexity and possible data errors.

We’re excited to support both experienced researchers working on complex research tasks and newcomers looking to explore the world of energy system modeling and Temoa’s capabilities. Of course, version 3 maintains OEO’s commitment to open-source data and analytic tools that are fundamental to quality, reproducible, and defensible research products. The fully updated project and database should be available to all researchers in the spring of 2024.

Jeff Hyink is the owner and principal of Western Spark LLC and is grateful for the opportunity to partner with the OEO team to refresh and expand the Temoa model. He can be reached at info@westernspark.us.

A typical optimization-based energy system model has three components:

• Objective function: The objective function describes the overarching goal to be achieved. Energy systems models typically track and minimize the total system cost.
• Constraints: Constraints can encode real-world limits and/or ensure the integrity and validity of the model. For example, constraints in energy systems models govern realistic mass/energy flows and operational constraints of power plants.
• Decision variables: Decision variables represent the levers that the model can manipulate to meet the objective function within the constraints. Examples may include decisions regarding the deployment of technology capacities.

However, unlike traditional optimization approaches that yield a single solution, MGA provides a portfolio of solutions. MGA is an optimization technique that was initially developed to address land and water management challenges and produce sets of planning alternatives. Consider Figure 1, illustrating the decision space of a typical energy system optimization problem. The solid blue lines delineate the constraints within which a solution must lie. If this problem were to be solved in a conventional linear optimization fashion, the red dot would represent this solution, obtained by minimizing the total system cost, while still respecting the constraints of the problem. By relaxing absolute adherence to minimizing total system cost, MGA can identify additional solutions that are near but not perfectly optimal, as shown by the green dots in Figure 1. The additional solutions MGA can identify are determined by: 1) the “slack” variable, which describes the degree to which solutions can deviate from perfect optimality and 2) the effectiveness with which the new decision space is explored. In the context of energy systems, MGA can identify many energy system designs that have very similar costs, but may be otherwise substantially different. Some of these near-optimal solutions may possess features deemed valuable enough to justify the associated cost premium or fall acceptably within the bounds of other sources of uncertainty (e.g., future conversion efficiencies).

A primary advantage of MGA is in its ability to identify diverse energy system configurations. For example, Figure 2 compares the absolute least-cost solution to two other pathways identified using MGA. The most significant differences between the MGA strategies relate to emissions from electricity, direct air capture (DAC), and carbon capture and sequestration (CCS). One MGA strategy reflects fewer emissions from electricity and more use of DAC. The second MGA strategy reflects more electricity use and CCS. We identified these alternatives as representative groups of many MGA results (using k-means clustering). These pathways differ drastically in system configurations but are otherwise near cost-optimal. Providing a portfolio of options can help decision-makers balance many competing criteria (such as equity or public acceptability) at negligible to modest cost increases. However, identifying representative groups using MGA requires thorough exploration of the near-optimal space to avoid biased results.

Recognizing that determining the true real-world optimum is practically unattainable due to inevitable uncertainties, MGA offers a pragmatic alternative. Rather than fixating on a single techno-economic optimal solution, which is unlikely to mirror the real-world scenario, MGA advocates for the comprehensive study of the entire near-optimal solution space. Each solution within this space represents a viable real-world alternative, providing decision-makers with a diverse array of options when strategizing for long-term energy systems.

The IRA establishes subsidies for “clean” hydrogen under the designation “45V,” though defining “clean” remains contentious. The Treasury Department released proposed regulations in December 2023 addressing this definitional challenge. Analysis reveals three foundational criteria called the “Three Pillars” governing cleanliness designation: incrementality, time-matching, and deliverability. Incrementality mandates that hydrogen production electricity derive from clean sources that would otherwise lack development incentive. Time-matching requires electrolyzer operations align with incremental clean electricity generation annually or hourly. Deliverability stipulates electricity sourcing within the same geographic region.

Previous studies examining Three Pillars requirements employed technologically sophisticated models with significant limitations. Their temporal scope typically covered only one or two years, geographic focus remained narrow, and analysis concentrated exclusively on electrical power system emissions while neglecting broader energy system implications and fossil fuel displacement across sectors. Such narrow framing resembles calculating solar panel manufacturing emissions without accounting for displaced generation benefits.

Researchers deployed the Tools for Energy Model Optimization and Analysis (Temoa) energy system optimization model to assess Three Pillars implications on comprehensive greenhouse gas emissions. Using an Open Energy Outlook Initiative energy system database, the team analyzed five scenarios between 2025 and 2039: one baseline scenario without 45V tax credits and four progressively stringent regulatory scenarios. The least stringent scenario permitted all electrolytic hydrogen to qualify for full tax credits regardless of electricity source. The most stringent limited full credits to hydrogen meeting complete Three Pillars requirements.

Key findings indicate that 45V tax incentives encourage early investment in hydrogen energy infrastructure, but development constraints result in comparable hydrogen production across all 45V scenarios analyzed. Initial period analysis (2025–2029) reveals hydrogen production modestly increased power sector annual CO2 emissions relative to non-credit scenarios. The second period (2030–2034) demonstrated incrementality requirements drove slight power sector emission reductions compared to baseline scenarios. By the third period (2035–2039), power sector annual emissions decreased against baseline scenarios even with less stringent electricity sourcing requirements. Notably, no noticeable difference in emissions from the power sector between the annual and hourly matching scenarios emerged.

Comprehensive energy system analysis revealed small increases in annual cross-sectoral greenhouse gas emissions associated with the 45V tax credits, regardless of their stringency. Incrementality requirements produced lower incremental system emissions than laxer electricity sourcing alternatives, while hourly matching provided negligible benefits. Cumulative greenhouse gas emissions differences across scenarios between 2025 and 2039 represented less than one percent, underscoring how system boundary selection critically influences consequential emissions factor calculations.

The modest system-wide emissions increase stems from sectoral energy demand shifts. The 45V tax credits incentivized hydrogen production induced natural gas and electricity demand changes, particularly affecting industrial sectors. Increased electricity demand for electrolytic hydrogen production corresponded with decreased industrial electricity demand and increased natural gas consumption. Hydrogen simultaneously found emerging applications in Fischer-Tropsch liquid production and heavy-duty fuel cell transportation.

Analysis findings emphasize the complexity of environmental impact assessment for policies like 45V tax credits. While Three Pillars direct effects on hydrogen production and power sector emissions appeared modestly favorable, broader implications regarding fuel demand shifts and cross-sectoral emissions warranted deeper investigation. The stringency level of 45V tax credits demonstrated minimal effect on annual system-wide emissions across the study period, with negligible differences likely falling within energy system model uncertainty margins. This underscores the importance of comprehensive energy system evaluation when assessing policy interventions, ensuring clean hydrogen incentives genuinely support greenhouse gas reduction and sustainable energy transition objectives.

Achieving net-zero energy system emissions by 2050 demands substantial hydrogen technology deployment. While 45V tax credits incentivize crucial early hydrogen production deployments, Three Pillars complexity creates administrative burdens for regulatory agencies and regulated entities. Such complexity risks resource misallocation, increased regulatory costs, and heightened litigation exposure, potentially discouraging hydrogen infrastructure investments through uncertainty and compliance expense elevation. Investment delays responding to Three Pillars requirements could intensify the “chicken and egg” dilemma where hydrogen consumers hesitate due to supply uncertainty while suppliers refrain due to insufficient demand confirmation. This dynamic threatens 45V policy effectiveness, delaying essential hydrogen infrastructure deployment and impeding net-zero emissions progress.

Energy system models serve as invaluable scenario comparison tools utilizing consistent assumptions, offering policy impact insights on hydrogen demand, electricity consumption, and emissions across entire systems. However, these models inherently cannot capture all real-world dynamics influencing results. Firm behavior and unforeseen market developments frequently exceed model scope. This limitation highlights result unpredictability from such analyses, illustrating causality and counterfactual application challenges within modeling frameworks.

The proposed 45V rulemaking represents a critical juncture in United States energy policy, reflecting clean energy commitment and integrated climate mitigation approaches. As the Treasury Department finalizes rules, balancing regulatory complexity against clean hydrogen production investment encouragement becomes essential for meeting Paris Agreement objectives. The evolving energy sector requires policy flexibility addressing unexpected outcomes and continuous comprehensive analysis incorporating technical and behavioral dimensions.

The researchers invite review of their formal comments submitted to the Treasury Department regarding the 45V hydrogen tax credit proposed rulemaking, which includes detailed figures summarizing analysis results and expanded finding discussions.

The debate over the environmental impact of using natural gas versus coal for electricity generation hinges on various factors, particularly when assessing their respective climate footprints. On average, using natural gas is likely to have a lower climate impact than coal. Only if we assume high methane leakage rates and a 20-year global warming potential is natural gas worse than coal, and such assumptions are likely unrealistic. Liquefying, transporting, and regasifying natural gas would increase its carbon footprint, as highlighted in a paper I authored in 2007. However, a more recent study published in 2022 provided insights into the broader picture. The authors found that increasing international access to US LNG would induce increased demand for natural gas but would also enable fuel switching from coal. The study estimated that exporting 2.1 billion cubic feet of natural gas per day in the form of LNG could result in a small reduction or a small increase in global greenhouse gas emissions. On average, the authors estimated a reduction of 8 megatonnes of CO2e per year, a figure that, while noteworthy, pales in comparison to the 36,800 megatonnes of CO2 emitted globally in 2022 and to the magnitude of projected annual reductions needed to be consistent with global climate change mitigation goals. Similarly, another 2022 paper suggests that any benefits from switching from coal to LNG for power generation in other countries would start decreasing after 2030 as a result of increasing emissions from LNG not used to substitute coal.

Within the context described above, the arguments that natural gas is “better than coal” are insufficient. The global commitment to the 1.5-degree climate mitigation target outlined in the Paris Agreement necessitates a substantial transformation of the global energy system. The continued use of coal is incongruent with the goal of reaching net-zero emissions, but so is a massive expansion of natural gas consumption. Thus, the questions we should be asking are: what role does natural gas play in a net-zero system, how can investments made today support that role and avoid stranded assets, and how do LNG exports from the U.S. fit with the goals of the Paris Agreement?

Modeling for the United States indicates that natural gas will likely persist in the energy mix even in a net-zero system. Similarly, in global mitigation scenarios, natural gas remains a factor in the equation. The extent of its role will depend on the availability and advancement of carbon dioxide removal technologies. However, a substantial expansion of global natural gas consumption beyond current levels conflicts with the 1.5-degree — and for that matter, the 2-degree — target. In other words, a bullish path for natural gas is, according to a large body of modeling studies, fundamentally at odds with reducing the risk of catastrophic climate change. Within this context, the delay in new LNG terminal approvals appears to align with the broader climate mitigation goals.

Some assert that Europe requires new natural gas sources to replace Russian imports, while China persists in developing coal infrastructure. Nevertheless, embracing more climate-friendly alternatives, such as electrification for space heating and industrial processes, coupled with an intensified focus on renewables and nuclear power, offers viable options to diminish dependence on Russian gas and Chinese coal. The argument that if we don’t export our natural gas, someone else will or China will keep building coal also lacks compelling force. This “if we don’t do it, someone else will” reasoning has been employed for thirty years to stall climate action. As we advance into the third decade of the 21st century, the U.S. bears a moral imperative to lead by example rather than relying on this well-worn justification for expanding its fossil fuel industry.

This is not to suggest that U.S. LNG exports are entirely superfluous or will sharply decline. The U.S. already holds the position of the largest LNG exporter. Existing LNG export terminals are set to process over 13 billion cubic feet of natural gas per day in 2024, and capacity could reach 24 billion cubic feet of natural gas per day by 2027, when already-approved terminals will be completed. Consequently, even without new LNG terminal approvals, the U.S. is likely to remain a significant LNG exporter. Yet, additional approvals would likely be inconsistent with global emissions pathways that would limit climate change to 1.5 degrees Celsius above pre-industrial levels.

When discussing this post with a colleague, he questioned the need for additional research on the life cycle climate impacts of LNG. From the perspective of the moral imperative of limiting fossil fuel expansion to levels that are consistent with the Paris Agreement, I contend that such additional research is likely irrelevant because we already know that natural gas has higher life cycle greenhouse gas emissions than other low-carbon energy sources. However, the Biden administration requires quantitative analysis to support potential decisions on new LNG terminal approvals, where additional scrutiny holds value. After a decade of evaluating methane emissions from U.S. natural gas production, researchers have identified substantial variability in emissions from natural gas-producing basins in the U.S. This variability could affect the marginal impact of each cubic foot of LNG used to displace coal, potentially reducing the climate benefits touted by proponents of new LNG terminals. Unfortunately, current global energy system modeling tools lack the capacity to consider the heterogeneity in life cycle emissions for all natural gas units. Such tools could include detailed supply curves for LNG coupled with estimates of respective life cycle greenhouse emissions. Thus, delaying approval decisions provides the government with the time needed to develop tools supporting a final determination.

In the context of my research, I am particularly interested in other critical questions about LNG that merit further analysis. With the U.S. set to be capable of exporting over 20 billion cubic feet of natural gas per day by 2027, understanding the “best use” of such LNG becomes crucial — is it for power generation (unlikely) or possibly for steel production? It’s also important to remember that any use of natural gas in line with the Paris Agreement will require carbon dioxide removal (CDR). However, technologies facilitating CDR, such as direct air capture and biomass energy with carbon capture and sequestration, are still in their early stages of development. More research and investments are crucial to ensure CDR technologies are available by the middle of the century. These issues inform my work at the Open Energy Outlook Initiative.

I first started doing research on the climate impacts of natural gas and LNG infrastructure in 2005, a time when the prevailing narrative labeled natural gas as a ‘bridge fuel’ to a cleaner energy system. Almost two decades later, it has become increasingly evident that we must move beyond the notion of natural gas as a transitional solution, particularly in power generation. The metaphorical ‘bridge,’ originally designed to guide us toward a destination, has been our trajectory for over a decade. Thus, it is imperative to step away from this metaphorical bridge. While natural gas will persist in the global energy mix, it should no longer be considered a bridge fuel. Moreover, a substantial increase in global natural gas consumption beyond current levels is likely to hinder our progress towards achieving global net-zero emissions by the middle of the century — a compelling reason to accelerate towards the path of net zero, leaving the metaphorical bridge behind in the rearview mirror.

There are several U.S. regions, though, where there is an equal or greater peak demand that occurs at the other extreme — on that cold morning in the dead of winter. In an Energy Policy paper with Adi Keskar and Christopher Galik, I explored winter peaking trends and uncovered a few things that surprised me. First off, winter peaking is far more common than I expected. In 2018 — which was admittedly a banner year for winter power peaking — we found that one-third of our regions² had a higher winter peak than a summer peak. Many of these regions are up in the Northwest or down in the Southeast corners of the United States. Looking at the local ambient temperatures can tell you why this trend emerged. Many of these places have winters that are chilly enough to require some heating, but not so cold that electric heating is rare. Not too many Mainers or Michiganders heat their homes with electricity, but plenty of North Carolinians and Oregonians do.

And we are seeing upward revisions to our expected peak demand play out right now. PJM’s latest peak load forecast, released in January 2024, sharply increases the peak demand forecast and shows markedly higher growth rates for the winter peaks.

A lot of power system planning focuses on meeting peak demand. How do we ensure that we have sufficient generation available when our electricity demand is at its highest? Investing in the infrastructure necessary to meet a growing peak electricity demand can be costly and drive up rates for customers. Building new generation capacity and expanding transmission takes time and money. The design of a power system that reliably meets a winter peak looks a bit different than a summer peaking system. Solar generation may be anemic or absent on that cold February morning and batteries’ performance can be impacted. And — as we saw unfold with Texas’s winter storm Uri — our natural gas infrastructure can freeze up, leading to cascading problems.

All of this, too, is playing out against a changing landscape of technologies. The growing success of electric heat pumps for space conditioning — while quite efficient — will further add demand on our grid during the coldest hours of the year. It remains to be seen, though, if this growth can outpace increasing air conditioner use, with both new adoption and record-setting climate-driven heatwaves. Electric vehicles are another wild card. Sales are robust — the key question is whether we can effectively incentivize charging outside of peak hours (in both summer and winter). I, for one, would be happy to avoid charging my car at these times, but I might need a reminder, or better yet, have this occur seamlessly in the background. As someone who thinks a lot about the power system and decarbonization, I still would appreciate it to require as little effort as possible.

Planning for a potential winter peak is a solvable problem, though. Some capacity markets are already bifurcated into summer and winter markets. Utilities can and should incorporate robust winter modeling into their integrated resource planning processes. It is essential, however, that these plans do not simply penalize renewable resources during winter peaks, but also recognize the higher failure rates of fossil infrastructure during extreme cold. Planning for winter peaking is not mutually exclusive with aggressive decarbonization efforts.

More information is needed on the potential of demand response to meet winter peaks. While summer demand response programs are more mature, we are still learning about the true potential for demand response solutions to a winter peak. An ACEEE white paper offers an optimistic view of the DR potential, finding that “demand response via electric heating, strategic electric vehicle charging, and controllable water heaters could provide gigawatts of capacity” to the Texas system. For this to be successful, though, people would need to give up a bit of control of their heating and hot water during those cold mornings. It may be tough to get volunteers. Broad-based efficiency measures — like good old insulation — can provide major reductions in winter electricity use, but historically many households have declined to install insulation, even when it may be cost-effective.

This is a lot for energy system modelers to consider, but getting it right is important. New technologies with different electricity demand profiles. A changing climate driving different energy service demands. Extreme weather. Unpredictable (dare I say, sub-optimal) human behavior. I’m heartened by the work done in this space, however, and confident that we can offer meaningful insights as we navigate our path to a decarbonized energy system.

In our recent publication, “Closing the Gap: Achieving U.S. Climate Goals Beyond the Inflation Reduction Act,” we identified different policy strategies that, when combined with the IRA, help the U.S. meet its long-term emission goals. First, we introduce a carbon price in addition to the IRA. We model two different carbon price schedules: a lower price schedule starting at $51/ton in 2025 rising to $78/ton by 2050, and a higher price schedule starting at $155/ton in 2025 rising to $240/ton by 2050. Whereas the IRA subsidizes the adoption of decarbonizing technologies, a carbon price disincentivizes greenhouse gas emissions. Neoclassical theory suggests a carbon price is the most efficient way to reduce emissions because it creates a market for greenhouse gases. In theory, a carbon price spurs innovation by allowing consumers and producers flexibility in how to manage their emissions.

However, carbon pricing has historically been politically intractable. Therefore, we also model so-called “command and control” strategies to reduce emissions in addition to those expected from the IRA. Relative to carbon prices, economists often refer to command and control strategies as “second best” alternatives because they are considered less economically efficient. The U.S. already uses some command and control strategies to manage emissions, such as mandating minimum fuel economy standards for vehicles and appliances. Our model assumes additional emissions reductions come from both new command and control strategies and making select existing standards more aggressive.

To quote our paper directly:

“Under the Low Carbon Tax scenarios, GHG emissions decline rapidly in the first ten years resulting in a 52% reduction in 2030 GHG emissions compared to 2005 levels. However, GHG emissions plateau after 2030 in this scenario. The low value of the carbon tax is insufficient to spur reductions in sectors that are more expensive to abate. For example, GHG emissions from the industrial sector do not decrease substantially with a low carbon tax. In contrast, the Standards and High Carbon Tax scenarios result in GHG emissions reductions economy-wide, including in the industrial sector, across the entire analysis period. By 2030, GHG emissions in the Standards and High Carbon Tax scenarios are 56% and 58% lower than in 2005, respectively. Furthermore, by 2050, GHG emissions in the Standards and High Carbon Tax scenarios are 74% and 76% lower than in 2005.”

While these scenarios do meet our NDC, they do not reach net-zero GHG emissions. The net-zero target is significant, as it keeps the U.S. in line with the 1.5°C target set in the Paris Agreement. While our model includes a characterization of negative emissions technologies like bioenergy with carbon capture and sequestration and direct air capture of CO2, the carbon tax isn’t high enough to spur their deployment.

We also compare each decarbonization strategy using the ratio of benefits to costs. All else equal, strategies with a higher ratio of benefits to cost are preferred. Our analysis includes the capital, operating, fuel, and maintenance costs of technologies. We monetize the benefits based upon our assumed social cost of carbon plus the benefits of reducing criteria air pollutant emissions (NOx, SO2, and PM2.5). All of our modeled policies have a benefit-cost ratio greater than one; the ratios range from 6–24. We also calculate the average abatement cost, or the total incremental cost divided by the total avoided emissions. The average abatement costs are all under $20/ton CO2, ranging from $5/ton to $19/ton CO2.

A key finding from our publication is that the combination of command and control policies across the energy economy can lead to emissions reductions approximately equal to those obtained under a $200/ton carbon tax, meeting (and exceeding) our NDC. These two policies both reduce emissions ~75% by 2050 relative to 2005. While not quite net-zero, policymakers could build upon the fuel and technology standards modeled here to reach net-zero emissions by 2050.

Our results demonstrate two alternative strategies supplemental to the IRA that could meet the U.S. decarbonization goals: a carbon price and more aggressive command and control policies. A relatively modest increasing carbon price schedule, starting at $51/ton in 2025 and rising to $78/ton by 2050, reduced 2005 emissions by 55% in 2050, which surpasses the U.S.’ nationally determined contribution. On average, this modest carbon price produces about $24 of benefits for each $1 of costs. A tripling of the assumed carbon price compares similarly to more aggressive command and control strategies, both reducing emissions by about 75% at similar benefits and costs. However, the benefit-cost ratio declines with increasing emission reductions. While a carbon price is theoretically most economically efficient, our results suggest policymakers could pursue command and control strategies towards similar ends. A command and control strategy has historically been more politically tractable for challenging environmental problems, as exemplified in the Clean Air Act, the Clean Water Act, federal Corporate Average Fuel Economy (CAFE) standards, and California’s Zero Emission Vehicle standards. This work proves the efficacy of a suite of politically feasible technology standards to decrease energy system emissions.

Nevertheless, addressing climate change requires transforming our energy systems. How can we effectively plan these changes given the complexity inherent in energy systems? What technologies and policies decarbonize our energy systems? At what cost? How can we de-risk these decisions? What are the tradeoffs? Answering these questions requires state-of-the-art decision support resources and thoughtful analyses.

Our team uses a method called Tools for Energy Model Optimization and Analysis (Temoa) to study energy systems. Temoa identifies the least-cost way to build and operate an energy system. Consider the example of constructing a house. You calculate the energy needed for lighting, appliances, water heating, space conditioning, and any other plug-in devices. Then, you gather data on prices for various models and energy sources like natural gas, solar panels, battery storage, and electricity from your utility. Optimization could determine the combination of energy supply, storage, and end-use technologies that fulfills your requirements at the lowest cost. Keep in mind, you might decide on a specific technology, say a certain appliance, that is not the absolute cheapest. However, the least-cost set of options gives you valuable insights into making an informed decision.

Our model, the Open Energy Outlook (OEO), similarly uses Temoa to analyze the entire U.S. energy system. Drawing from a rich database describing thousands of energy-system assets, OEO identifies the least-cost portfolio of energy sources and technologies that meet our expected demands for energy from 2020 to 2050 for 9 regions in the U.S. OEO models interdependencies among technologies and different energy sources; seasonal, daily, and hourly fluctuations in supply and demand; the ramping up and down of thermoelectric power generators, and tracks the flow of greenhouse gases from energy sources into potential sinks, such as biomass or negative emission technologies. A key distinguishing feature of Temoa is that it considers different end-use technologies in identifying least-cost energy provisions, whereas other energy system models specify end-use technologies as inputs. OEO also tracks and reports greenhouse gas emissions as well as select criteria air pollutants. For a more comprehensive list of OEO capabilities, refer to the documentation for Temoa.

Most importantly, Temoa and OEO are both open-source. The OEO team believes that transparency, accessibility, and replicability are essential to achieving our decarbonization goals. As opposed to proprietary models and data, open-source resources better democratize understanding of the challenges and opportunities and facilitate consensus-driven strategies. An academic review of 31 “mostly” open source models found Temoa to be amongst the best.

OEO proves especially valuable in examining the dynamic interplay between policy and energy systems. OEO can replicate the impact of a monetary incentive, like a subsidy or tax break, by adjusting the assumed prices of relevant technologies. It could identify the least-cost routes to meet emission reduction targets or, conversely, to maximize emission reductions within a specified cost range. Additionally, OEO can assess the influence of technology learning rates on both the energy system and emissions. These examples highlight just a few of the diverse applications where OEO can provide valuable insights.

Over the last year, our team has been using OEO to estimate how the Inflation Reduction Act will reduce emissions (and how to achieve net zero), identify lots of net-zero emissions strategies that are near least cost, build emission mitigation curves, estimate how hydrogen policy will impact emissions, and identify reduction opportunities by state. We also published our first Open Energy Outlook in September 2022.

We would not have had these opportunities without contributions from our core team, collaborators, and sponsors. We thank the Sloan Foundation for their continued financial support, and our many contributors to OEO and Temoa.

First, we’re focusing on ways to decarbonize the whole energy system. To perform analysis across the full energy system, we employ Tools for Energy Model Optimization and Analysis Temoa, an open-source energy system optimization model. The model employs linear optimization to make technology-specific investment and utilization decisions that minimize the cost of energy supply subject to a set of constraints. The OEO input database for Temoa includes a representation of fuel supply, electricity, residential, commercial, industrial, and transportation sectors. This approach allows us to consider sector-specific technology alternatives, such as electric heat pumps or LED lighting in the residential sector, while also capturing cross-cutting pathways, such as power-to-X in which water is electrolyzed to produce a variety of synthetic fuels that can be utilized across the end-use sectors.

Second, our effort uses open-source modeling tools and data. We make use of open-source software elements wherever possible, which helps to minimize the barriers to entry for new energy modelers. Temoa is publicly available on GitHub, and the OEO input database used to perform the model runs is also publicly available in a dedicated OEO repository on GitHub. The project also leverages other open source efforts, including electric sector data made available through the Public Utilities Data Liberation Project PUDL, and PowerGenome, which is used to programmatically create the electric sector input data for the modeling effort.

Third, we strive for transparency. We’re working hard to make the model and input data as transparent to external stakeholders as possible. The Temoa source code is well-documented, and the user manual includes a detailed explanation of the model equations and functionality. In addition, we are using jupyter notebooks to document the OEO input data. Using jupyter notebooks allows us to embed database queries directly into the documentation, enabling users to interact with the data by selecting fields and producing dynamically rendered tables and graphs that remain up-to-date with the latest version of the database.

Fourth, we’re building a community around the effort. We have a large team of researchers who are actively engaged in providing assistance on the project. In addition, by making our research effort open source, the broader community can provide feedback on our code and data, and use it in their own analysis. Our effort is an intentional shift towards a more distributed, collaborative modeling team, allowing access to a much wider array of disciplinary and domain expertise to inform the analysis. We have articulated our approach in a paper published in 2020.