Power sector impacts of the Inflation Reduction Act of 2022

Although IRA contains a range of provisions for the power sector (table S3), there are several key incentives that are expected to account for a large share of emissions reductions. The PTC and investment tax credit (ITC) are the most significant provisions. IRA expands tax credit eligibility to give new wind, solar, and several other generating technologies the flexibility to take the PTC or ITC, depending on which is more lucrative for a given project, and offers the ITC to standalone energy storage for the first time. Historically, solar projects could only take the ITC, and the frequent expiration or last-minute extension of both credits over time (figure S2) contributed to boom-and-bust cycles of investment. Under IRA, the PTC and ITC become technology-neutral by 2025, and all zero-emitting resources can receive these tax credits at full value for projects that commence construction before the end of the year after power sector CO₂ emissions reach 25% of their 2022 levels or 2032, whichever is later. The credits then begin to phase out over three years. Eligible projects receiving the PTC are awarded per megawatt-hour of electricity output over the first ten years of production, while the ITC is awarded as a fraction of investment costs. In addition to the PTC and ITC for new low-emitting capacity, IRA offers a production credit for existing nuclear power, recognizing potential retirement risks for these power plants, which offers up to $15/MWh (depending on electricity revenues and support from other programs) and ends after 2032.

IRA's PTC and ITC contain bonus credits that can increase credit levels for eligible projects (figure S2). Labor standards—which add requirements on prevailing wages and apprenticeships—are required for full eligibility of the credits (about $27.5/MWh in 2022 dollars for the PTC and 30% for the ITC). Most models assume these standards are met by all projects, as failing to do so reduces the value of the credits by 80%, making these projects less competitive with ones that are compliant with labor standards. There are also bonuses that can provide a 10% increase for the PTC and 10 percentage point increase for the ITC for projects that are sited in specified 'energy communities' (e.g. communities with recent coal closures or fossil fuel employment) and/or use a minimum share of domestic content—with projects eligible to claim both bonuses. There are additional bonus credits for certain small-scale projects sited in low-income communities or on Native American land, though these bonuses have annual caps on their use. IRA also allows for direct pay for nonprofits and tax-exempt entities (enabling tax credits to be directly delivered in cash) and transferability (allowing tax credits to be transferred to other taxpaying entities in exchange for cash payment).

There are also several multi-sector credits that may impact the power sector. IRA offers an extension and enhancement of credits for captured CO₂ (45Q), which offer up to $85/t-CO₂ for captured and stored CO₂ for electricity, industry, and fuels and $60/t-CO₂ for utilization. Clean hydrogen production also receives a new PTC (45V) based on the lifecycle GHG emissions intensity of production with a 10 year eligibility for projects, which may be up to $3 per kilogram of hydrogen.

More detailed descriptions of IRA incentives are available in [3, 9, 15].

To examine the effects of IRA on the power sector, this analysis uses 11 models that vary in structure, inputs, and IRA representation (see SI S1 for more detail on participating models). Participating models can be divided into two categories—six energy-economy models that represent sectoral feedbacks (e.g. impacts of transportation choices on electricity consumption) and five power sector models that can bring greater spatial, temporal, and technological detail (see SI S1). The representation of the power sector in most models includes decisions about least-cost investments, retirements, and dispatch in generation, energy storage, and transmission resources. Electricity demand is endogenous in the six energy-economy models (table S1), while service demand (e.g. vehicle miles traveled) and non-energy prices are exogenous for most models in the study as are macroeconomic variables such as inflation, interest rates, and gross domestic product. Other papers in the literature examine IRA's impacts on macroeconomic [9] and distributional [16, 17] outcomes using alternate modeling approaches.

Two of the largest differences across models are choices about temporal and spatial resolution (SI S1), which can shape investment decisions, especially under deeper decarbonization [18]. Figure S1 compares temporal and spatial resolutions across participating models and illustrates how models with more time segments generally have fewer geographical regions (and vice versa), reflecting computational tradeoffs between these two dimensions. IRA's tax credits accelerate deployment of low-emitting technologies by lowering their costs, and although generation costs of wind and solar are lower than alternatives in some markets already (hence leading to some deployment even before IRA incentives), the market value of variable resources declines as deployment increases [19, 20], which can mean that other resources are deployed as part of a cost-minimizing system mix and that additional cost reductions for low-emitting options will increase deployment ¹⁹ . Models' abilities to capture value deflation depends in part on decisions about temporal and spatial resolution [21]. As described in [8], most participating models do not incorporate expansion constraints on the rate of wind and solar additions, though deployment constraints on transmission, nuclear, and carbon capture are more common.

Results of IRA scenarios are compared with counterfactual reference scenarios without IRA to assess potential impacts. The policies in the reference scenarios include but are not limited to national policies such as the Infrastructure Investment and Jobs Act (IIJA) as well as subnational renewable and clean electricity portfolio standards, carbon markets, incentives, and other state and regional policies, as discussed in supplementary note 2. The IRA incentives are then added to these assumptions to examine the impacts of IRA (ceteris paribus).

Models vary in their coverage of IRA provisions (table S3) and how these incentives are represented (see SI S2). While no model includes all IRA provisions, every model includes the main power sector provisions (table S3), which are expected to account for the majority of emissions reductions. Each model has different methods for calculating which technologies utilize which tax credits, what bonuses apply for which projects, when tax credits phase out, and the impact of direct pay and transferability, among other policy implementations. The major power sector tax credits—PTC, 45Q credits for captured CO₂, 45V credits for clean hydrogen, and existing nuclear credits—are adjusted for inflation over time in the bill text ²⁰ .

For the core IRA provisions, low and high sensitivities have been conducted to examine how uncertainties related to IRA implementation could affect model outcomes. These sensitivities vary assumptions about bonus credits, tax-equity haircuts, eligibility for end-use credits, and technology-specific deployment constraints (see SI S2, table S4 for specific scenario assumptions). Fuel costs, macroeconomic conditions, and technology costs are consistent across these runs.

Input assumptions about the costs of supply- and demand-side technologies vary across models. Two important assumptions for power sector outcomes are capital costs of generation resources, which are compared in figure S3, and natural gas price projections over time, which are shown in figure S4. Two popular public data sources are the Energy Information Administration's (EIA) Annual Energy Outlook (AEO) and National Renewable Energy Laboratory's (NREL) Annual Technology Baseline (ATB). These different datasets and cost assumptions are aggregated and downscaled to the relevant resolution of each model. Both EIA's AEO and NREL's ATB also include sources for demand-side technology costs and performance.

Figure S3 illustrates how input assumptions for generation and energy storage costs are similar across models, largely due to the widespread use of NREL's ATB for cost assumptions. Figure S4 suggests that assumptions about natural gas prices vary more across models. Although several models use EIA's AEO for fuel price assumptions, modeling teams use different near-term adjustments to account for recent observed price increases and alternate AEO scenarios for changes over time.

Cross-model differences reflect several sources of uncertainty:

• Some IRA provisions await governmental guidance and rulemakings regarding interpretation and implementation as of the time this analysis was conducted (e.g. 45V credits for hydrogen).
• Technology and fuel costs depend on external uncertainties (e.g. other policies, inflationary drivers, technology trends, domestic and international economies).
• Several implementation factors are challenging to model but may influence adoption, including barriers to siting and permitting; uncertain infrastructure buildout (e.g. transmission); interconnection queue delays; scaling supply chains, workforce, and materials; other grid system upgrades to maintain reliability with the retirement of existing capacity; and non-cost factors influencing adoption.
• There are unknowns about the impacts of electrification on demand growth, profiles, and flexibility, and some models do not capture these changes endogenously with IRA incentives (table S1).

Effects of IRA depend on responses to IRA provisions in terms of clean energy adoption and producer choices, changes in policy, technological progress, and others. Models attempt to represent key economic and policy factors that shape emissions and energy systems trajectories and are typically designed to find least-cost system pathways.

These uncertainties and model simplifications imply that results should be interpreted as conditional projections based on economic incentives and not as predictions of policy-induced outcomes.

IRA accelerates deployment of low-emitting resources—defined here as renewables, nuclear, and generation with carbon capture and storage (CCS)—and leads to fewer near-term nuclear retirements relative to the reference case without IRA (figure 1). This sustained deployment of low-emitting resources and energy storage, spurred by technology-neutral PTCs and ITCs and credits for CO₂ capture, is projected to be the most extensive in U.S. history. IRA increases deployment of low-emitting capacity, with 23–120 GW added annually (67 GW yr⁻¹ mean across models) through 2035 across models. This is compared to 13–61 GW yr⁻¹ in the reference without IRA (30 GW yr⁻¹ average), and the annual record of 65 GW added in 2002 for all power resources, which was primarily gas-fired capacity. These increases in low-emitting capacity deployment continue through the 2030s due to most models not crossing the 25% of 2022 CO₂ emissions threshold until after 2035 (as discussed in section 3.3), which means that eligible technologies could receive the PTC or ITC over a longer time horizon than other IRA tax credits that expire after 2032 ²¹ . Such unprecedented levels of deployment require substantial scale ups of supply chains, infrastructure, and workforce, which elements of IRA aim to address and some models represent through build rate constraints, mostly for emerging technologies [8].

Zoom In Zoom Out Reset image size

Although there are model-specific differences in the extent of changes, most models agree that wind and solar lead IRA-induced increases in adoption (figure 1). Average growth rates through 2035 for wind and solar capacity range from 10–99 GW yr⁻¹ (56 GW yr⁻¹ average). This is both higher than the reference (12–53 GW yr⁻¹ with 27 GW yr⁻¹ average) and the annual record of 32 GW (utility-scale and distributed) installed in 2021. This expansion leads to 1.4–6.2 times current installed wind and solar capacity by 2035.

Annual installations of energy storage also increase under IRA with 1–18 GW added annually and a 7 GW yr⁻¹ average (compared with 0.4–8 GW yr⁻¹ and 3 GW yr⁻¹ average in the reference). Since cumulative battery storage deployments in the U.S. are 7 GW as of 2022, IRA projections indicate exceeding 2022 installations annually (on average) over the next decade. Note that uncertainties in future cost and performance assumptions for energy storage and other technologies (figure S3) may increase the range of projected outcomes.

IRA amplifies deployment of relatively mature technologies, such as wind and solar, while also increasing the adoption of emerging technologies. IRA includes an extension and enhancement of 45Q credits, which offer up to $85/t-CO₂ for captured and stored CO₂ for electricity, industry, and fuels. The extent and composition of CCS deployment varies by model (figure S9) and ranges from 0–150 GW (37 GW average) by 2035 (figure S5). 4 of 11 models show <1 GW deployment of CCS in the power sector by 2035 (though each shows CCS deployed for non-power applications), 1 model shows more CO₂ captured from gas CCS compared to coal CCS, while the remaining 6 models expect more coal CCS. Coal CCS capacity is driven by greater subsidies from IRA per unit of output for these plants relative to gas, due to coal's higher emissions intensity before capture. Models with higher gas CCS deployment also tend to have lower capital cost assumptions for this technology (figure S3), indicating that capital cost assumptions (combined with fuel prices) influence the extent of deployment. CCS deployment is one of the biggest departures from the reference, where little capacity is deployed by 2035, though CCS is deployed in deeper decarbonization scenarios [22, 23]. Because commercial-scale CCS and pipeline capacity are relatively nascent, projected deployment depends heavily on technological assumptions, permitting and siting timelines, CO₂ storage and transportation costs, and project risk.

More existing nuclear capacity remains online with IRA through 2030, between 80–99 GW (compared with 55–100 GW in the reference), due to incentives in IRA. But after subsidies expire, the spread in model results widens significantly, ranging from 5–97 GW in 2035 (figure S5). The models with the greatest decline in nuclear generation also tend to have lower natural gas price assumptions in 2035 (figure S4). This comparison suggests that, when subsidies for existing nuclear expire, retirement risks increase, especially with lower natural gas prices and wholesale power prices. This finding aligns with the literature indicating that retirements of existing nuclear capacity are linked to natural gas prices, ongoing costs, and policy [24].

IRA lowers coal and natural gas generation while increasing low-emitting generation shares (figure 2). In 2035, unabated coal reductions from 2021 levels range from 44%–100% with IRA (84% average) versus 12%–63% in the reference (38% average). In addition to retirements, lower coal capacity factors drive this change—ranging from 7%–64% with IRA in 2035 (32% average) versus 38%–74% in the reference (57% average). Gas-fired generation shares in 2035 range from 7%–38% with IRA (21% average) versus 16%–47% in the reference (33% average), down in many scenarios from 39% in 2021. Retirements of natural gas combined cycle (NGCC) and natural gas turbine (NGGT) capacity are lower than for coal (figure 1). NGCC capacity (figure S5) declines less than its generation, as units can operate at lower capacity factors (figure S7), providing capacity in periods of system stress. NGGTs are less impacted by IRA provisions, because they are primarily used for peaking capacity rather than generation. Figure S11 illustrates the influence of gas price assumptions on generation shares, and although gas prices do not uniquely determine technology-specific deployment, these prices are generally positively correlated with wind, solar, nuclear, and coal shares and negatively correlated with gas generation shares.

Zoom In Zoom Out Reset image size

Generation shares from renewables, nuclear, and CCS-equipped generation in 2035 range from 59%–89% with IRA (77% average) versus 46%–74% reference (55% average), as shown in figure 2(B). Note that emissions reductions do not track exactly with low-emitting generation shares, since the relative share of coal and natural gas displaced by new generation also drives CO₂ changes.

Given regionally varied resources, the increased deployment of renewables contributes to additional transmission expansion. IRA model projections show transmission infrastructure growing at a rate of 0.4%–2.3% per year through 2035 versus 0.2%–0.8% in the reference, which is greater for many models than the recent annual rate of ∼1.2% but similar to the longer-term growth rate of 1.9% experienced from 1978–1999 [25, 26]. However, transmission expansion is greatly influenced by the underlying model structures (e.g. spatial resolution, endogenous expansion) and input assumptions (e.g. costs, constraints).

Electricity demand increases with IRA from incentives for end-use electrification, including subsidies for electric vehicles and heat pumps. Energy system models indicate an 8%–36% increase in load by 2035 (22% average) from 2021 levels (7%–31% in the reference, 19% average). This sustained demand growth is a notable departure from the roughly constant national demand for electricity experienced in the U.S. since the mid-2000s [27]. The combination of peak load growth from electrification and deployment of variable renewables leads to greater capacity deployment than generation growth, though energy efficiency may partially offset these increases.

Although hydrogen use for power generation is nearly zero (figure 2(A)), IRA incentives for hydrogen shift production toward lower-emitting technologies (figure S21) and expand non-electric hydrogen consumption. Models differ in the degree of production from electrolysis and CCS-equipped steam methane reforming as well in the extent of growth in end-use demand for hydrogen, which ranges from near-current levels to a doubling by 2030. Models with electrolytic hydrogen production have corresponding increases in electricity demand, though this growth is modest relative to other end-use demand (figure S19).

SI S3 provides additional comparisons illustrating the relative roles of input assumptions and model structure on the power sector mix across models. Many of these factors interact to determine technology-specific deployment, though assumptions about capital costs, discount rates, and gas prices as well as structural features such as a model's temporal resolution all influence modeled effects of IRA.

IRA accelerates emissions reductions in the power sector, which cuts emissions faster and deeper than other sectors. As shown in Bistline et al [8], the reduction in economy-wide GHG emissions from IRA by 2030 is on average 9 percentage points (ranging 3–15 p.p.), while power sector CO₂ emissions are reduced an average of 18 p.p. (ranging from 5–34 p.p.) from reference levels. The electric sector constitutes most of the economy-wide GHG mitigation from IRA with declining emissions in the power sector contributing 38%–80% of reductions relative to the reference (64% average) in 2030, and the reduced emissions intensity of electricity generation amplifies additional emissions reductions in transportation, buildings, and industry due to electrification (figure S19).

Models indicate consistent trends in power sector emissions reductions but with variation in the magnitude of reductions. By 2030, projected electricity CO₂ emissions are 47%–83% below 2005 levels with IRA (68% average) compared to 41%–60% (50% average) in the reference (figure 3(A)). By 2035, the range of reductions tightens to 66%–87% (78% average) given additional time for lower costs, retirements, and ratcheting state policies. Nevertheless, the power sector falls short of reaching the U.S. goal of 100% 'carbon pollution-free electricity' [10] by 2035 in all models. Additional actions at federal, state, and local levels can narrow the gap further, given how IRA lowers costs associated with adopting policies and may encourage technological progress.

Zoom In Zoom Out Reset image size

Under IRA, the technology-neutral production and ITCs for zero-emitting resources could continue after 2032 if power sector CO₂ emissions are above 25% of their 2022 levels. Only 3 of 11 models reach this threshold by 2035, and emissions in many models exceed these levels even in 2050 (figure S13). All three models that cross this threshold by 2035 eventually increase emissions and exceed this threshold in later periods. This timing has important implications for electric company planning, as these scenarios indicate that tax credits could be sustained well beyond 2032 to incentivize decarbonization. Nuclear retirements, load growth, and expiring CCS incentives play roles in potential emissions rebounds after 2032, which delay the point when the emissions threshold is crossed, while accelerated coal plant retirements and additional policies could mean that the threshold is reached sooner.

Moving away from carbon-intensive fuels also lowers conventional air pollutants, which can help improve air quality and lower health burdens on local communities [5, 17, 28]. Figure S14 illustrates declines in NO_x and SO₂ emissions over time, and IRA-induced reductions relative to the reference mirror similar trends in CO₂.

Although IRA's power sector credits lower long-run costs (including all capital and operating costs after accounting for tax credit values), costs in any given year may increase or decrease, especially after accounting for IRA's electrification incentives. IRA accelerates the shift to low-emitting resources, which generally entail higher capital costs while lowering fuel and operating costs [18, 23]. Models disagree about whether IRA would increase or decrease power sector costs relative to the reference (figure S16). Increases could come from greater electricity demand (from electrification) or from higher electricity prices (due to accelerated investments). The balance of these effects depends on which tax credits are used (e.g. investment credits lower upfront costs, while production credits lower operating costs over time), and the timing of capacity deployment. However, wholesale and residential retail electricity prices decline across nearly all models and time periods between the reference and IRA scenarios (figure S17), which suggests that greater demand is the main driver of increased electricity expenditures.

IRA could lead to lower total energy costs for households and businesses, as lower fossil fuel expenditures more than offset potential increases in electricity spending (figure S16). Although net costs decline for most economy-wide models and periods, there are increases in total electricity costs due to higher electricity demand from electrification. Models showing declining electricity costs in figure S16 are electricity-only partial equilibrium models that do not account for IRA-driven electrification (figure S19). Net costs vary by model and over time: Cost declines with IRA are $10–52B per year (about $73–370 per household) by 2035 relative to the reference. Households benefit directly from energy spending and indirectly from lower prices of goods and services throughout the economy, which are both included in this total. Ultimately, the distributional impacts of IRA depend not only on direct expenditures on fuels but also on end-use incentives in IRA, how revenue is raised/recycled, and other changes in the tax code through IRA. Such comparisons require computable general equilibrium models with detailed treatments of the tax code and assumptions about pass-through of tax credits, which are beyond the scope of this analysis but are discussed in a few recent studies [9, 16, 17].

Projected federal tax credit expenditures vary across models and over time in terms of their magnitudes and compositions (figure S18). Cumulative power sector tax credit values range from $81–410B through 2030 ($230B average). These cumulative tax expenditures increase to $240–960B through 2035 ($530B average) due to the extended eligibility of power sector credits. Annual expenditures range from $28–100B in 2035 ($56B average). Tax credit expenditures across provisions with uncapped incentives, including the power sector production- and investment-based credits as well as 45Q credits for stored CO₂, reflect the relative uptake across models. Note that the implied fiscal costs per unit of emissions reduced are higher in instances where there are considerable investments in subsidized resources under reference conditions even without IRA tax credits (figure 1).

To understand how alternate assumptions about IRA implementation and related assumptions could alter power sector outcomes, low and high IRA sensitivities were conducted used harmonized scenario assumptions (see table S4 in SI S2) in 7 of 11 models. These sensitivities vary assumptions about IRA bonus credits, tax-equity haircuts, eligibility for end-use credits, and implementation details while holding other assumptions constant (e.g. fuel and technology costs). These sensitivities highlight how these IRA assumptions can shift generation shares and levels of CO₂ reductions in the power sector (figure S22). In 2035, CO₂ reductions in the 'Mid' IRA case range from 66%–87% from 2005 levels, which broadens to 60%–92% under the 'Low' and 'High' sensitivities. Similarly, the 2035 low-emitting generation share ranges from 59%–89% in the 'Mid' IRA case, which broadens to 59%–95% under the 'Low' and 'High' sensitivities. Although these scenarios broaden the range of possible outcomes, they also suggest that variations in the generation mix across models can be more significant than differences across sensitivities within a single model, highlighting the value of model intercomparisons like this one.