According to SMC, for an energy strategy to conform to ZERO, the actual or perceived economic costs of energy must never become significantly higher than if unabated fossil fuel use were continued. Relatedly, if an energy strategy gives emitters economic incentives to move elsewhere where they can continue to emit, the strategy fails to conform to ZERO. And finally, in order to conform to ZERO, an energy strategy should not count on the import of emission-free energy from elsewhere unless there is a global abundance of exportable emission-free energy, an unrealistic scenario for the next few decades.
The greatest economic risks to energy strategies that aim for ZERO come from sectors/activities in which no candidate substitutes for fossil fuels have been deployed at scale and/or all are much more costly. According to Davis et al. (2018), these are the following: load-following electricity (12% of global emissions in 2014), iron and steel production (5%), cement production (4%), shipping (3%), aviation (2%), and long-distance road-transport (1%). Electricity production outside load-following electricity is an even bigger source of emissions (26%). Eliminating emissions from electricity at affordable costs is widely regarded as a key step in the decarbonization of other activities, including hard-to-decarbonize ones. The hope is that these will be electrified (e.g., short- and medium-distance road transport) or shifted to synthetic fuels, to be produced with emission-free electricity. For such synthetic fuels to have costs similar to those of fossil fuels, the emission-free electricity used to produce them will likely have to be cheaper than fossil fuel-based electricity.
Decarbonizing Electricity with Variable Renewables
Aspects of the economic costs of electricity sources are expressed in terms of their levelized cost of electricity (LCOE). This quantity specifies the required revenue from selling electricity to recover the capital investment of construction as well as maintenance, operating, and (where applicable) fuel costs. The most promising recent trends in the economics of emission-free electricity are cost reductions of solar and wind energy in the last few decades. The LCOE of both these sources have decreased rapidly due to learning effects in combination with economies of scale in the production of solar panels and wind turbines.
To be specific, according to the International Renewable Energy Agency (IRENA), median LCOE have fallen from 378 $/MWh (in 2010) to 39 $/MWh (in 2021) for solar PV and from 86 $/MWH to 43 $/MWh for onshore wind (IRENA, 2020, p. 14, Fig. S2). Significant reductions have also been achieved for offshore wind and concentrated solar power (CSP). Further reductions of these LCOE can be expected through learning and economies of scale. In many regions of the world, solar PV and onshore wind now undercut fossil fuel plants in terms of LCOE, sometimes even existing fossil fuel plants.
These promising developments notwithstanding, it would be premature to conclude that solar and wind energy are on the verge of making fossil fuels redundant as sources of electricity. The main challenge arises from an effect known as “value deflation” (Hirth, 2013; Sivaram, 2018), which occurs because the production profiles of solar power installations are highly positively auto-correlated (day/night and weather), and analogously for wind turbines. Electricity supply from these sources can be very high at specific times of favorable weather even while their average share of electricity production is still small. And it may vanish almost completely at other times, sometimes for several weeks, when wind speeds are low and there is a persistent cloud cover, or during the night.
As a consequence, if solar and wind energy already contribute a significant share to the overall electricity mix, further additions will predominantly deliver at times when supply is already strong. The economic value of such additions will be low in comparison with “flexible firm” sources, often fossil fuel-based ones, which can deliver at times of low wind and solar output, and which have to remain connected to the grid in order to supply reliable electricity. As a consequence, the overall costs of electricity provision will likely increase, potentially dramatically so, when high shares of solar and wind are being reached — even if these sources have a lower LCOE than other sources (NEA 2018; Loftus et al., 2015).
This value deflation can be studied in the German electricity system. According to an analysis by the Fraunhofer ISE for the year 2019, with wind and solar PV shares of 24.6% and 9.0% (Burger, 2020, p. 13), respectively, the average market values of these sources were reduced to 87.1% and 92.6% (p. 55) of the overall average market value of electricity. The marginal values of the newest additions are even lower. Conversely, the relative value of fossil fuel sources was increased, to 113.2% for fossil gas and 119.2% for hard coal. For the time being, wind and solar power are largely shielded against value deflation through fixed feed-in tariffs. The more these are needed, however, the higher the overall costs of electricity provision, in tension with ZERO, if SMC is assumed.
Another way to see why this is problematic is by recalling that, among all hard-to-decarbonize sectors/activities, load-following electricity is the single largest contributor. Expanding solar and wind energy effectively increases the need for load-following electricity, since fluctuations in demand are compounded with fluctuations in production.
Moreover, dispatchable energy sources such as fossil fuel plants are typically located close to centers of consumption (e.g., large cities) on relatively small industrial sites, which facilitates grid connection. By contrast, solar panels and wind turbines are often spread out over large areas and may be far removed from centers of consumption, in areas with suitable weather conditions (e.g., offshore wind). To bring renewable energy to centers of consumption, as well as to smooth out weather variations across large regions, the associated costs of grid infrastructure are far higher than for dispatchable sources such as fossil fuels or nuclear energy (Jenkins, Luke, & Thernstrom, 2018).
A Zero-Emission Grid with Mostly Variable Renewables?
Load-following electricity is produced by power plants that directly adjust their power output in response to fluctuating demand, which means that they will be often inactive or in waiting-mode. From an economic view, natural gas plants are well suited for this role, or, where available, coal plants or hydroelectricity. One major worry is that, in many cases, emission-free alternatives to fossil fuel plants will have far higher economic costs. By SMC, eliminating fossil fuel plants that provide load-following services from the grid will then become practically impossible.Footnote 5
Various steps can be taken to prevent such an impasse. One, already noted above, is to build out electricity transmission and connect electric grids at scales that are large enough for weather correlation to be low or non-existent. Another partial solution is to provide incentives to customers for adjusting demand to production (“demand response”). Finally, electricity storage can be deployed to smooth out the non-alignment of production and demand. Batteries and, where geographically available, pumped hydro storage are among the prime options for short term storage of limited volume, though their use is economical only when sufficiently utilized — a few times per year is likely not enough. In the more distant future, longer term storage might be provided by “green” hydrogen (produced by solar and wind power through electrolysis), even though for now the conversion losses are substantial (efficiency in the range of 50–70% (IRENA, 2019) and long-distance transportation from renewable-rich areas entails further losses).
According to some energy system modelers, if these steps are taken in an intelligent way, a future electric grid relying far more on solar and wind power than today can be achieved without cost escalation. For instance, MacDonald et al. (2016) find that, by expanding solar and wind power, CO2 emissions from US electricity can be reduced by 80% in 2050, compared to 1990 levels. Similarly, NREL (2012) concludes that renewables can reliably deliver 80% of US electricity by 2050 at costs comparable to today’s. Both studies identify far bigger hurdles when trying to reach the goal of 100% renewables. Mechanisms giving rise to these hurdles are elucidated by Shaner et al. (2018) using a simple continent-sized toy model with historical hourly weather data. As these authors conclude, “to reliably meet 100% of total annual electricity demand, seasonal cycles and unpredictable weather events require several weeks’ worth of energy storage and/or the installation of much more capacity of solar and wind power than is routinely necessary to meet peak demand” (Shaner et al., 2018, abstract).
Future cost developments in solar and wind power as well as electricity storage may render possible the complete decarbonization of US electricity, or even of the entire global energy system, by relying only on renewables, if trends of the last decade continue for several decades and global final energy does not change dramatically until the 2040s (Way et al., 2021). Potential realization pathways are indicated in (Jacobson et al., 2018; Ram et al., 2019; Teske & ed., 2019), based on optimistic assumptions about future costs. For instance, Ram et al. (2019) assume capital cost reductions for the cheapest solar PV technology from 1000 €/kW in 2015 to 246 €/kW in 2050 (p. 238) and for electrolyzers from 800 €/kW in 2015 to 248 €/kW in 2050 (p. 285).
A broader range of potential future cost developments is studied by Sepulveda et al. (2018). They find that, if the emission limit is lowered from 200 gCO2/kWh down to 0 gCO2/kWh and if only variable renewables are deployed, average electricity costs rise by a factor of 1.5 or 2 already in the lowest-cost projection, but by a factor of 3 to 4 in more cost-conservative projections. With such steep cost increases, emission-free electricity would fail to provide alternatives to fossil fuels via electrification in sectors that are hard to decarbonize. By SMC, this would lead to decarbonization failure.
This finding motivates investigating whether inclusion of nuclear energy in one’s energy strategy may reduce the risk of cost escalation and thus facilitate decarbonization.
Economics of Nuclear Energy for Electricity Production
Nuclear energy starts out with an economic advantage over fossil fuels because, to produce a certain quantity of energy, little fuel is needed, in terms of volume or mass. In a standard light water reactor, which requires enriched uranium-235, fuel costs (including enrichment and conversion) amount to about 4 $/MWh, which is typically between two and five times lower than for coal or gas plants, even in the absence of a carbon price. Some nuclear fuel cycles do not require enriched uranium, and their fuel costs are lower still. Nuclear plants tend to have higher operation and maintenance costs than fossil fuel plants, but existing nuclear plants are typically cost competitive with (new, often also existing) fossil fuel plants in a variety of market condition, even in the absence of a carbon price and with the costs of decommissioning and waste disposal taken into account (IEA, 2019a).
The costs of nuclear energy are dominated by its capital costs. As the World Nuclear Association acknowledges: “In general the construction costs of nuclear power plants are significantly higher than for coal- or gas-fired plants because of the need to use special materials, and to incorporate sophisticated safety features and backup control equipment.” (WNA, 2020) Incidentally, it is this need for specific safety features which has made nuclear energy impractical in road transport or aviation.
The capital costs of nuclear power plants consist of the “overnight” construction costs and the costs of financing during construction. The latter depend on interest charges and construction duration. Historically, where overnight construction costs have been in the range of 1000–3000 $/kW of installed capacity, financing costs were comparatively low, and construction durations in the range of 4 to 6 years, nuclear plants have been cost-competitive with fossil fuels plants. This applies to recently completed reactor build projects in China and South Korea (Lovering et al., 2016), and to many reactors built in Western countries in the 1960s, 1970s, and 1980s. France and Sweden provide examples of electric grids with very low carbon intensity (in the range 10–50g of CO2 per kWh), in large part thanks to nuclear power. Reactors there were largely constructed in the 1970s and 1980s, at overnight costs mostly in the range between 1000 and 2000 EUR/kW (Lovering et al., 2016, Fig. 5, EUR normalized to 2010 value). By contrast, the most recent Western reactor build projects, started in the 2000s and 2010s, have overnight capital costs around $8,000/7,000EUR per installed kW (Buongiorno et al., 2018, p. 36, Fig. 2.3), some of them still subject to ongoing cost increases, and with construction times two to three times as long. Eash-Gates et al. (2020, p. 2350) attribute the observed cost increases to “reactor upscaling, a lack of technology standardization, fragmented industry structure and plant ownership, and increasing plant complexity including increases in the number of plant components, new control systems, redundancy in equipment, and added safety features.”
As a result, building new reactors will typically not be part of strategies that conform to QUICK in Western countries, unlike lifetime extensions for existing reactors, which are among the lowest-cost options of avoiding emissions (IEA 2019a). The picture changes when we consider longer time scales.Footnote 6 Some have interpreted the rising capital costs of nuclear plants in Western countries in terms of an inherent “negative learning curve,” but one can also interpret the historical record as constructive evidence that far lower capital costs and shorter build times than today are feasible, even with 1970s inferior levels of technological development.
In order to bring down construction costs and construction times to 1970s levels (and preferably below), learning effects from serial construction will be indispensable. However, as shown by Eash-Gates et al. (2020), serial construction is not sufficient for cost reductions. In the US, construction costs of repeated designs actually rose more often than they fell (Eash-Gates et al., 2020, pp. 2351–2352). According to these authors, the two most important causes of cost increases are decreasing labor productivity and increasing commodity use. They suggest that pursuing designs that lead to reduced commodity use and the automation of construction processes may be the most promising route toward cost reductions.
There are two views on how economies of scale are to be leveraged to make construction of nuclear reactors as cost-competitive as possible. The first is reflected in the fact that, historically, all nuclear reactor build-out programs moved to ever larger designs, with the EPR currently being the largest reactor in the world, at a capacity of 1.65 GW. The main benefit of larger reactors is that, in terms of installed capacity (in MW), they tend to be cheaper than smaller reactors of similar design. Each new reactor involves an investment at the scale of billions of dollars, however, which means that, unless state-actors provide a secure long-term financing, investors will regard such projects as risky. Adverse experiences with first projects of any given design will lead to even more hesitancy, which will further reduce learning opportunities. Experience with cost developments under this strategy is mixed: costs trends rose almost uniformly in the US, but remained largely stable in France (see Grubler (2010) and Escobar-Rangle and Leveque (2015) for nuance) and, more recently, in Japan, China, and South Korea (Buongiorno et al., 2018). Berthélemy and Escobar Rangel (2015) credit design standardization and stable architect-engineer teams with successes in preventing cost escalations and enabling modest temporary cost reductions in the French program.
The second view is that comparatively small “modular” reactors (SMRs) should be constructed at industrial factory-like facilities in large numbers.Footnote 7 Historical records of cost development are absent here since SMRs have not been built at any meaningful scale so far. However, observations about granularity provide grounds for optimism (Sweerts et al., 2020; Wilson et al., 2020): energy technologies with smaller unit sizes have a track record of higher learning rates. Learning rates drop by a few percentage points per order of magnitude of the units built and turn negative for units above certain threshold sizes.
SMRs are likely to experience a “valley of death” in that the first units will be more expensive (per unit of capacity) than more traditional large units, and realizing the envisaged economies of scale will require some stamina of investment. For an agent focused on QUICK, such a long-term investment may not seem appealing, compared to investments in mature renewable energy sources, which promise quick and relatively cheap emission reductions, especially at low levels of VRE penetration. By contrast, investment in SMRs will be attractive for actors guided by FACILITATE and ZERO. Purchasing SMRs, if it enables learning effects and allows some designs to overcome the “valley of death”, may be a cost-effective step in global emissions reduction, similar to the German feed-in tariffs for solar PV in the 2000s mentioned in Section 3.
A more principled worry is that the operating costs of SMRs may be higher than those of large reactors because of a less favorable staff-to-output ratio. Systematically assessing the potential of SMRs in comparison to larger units is beyond the scope of this paper. We suspect that the question of which option will be more beneficial for which actor will depend on geography, technological advancement, and market design.
Nuclear Energy in Zero Emissions Energy Systems
Because of their high capital and low operating costs, nuclear reactors are most suited to provide constant electricity at high power output, covering “base-load.” Although several modern reactors can be used for load-following (Locatelli et al., 2015, 2018), their economic profile makes them not ideally suited for that role. A middling “flexible base” mode is also possible, however, in which reactors generate maximal output for most of the time but modulate down during periods of high solar and wind production and/or very low demand.
Recent modeling (see Jenkins, Luke, and Thernstrom (2018) for a review) of scenarios with different electricity generation mixes concludes that — at close-to-zero emission limits — mixes including at least one “firm” (available with high reliability) low-carbon source tend to be significantly cheaper than those relying exclusively on variable renewables and storage. Notably, this holds for a majority of the 912 scenarios for the Northern and Southern US electricity systems considered by Sepulveda et al. (2018). In the least-cost system, the firm low-carbon source typically runs in the “flexible base” mode. If a zero emissions constraint is imposed, having at least one such firm zero-carbon source reduces costs by 10–62% (see Long et al. (2021) for a more recent effort coming to a similar conclusion for a specific region). Alternatively, if emission limits are imposed by imposing a carbon price, the share of nuclear energy (or some other source with similar economic characteristics) increases, at the expense of unabated fossil fuels and sometimes even variable renewables (Hirth, 2013, 229).
This finding is plausible in the light of real-world examples of low-emission electricity systems (including those of New Zealand, Norway, Sweden, Iceland, Costa Rica, France, and Switzerland), which all rely on a high share of at least one firm low-carbon source: hydroelectric energy, nuclear energy, or geothermal energy.Footnote 8 Indeed, in view of the limited geographical availability of hydropower and geothermal energy, nuclear energy is often the most scalable and geographically flexible firm low-carbon source (Buongiorno et al. (2018), NEA (2018), and IEA (2019a)).
There are upper cost thresholds above which nuclear energy no longer figures in least-cost zero-emission systems. In the modeling of Van Zuijlen et al. (2019) for Europe, for instance, this is the case for capital costs of 7900 €/kW and build times of 10 years per reactor. By contrast, if capital costs of 5300 €/kW are achieved and build times below 7 years, nuclear energy contributes “between 30 and 45% of total demand” (van Zuijlen et al., 2019, p. 13) in least-cost scenarios.
Because electricity from existing reactors is generally cheaper than from new reactors, one may think that there is a clearer ethical case for life-time extensions than for new reactor projects. Indeed, from the point of view of QUICK, this is generally true, and, depending on the exact time scale of quick emission reductions considered, building new reactors will tend to be disfavoured compared with expanding renewables. From the perspective of ZERO and FACILITATE, however, the verdict can be the reverse: reactor lifetimes can often not be extended until 2050, when a zero-emission system should be reached, whereas new reactors can contribute to such a system. (The same applies, of course, to wind turbines and solar panels, which have shorter life spans, in the range of 20–30 years, as opposed to 60–80 years for nuclear plants.) From the perspective of ZERO, the main benefit of reactor life-time extensions may well be that, by obviating the temporary need for new fossil fuel plants, they help avoid the lock-in of fossil fuel infrastructure. However, it is only through successful new reactor projects that learning effects can be enabled which are needed for making nuclear-driven decarbonization more economically attractive for others. Accordingly, from the perspective of FACILITATE, new reactor projects will often be ethically preferable over life-time extensions, even if the latter have a cost-benefit ratio that is better prima facie. Indirectly, life-time extensions may facilitate new reactor projects by preserving know-how and infrastructure.
Recent modeling by Wealer et al. (2019) suggests that new reactors in Western liberalized electricity markets will not be profitable for the foreseeable future. On the basis of this finding, these authors advise against any role for new nuclear reactor builds in climate change mitigation. However, if we combine this finding with the results reviewed above on least-cost near-zero emissions electricity mixes, the real lesson is different. Markets in which renewables are largely shielded from value deflation and where carbon prices are too low to impose near-zero emission limits, but which are otherwise “liberalized,” may not incentivize strategies that are in conformity with ZERO. Moreover, the historical cost records for new nuclear reactors show that capital costs below 3000 $/kW are obtainable. Below this threshold, according to Ingersoll et al. (2020), investments in new reactors become economically attractive even in current liberalized electricity market designs. For actors oriented toward FACILITATE, helping to make such reactors globally available is a natural goal.
Beyond electricity, further potential contributions of nuclear energy to zero emissions energy systems include shipping (Hirdaris et al., 2014), industrial heat (Friedman et al., 2019), and synthetic fuels based on hydrogen. In hydrogen production, nuclear energy’s constant production profile is an advantage because it enables high utilization of commodities such as electrolyzers, as well as higher temperatures for more efficient hydrogen production (WNA 2021). However, to make such hydrogen competitive fossil fuel-based hydrogen, both capital costs and commodity costs must be very low (IEA, 2019b).