Main

Meeting the Paris Agreement’s target of carbon neutrality by mid-century1 will require a large increase in low-carbon generation2,3, and this is especially true in the developing world. The United Nations 2030 Agenda for Sustainable Development includes a goal to secure reliable energy access for all4,5,6,7,8. Renewables are considered essential for meeting both of these goals. However, in its 5th Assessment Report, the Intergovernmental Panel on Climate Change included nuclear power among the low-carbon generation technologies that could be used to limit carbon emissions9. Increasingly, small modular reactors (SMRs) and micro modular reactors (MMRs) have been discussed as the future of nuclear energy, but as yet, no market demand has materialized for these machines. While there is no firm standard, microreactors are often considered to be ≤20 MWe, with SMRs being up to ~300 MWe (ref. 10). Gilbert and Bazilian11 and others highlighted the potential for reactors at these scales to be used in a distributed fashion to help address electricity poverty12. With over 730 million people estimated to live in electricity poverty in 2022 (ref. 13), the potential electricity market is large. However, nuclear power has unique security, oversight and proliferation issues that impact its deployment, and economics has often been an issue relative to other generation technologies, such as solar systems coupled with large-scale pumped hydroelectric storage14.

Siting constraints are critically important considerations affecting where small reactors and microreactors could go15,16. Physical characteristics of potential sites must be analysed (for example, hazards, population proximity and water access). Regulatory and proliferation oversight present other necessary angles to contemplate. Many countries would need to build at least part of the regulatory infrastructure from the ground up, and it is estimated that two-thirds of the known or suspected nuclear-weapon programmes have been linked to civilian nuclear programmes17,18. The Russian invasion of Ukraine has also highlighted the potential for nuclear reactors to become involved in military conflicts. Cost is always a core consideration in power systems. Recent estimates for Nth-of-a-kind microreactors give a levelized cost of electricity (LCOE) range from just under US$0.10 kWhe−1 to over $0.30 kWhe−1 depending on the assumed learning rates19. While high, these prices are similar to those currently being paid in many rural areas, even in developed countries20 but they are above those of renewables and energy storage in many developing regions21.

The viability of using nuclear energy to increase electricity access will depend on long-standing issues of sitability, cost, governance, safety, security and public perception. To achieve and sustain Nth-of-a-kind reactor costs, a substantial market will be needed22. The developing world could offer this market by aiming at the World Bank’s Energy Sector Management Assistance Program (ESMAP) ultimate target of at least 8.2 kWhe daily power per capita (Supplementary Table 1 in Supplementary Note 1). In this Article, we consider the potential for SMRs and MMRs ≥1 MWe to help address electricity access by considering need, sitability, regional conflict, regulatory capacity and economics. Locations that meet scale-specific siting constraints are identified on the basis of US and International Atomic Energy Agency (IAEA) criteria. We compute the scale of generation capacity that would be required for the unelectrified to achieve an ESMAP tier 5 level.

Technical siting constraints

A geospatial analysis is used to determine where micro and small modular could be sited (Fig. 1). Scale-specific reactor sites are identified using US and IAEA criteria. These include proximity to water (for cooling and emergency use), avoidance of hazards (for example, seismic, flood, landslide and so on) and exclusion of protected areas such as wetlands15,16 (Table 2). Details of these criteria are in Methods. The results are shown in Fig. 2. The wide geographic compatibility of microreactors is largely due to their minimal water requirements. To determine how viable regions overlap with energy demand, night satellite images23 are used in conjunction with ambient population data24. Electrification needs are assumed to exist only in regions with no visible night-time light. Figure 3 shows the geospatial distribution of capacity needed at a 0.5° resolution. The data are truncated to countries with incomplete electrification rates25. An interesting observation is that some locations would support large-scale nuclear facilities, particularly in Sub-Saharan Africa (for example, parts of Kenya). Small reactors (50 MWe and below) would be site compatible with the electrification of up to 70.9% of people (~733 million people) living in regions with no visible night-time light.

Fig. 1: Geospatial analysis for the ability of MMRs and SMRs to address electricity poverty.
figure 1

Visible night-time light and ambient population are used in conjunction with electrification rates to identify the number of people in electricity poverty as a function of location. Multiple siting and operational criteria (left) are used to exclude regions that would not be appropriate for reactor systems. Suitable locations and need (centre) are then mapped globally. The current analysis is then confined to all regions where electrification rates are incomplete nationally. LWR, light water reactor.

Full size image
Fig. 2: Locations that meet basic siting criteria by reactor scale.
figure 2

The figure shows regions that could host reactors of different scale based on landslide potential, protected areas, access to water, population density, exclusion zone, and seismic and flood risk. Several of these (for example, population density and proximity to water) depend on the size of the reactor, and the scale-dependent applicability of a region is shown by shades of green. This figure provides site compatibility and not an estimate of future deployment, and does not supersede the necessity of further site-specific studies. MMRs, SMRs and LWRs are shown.

Full size image
Fig. 3: System size needed to meet tier 5 demand for the no-light population.
figure 3

The heat scale shows the capacity needed to meet tier 5 demand for countries whose electrification rate is reported below 100% (ref. 25). The spatial resolution is 0.5° (approximately 50 km by 50 km).

Full size image

To estimate where MMR- and SMR-based minigrids would be viable, we assess the effects of competition from large grids and other minigrid alternatives. We consider the distance of potential regions from known grid infrastructure26. We find that 62.6% of the population without visible night-time light live within 5 km of an existing transmission line, 76.8% within 10 km, and 87.5% within 25 km. To connect these populations to existing grid infrastructure, transmission lines would need to be extended and capacity upgraded to support the increased load. We adopt conventional recommendations for locating a minigrid27 (Methods). Considering now only this minigrid market of close to 250 million people and associated technical siting constraints, most (78.4%) of the population identified as unelectrified and more than 10 km from existing transmission lines could be served by small nuclear reactors up to 50 MWe. Almost 200 million people fall into this category. Providing tier 5 access to these grid-remote communities would require more than 70,000 1 MWe reactors (Fig. 4), which constitutes one of the largest potential markets for small and micro modular reactors (SMMRs) identified thus far.

Fig. 4: Number of people potentially served by MMRs of 1 MWe.
figure 4

The heat scale shows the number of people living in locations with no visible night-time lights that could be electrified by 1 MWe reactors. The data are truncated to countries with electrification rates reported below 100%.

Full size image

Governance and conflicts constraints

Many of the countries that suffer from energy poverty also rank poorly on indicators of governance quality and capacity28. Four Worldwide Governance Indicators28 are used to estimate the capacity of countries to develop effective institutions of oversight: control of corruption, rule of law, government effectiveness and regulatory quality. While all of these could be considered essential, some countries with successful nuclear programmes also rank poorly on these measures. Figure 5 shows these four indices mapped relative to their values in Ukraine, which ranks in the bottom quarter but obtains about 50% of its electricity from legacy nuclear plants inherited from the Soviet Union and operates its industry with substantial foreign assistance. We use Ukraine as a baseline for minimum governance levels. A sensitivity analysis is given in Supplementary Fig. 1 in Supplementary Note 3.

Fig. 5: Indicators of governance in countries with incomplete electrification, relative to Ukraine.
figure 5

a–d, For countries with less than 100% reported electrification rates, the Worldwide Governance Indicators28 for control of corruption (a), rule of law (b), governmental effectiveness (c) and regulatory quality (d) are shown relative to their values in Ukraine. The majority of these countries surpass Ukraine in control of corruption, rule of law and governmental effectiveness. Regulatory quality is seen to be a bigger issue in these developing regions.

Full size image

The physical security of nuclear plants is another consideration, especially in conflict regions. The Chernobyl and Zaporizhzhia reactors were taken over by Russian forces in 2022 as part of Russia’s invasion of Ukraine. While the reactor internals were not harmed, shelling caused disruptions that led to unsafe conditions29. These events show that nuclear facilities can make attractive targets or economic hostages30. In addition to kinetic attacks, there have been cyber infiltrations, and this type of vulnerability remains poorly understood31.

To assess the impact of military conflict, regions with recent conflict are also excluded. A convex hull method is used to determine regions where conflict events cluster over time32. Events of interest here are those that involve either a state entity or an organized armed group. Figure 6 shows conflict events against the state in countries with incomplete electrification rates. More than a third of the population (39.7%) that could be served by some types of SMMR are removed from the market if conflict cluster zones are excluded, and 84.4% are removed if all four governance indicator metrics are additionally required to be at a Ukraine-equivalent minimum.

Fig. 6: Conflict events against the state in countries with high rates of electricity poverty (2016–2020).
figure 6

Of the 97 countries with electrification rates below 100%, 46 have experienced conflict events against the state between 2016 and 2020. The heat scale shows the number of conflict events in each country, while the green dots display the actual location of each battle. Risk zones based on a clustered convex hull analysis of the battle events are shown with a hashed pattern. Countries that were removed from this analysis (complete electrification rate) are shown in grey.

Full size image

Table 1 summarizes the potential limitations posed by conflict and governance issues. However, what degree of country-level corruption should preclude the development of nuclear power remains an open question, and Supplementary Note 3 gives a sensitivity analysis of these metrics.

Table 1 Potential exclusions based on conflict and governance indicators levels
Full size table

Economic considerations

Reactors will need to compete with other low-carbon energy sources. Academics have issued pre-engineering levelized LCOE targets for small reactors (not microreactors) between $40 MWhe−1 and $70 MWhe−1 (2019 dollars)33,34. A few conceptual designs have matured to the point where an empirical estimate of LCOE is possible. Argentina’s CAREM25 started construction in 2014, and at current forecasts, its electricity should cost roughly $370 MWhe−1 (assuming a 7.7% weighted average cost of capital and a completion date in 2027 (refs. 35,36). The US-designed NuScale is still in the post-licence, pre-construction engineering phase. The project is being built as a collection of six reactors sharing a site and turbine hall, resulting in a power output equal to a full-scale conventional reactor, to achieve more favourable economics. It is therefore a lower bound on the cost37. NuScale’s pre-construction LCOE estimates is currently $120 MWhe−1 (refs. 38,39) but could rise further considering that every nuclear reactor project started in Western nations within the past two decades has in the end cost more than double the pre-construction estimate40,41,42,43,44. An analysis of predicted costs for 19 SMRs showed median LCOE as low as $116 MWhe−1 for high-temperature reactors and $218 MWhe−1 for pressurized water reactors45.

Microreactors have less favourable economics because of their smaller size. In 2019, the Nuclear Energy Institute (NEI) published LCOE estimates of $140–410 MWhe−1 for first-of a kind reactors based on assumed capital and operating costs that were not supported by empirical data19. In 2021, Buongiorno produced a similar type of estimate at $100–150 MWhe−1 for first-of-a-kind reactors46. The lower end of this range is consistent with the 2019 estimate from the US NEI for Nth-of-a-kind microreactors19. A bottom-up cost estimate of an actual design carried out by Idaho National Laboratory estimated the LCOE from a first-of-a-kind microreactor to be $2,200 MWhe−1 (ref. 47). These estimates vary by a huge amount, illustrating the poor state of knowledge. Moreover, all of these estimates, for both small reactors and microreactors, assume that the reactor operates at full power 95% of the time, which is not possible for applications where load following is required, such as with minigrids. Load following would roughly double the LCOE.

It is common to compare nuclear LCOE with the existing electricity cost to establish market potential. Utility costs are not available in detail for all the countries considered here. However, a study of 39 African countries by Kojima and Trimble48 showed that a utility’s costs for electricity ranged from $80 MWhe−1 to more than $600 MWhe−1. Superficially, the NEI estimates for microreactor costs appear to be compatible as more than 25 of the countries studied by Kojima and Trimble have costs in excess of $200 MWhe−1. However, this type of comparison provides an incomplete view of the situation. Kojima and Trimble (2016) also report that, of the 39 countries considered, utilities recovered their full costs in only two, and failed to recover revenues to meet their operating expenses in 18 (refs. 48,49). If utilities are already not financially solvent, it is unclear how services can be expanded.

Any expansion of electricity services must also consider the cost of alternatives. Several studies have recently looked at the potential for solar power and battery energy storage to address electricity access in the developing world21,50 and generated geospatial distributions for its LCOE. L’Her et al.21 in particular does this for all countries with incomplete electricity access. By comparing these LCOE distributions with an optimistic Nth-of-a-kind, load-following microreactor LCOE target of $150 MWhe−1, 86.7% of the identified potential for SMMR and 87.4% of the SMMR-powered minigrids for remote communities disappears. Supplementary Fig. 2 in Supplementary Note 4 shows a sensitivity analysis of the impact of nuclear minigrids costs on the estimated population served. Cash-stranded utilities in these countries, like those in many developed regions, will be under pressure to look for the cheapest generation options. Additionally, security and maintenance are essential components of nuclear safety, and the potential impact financially strain may have on the safe operation of nuclear power cannot be ignored. Combining the economic limitations with the governance limitations discussed previously, only 5% of the identified potential for SMMR and less than 2% of the SMMR-powered minigrids for remote communities remain.

Regulation and proliferation

Proliferation concerns do not directly limit the deployment of reactors, but they should not be ignored. By some estimates, 70% of the 34 known or suspected nuclear-weapon programmes have had ties to civilian nuclear programmes17,18. Studies find that interest in developing a nuclear-weapon programme is tied to intercountry conflict, economic stress and the inability to maintain security through non-nuclear means51,52. The majority of the world’s population currently living without access to electricity resides in African countries that have some or all of these factors53. Intercountry conflict and civil nuclear cooperation are the two most consistent predictors of whether a country will start a nuclear-weapon programme54. If nuclear proliferation risk is indeed correlated to nuclear cooperation agreements, then developing a civil nuclear programme would increase the risk of proliferation after the fact. Although far from being a guarantee of non-proliferation, safeguards can reduce the risk that declared facilities are used for nuclear proliferation, and make undeclared facilities more difficult by allowing IAEA to oversee necessary supply chains. The Iran nuclear agreement provides a possible example for how rules could establish voluntary limits to fuel-cycle infrastructure that are the riskiest for proliferation18.

In addition to technical feasibility, a variety of regulatory factors bear on the safety and security of nuclear power plants. The IAEA has developed the ‘milestones’ approach, which enables a phased development process for a nuclear power programme over a 10–15 year period55. However, many countries may encounter challenges to set up efficient, independent and well-funded regulatory infrastructure. The availability of trained staff is part of this hurdle. We estimate that at least 57,000 regulatory staff are needed to oversee the theoretical technical market of 70,000 1 MWe microreactors for remote communities in developing countries—almost the number of people employed across the wider US nuclear industry. Additional discussion can be found in Supplementary Note 2. Regulatory effectiveness is another potential hurdle. ISO 17020 is an international standard that specifies the requirements for the competence, impartiality and consistency of bodies that perform inspection activities56. In this context, ‘impartiality’ would encompass independence from political or industry influence. At the same time however, regulatory authority would also have to be created and funded by respective governments in a way to ensure the practical competence and effectiveness of its staff. These factors are central to an effective nuclear regulator57. The World Bank’s ‘regulatory quality’ metric (Fig. 5d) captures these essential features. Future work could assess how this metric translates specifically to the creation and operation of effective nuclear regulator.

There are several limitations with the current study. Because of sensitivity thresholds, night-time light data from the VIIRS satellite will fail to register some areas that do in fact have limited electricity. We also assume a tier 5 level of electrification as an aspirational standard, though clearly lower levels of electricity access would also be an improvement. Despite being based on current best practices27, there is no one-size-fits-all service area for a minigrid system, and transmission grid development and costs are important additional considerations. Predicting political stability (governance and conflicts) over long timeframes is difficult, and historical data paint an informative but incomplete picture. The Worldwide Governance Indicators are the best available but do not capture local variability. It is also noteworthy that ranking poorly on an indicator like corruption does not necessarily mean that infrastructure will fail to be built correctly58. Some technical siting criteria could be impacted by climate change, such as the access to water under more severe droughts. While a recent historical availability assessment was considered, ongoing climate changes may remove some potential locations. Finally, while recent costs analyses for microreactors are available, they have not been demonstrated and are likely to be optimistic. Similarly, projected costs for solar systems with sufficient energy storage are still unproven, and economy of scale might prove difficult to sustain when it comes to batteries. Where these costs sit, relative to utility costs, is also only available for utilities in 39 countries in Africa, not for all the countries with incomplete electrification rates.

Conclusions

Based on siting criteria alone, large markets appear to exist where SMRs and MMRs could technically be sited. Close to 200 million live farther than 10 km from existing transmission infrastructure and might be candidates for minigrid with reactors in the 1–50 MWe range. Microreactors will be viable only if they are economically competitive. Utility-scale solar with battery storage are seen to have lifetime LCOE below those for estimated Nth-of-a-kind microreactors in many regions. We find that only ~13% of the nuclear-potential grid-remote population identified as living in regions without visible night-time light would favour nuclear over utility-scale solar with battery storage with near-future cost estimates. This represents approximately 18 GWe of microreactor capacity. Although this number is not an unsubstantial potential market—and is larger than other previously studied electricity markets for microreactors—one must also consider that many of the utilities in these regions operate at a deficit and therefore lack the revenues to safely support nuclear power. The scale and cost of standing up the regulatory and oversight apparatus for microreactors is also a major consideration. Simultaneously accounting for governance issues would further lower the realistic grid-remote market down to less than 2% (~2 GWe). Financial strain could also affect the ability of utilities to maintain security and meet the stringent operations and maintenance obligations of a nuclear generator, suggesting that nuclear might be poorly suited to developing regions.

Proliferation and US export policy may favour entry by countries like Russia; however, liberalization of export policies will not overcome the more fundamental economic and regulatory challenges. Recent research is pointing at the risk of letting Russia foster such agreements when it comes to nuclear technologies59, and may force the Western world to develop a suboptimal market.

Conflict and poor governance were found to be the dominant factors that could limit where SMMRs could go. More than a third of the population (39.7%) living in regions with no visible night-time light that could be served by some type of nuclear reactor reside in clustered conflict zones. Over 80% of the population living in regions without visible night-time light, and where nuclear power could be an option, are within countries where one or more of the governance indicators fall below that of the Ukraine benchmark.

Methods

Data layers

The US Nuclear Regulatory Commission published guidance on the siting requirements for nuclear reactors60. The three hazards considered are seismic61, flooding62 and landslide63,64. Societal constraints are added, with limitations based on population density24 (Supplementary Fig. 3) and the unfit nature of wetlands65 and protected areas66,67. Additionally, any reactor must be sited on land68,69, and access to sufficient water sources is used to validate a location70,71. Details are given in Supplementary Note 5. This study also accounts for any site to perform reasonable and cheap vegetation management to mitigate wildfire risks and for current design, able to withstand high-velocity windspeed. Those different layers are collated to filter out unsuitable locations for nuclear reactors of varying sizes. Water requirement for cooling and emergency operation is defined relative to the power capacity71. The adequate urgent protective action planning zones for SMRs and MMRs are still up for debate. A linear interpolation interpretation of international guidelines is used in this study72. Table 2 presents the preliminary siting requirements used in this study.

Table 2 Siting requirements considered for reactors of different sizes
Full size table

Population living without access to electricity

A high-resolution assessment of electricity poverty was performed using satellite imagery data, night-time light and population estimations (Supplementary Note 6). Using locations with population but no night-time lights, the total estimated population living in areas without visible night-time light is approximately 1.5 billion persons (Supplementary Figs. 4 and 5). However, it should be noted that visible night-time light is a proxy for electricity poverty, and in very sparsely populated regions or areas where light pollution is a concern, a lack of visible light may not be indicative of poverty. Data from the World Bank on electrification rate per country in 201925 are used to filter only the countries presenting an electrification rate below 100% (Supplementary Fig. 5). Ninety-seven countries fall into this category, most of them in Sub-Saharan Africa. Using this filter, 1,047 million persons are found to be in electricity poverty. This is around 40% above estimates from the World Bank (759 million)25. It is, however, important to note that electrification does not automatically mean a tier 5 access, and the discrepancy can be explained by the capture of low-level of electricity access using the night-time lights methodology.

Reactor service region size

The reactor service area is determined using a best practice estimate for the minimum localized population density for a minigrid at ≥50 people per square kilometre and farther than 10 km from the existing grid26. The population density here is for people without electricity access and is used to derive the maximum service area (Supplementary Table 2 in Supplementary Note 6). Locations with <50 unelectrified people per square kilometre are removed. Within a service region, reactors are assumed to supply electricity to the fraction of the unelectrified population that their capacity allows.

Ambient population density24 is used in conjunction with global data on transmission line locations26 to determine the number of people living within different distances from existing grid infrastructure. Of the 1,047 million people in regions with no visible night-time light, 76.8% (804 million) live within 10 km of a transmission line.

Conflict risk

Unstable regions are equally unfit to develop nuclear power due to safety concerns. Recent terrorism activity and violence against the state should thus be taken into account. The Uppsala Conflict Data Program Georeferenced Event Dataset73,74 is used to identify the location of conflicts. No method exists to assess the potential risk zone extent of a conflict at a local level. In this study, we follow the guideline of the Uppsala Conflict Data Program Georeferenced Event Dataset on defining a convex hull of battle event clusters32. Additional detail is given in Supplementary Note 7.

Governance

This study considers four Worldwide Governance Indicators28: control of corruption, rule of law, government effectiveness and regulatory quality. These indicators are used to quantify the capacity of countries to develop effective institutions of oversight. It is important that a regulatory agency be supported by the local government (funding, authority and manpower) while remaining independent. Scoring low on the Worldwide Governance Indicators indicates a risk to fail this mission. We use Ukraine as a baseline for minimum governance levels, with a sensitivity analysis given in Supplementary Note 3.

Cost competitiveness

The World Bank is one of the main funding sources for energy projects in developing regions of the world. Most of their new projects are hydroelectricity expansion and utility-scale solar plants with battery storage, notably due to the quantity of these resources in the developing world. An energy flow model, using projected costs at different horizons and discount rates, was developed to combine solar irradiance resource, population electricity demand by ESMAP access tier and plant sizing21. Projected LCOE were computed for coupled utility-scale solar-battery systems to address electricity poverty. These projected cost estimates are coherent with other research, such as ref. 50, which looked at rooftop systems in sub-Saharan Africa.