Nuclear against the alternatives, one metric at a time
Each bar is one published figure with its year, region and definition; hover for the source, open the table for every row of the metric, and switch any technology to another row. Figures with different units stay in the table rather than on the axis. The default rows follow a written rule (D44).
Capacity factor, %
Choose which row each technology shows (default: the headline row, rule D44)
Dispatchability, inertia, wildlife, landscape, waste toxicity
These are statements, each with its source, not numbers: whether a plant can follow demand, whether it adds rotating inertia to the grid, what it does to birds, bats, fish and rivers, what it does to a landscape, and what its waste is.
| Technology | Dispatchability | Grid inertia | Wildlife | Landscape | Waste toxicity |
|---|---|---|---|---|---|
| Nuclear | Dispatchable, normally run as baseload at 85-94% capacity factor; modern PWRs can load-follow - the NEA's 50%-VRE scenario has nuclear ramping to 30-35% of capacity within one hour. OECD-NEA, The Costs of Decarbonisation: System Costs with High Shares of Nuclear and Renewables (2019), executive summary, 2026 · reputable secondary | Synchronous generator: the rotating turbine-generator mass inherently provides system inertia (kinetic energy that resists frequency change in the first seconds after a disturbance). Large nuclear units (1,000+ MW) are among the biggest single contributors of inertia on a grid. ENTSO-E (16 Dec 2020), 'Inertia and Rate of Change of Frequency (RoCoF)', Version 17, System Protection & Dynamics Inertia TF, 2020 · primary source | 0.6 avian fatalities per GWh Sovacool, B.K. (2012), 'The avian and wildlife costs of fossil fuels and nuclear power', Journal of Integrative Environmental Sciences 9(4): 255-278, doi 10.1080/1943815X.2012.746993 (abstract via Tethys), 2012 · not re-verified | Compact: a 1,000 MW plant occupies about 1.3 square miles (3.4 km2) of site (NEI, median of 59 US sites); visual impact concentrated in cooling towers (up to ~200 m) or a coastal/lakeside site with the dome. Fuel-cycle land (uranium mining) is small per TWh (Lovering 2022: 7.1 ha/TWh/y total). Nuclear Energy Institute (9 Jul 2015), 'Land Needs for Wind, Solar Dwarf Nuclear Plant's Footprint' (citing NEI 'Land Requirements for Carbon-Free Technologies', 2015), 2026 · reputable secondary | High-level waste is ~3% of the volume but ~95% of the radioactivity of all radioactive waste; it is heat-generating and requires shielding and isolation for thousands of years; radioactivity of used fuel falls to one-thousandth of its initial level after 40 years. Low-level waste is ~90% of volume and 1% of radioactivity. World Nuclear Association, 'Radioactive Waste Management' (updated 22 Nov 2025), citing IAEA inventory estimates (published Jan 2022, end-2016), 2025 · reputable secondary |
| Solar PV | Variable (non-dispatchable): output follows daylight and weather; can be curtailed but not increased on demand. Firming needs storage or backup; US fleet capacity factor 23.2% (2024). OECD-NEA, The Costs of Decarbonisation: System Costs with High Shares of Nuclear and Renewables (2019), executive summary, 2026 · primary source | Converter (inverter)-connected: 'delivers, therefore, in principle no additional inertia to the power system' (ENTSO-E); synthetic inertia or fast frequency response can be programmed but is not inherent. ENTSO-E (16 Dec 2020), 'Inertia and Rate of Change of Frequency (RoCoF)', Version 17, System Protection & Dynamics Inertia TF, 2020 · primary source | birds killed per year at US utility-scale solar facilities Walston, L.J., Rollins, K.E., LaGory, K.E., Smith, K.P., Meyers, S.A. (2016), 'A preliminary assessment of avian mortality at utility-scale solar energy facilities in the United States', Renewable Energy 92: 405-414, doi 10.1016/j.renene.2016.02.041 (abstract via IDEAS/RePEc), 2016 · reputable secondary | Extensive but low: ground-mounted PV covers 7-9 acres per MWac (Ong 2013; ~3 ha/MW) across the whole site, changing land use from farmland/desert to fenced arrays 2-4 m high; rooftop PV has near-zero additional landscape impact. Ong, S., Campbell, C., Denholm, P., Margolis, R., Heath, G. (June 2013), 'Land-Use Requirements for Solar Power Plants in the United States', NREL/TP-6A20-56290, 2026 · primary source | Crystalline-silicon panels are mostly glass, aluminium frame, silicon and polymer with small amounts of copper, silver, tin and lead solder; thin-film CdTe panels contain cadmium. Panels are bulky but low-toxicity waste; recycling recovers glass, aluminium and copper. IRENA and IEA-PVPS (2016), 'End-of-Life Management: Solar Photovoltaic Panels', 2016 · not re-verified |
| Wind, onshore | Variable (non-dispatchable): output follows wind speed; curtailable; US fleet capacity factor 34.3% (2024). Firming needs storage, backup or interconnection. OECD-NEA, The Costs of Decarbonisation: System Costs with High Shares of Nuclear and Renewables (2019), executive summary, 2026 · primary source | Converter (inverter)-connected: 'delivers, therefore, in principle no additional inertia to the power system' (ENTSO-E); synthetic inertia or fast frequency response can be programmed but is not inherent. ENTSO-E notes wind 'synthetic inertia' can reduce the frequency nadir and RoCoF but may cause a second nadir. ENTSO-E (16 Dec 2020), 'Inertia and Rate of Change of Frequency (RoCoF)', Version 17, System Protection & Dynamics Inertia TF, 2020 · primary source | 234012 birds killed per year, US land-based wind (median; min-max) US Fish & Wildlife Service, 'Threats to Birds: Migratory Bird Mortality - Questions and Answers' (estimates as of 2017; Loss et al. 2013b for wind), 2017 · primary source 0.27 avian fatalities per GWh Sovacool, B.K. (2012), 'The avian and wildlife costs of fossil fuels and nuclear power', Journal of Integrative Environmental Sciences 9(4): 255-278, doi 10.1080/1943815X.2012.746993 (abstract via Tethys), 2012 · reputable secondary | Dispersed and tall: turbines installed in the US in 2023 average 3.4 MW with 103 m hub height and 134 m rotor diameter (DOE), visible for tens of kilometres; the project area is 34.5 ha/MW but only ~1 ha/MW is physically occupied (Denholm 2009), so farming continues between turbines. US DOE EERE, 'Wind Turbines: the Bigger, the Better' (updated Aug 2024), citing the Land-Based Wind Market Report: 2024 Edition (LBNL), 2026 · primary source | n/a |
| Wind, offshore | Variable (non-dispatchable) but steadier and higher capacity factor than onshore (Lazard assumption 45-55%; IRENA 2024 global average 42%). Lazard, Levelized Cost of Energy+ (LCOE v19.0, LCOS v11.0), July 2026, 2026 · primary source | Converter (inverter)-connected: 'delivers, therefore, in principle no additional inertia to the power system' (ENTSO-E); synthetic inertia or fast frequency response can be programmed but is not inherent. Offshore wind connected via HVDC cannot yet provide inertia ('not state of the art yet', ENTSO-E). ENTSO-E (16 Dec 2020), 'Inertia and Rate of Change of Frequency (RoCoF)', Version 17, System Protection & Dynamics Inertia TF, 2020 · primary source | n/a | Marine rather than land area: turbines (often larger than onshore) stand in the sea; visual impact from shore depends on distance (OWID treats offshore area as marine, not land), plus onshore substations and cable landfalls. Ritchie, H. (2022-06-16), 'How does the land use of different electricity sources compare?', Our World in Data, 2026 · reputable secondary | n/a |
| Gas, combined cycle | Dispatchable, mid-merit to baseload: US fleet capacity factor 60.5% (2024); start-up in ~1-3 hours, flexible ramping; Lazard models 30-90% CF. US EIA, Electric Power Annual, Table 4.08.A Capacity factors for utility scale generators primarily using fossil fuels, 2026 · primary source | Synchronous generator: the rotating turbine-generator mass inherently provides system inertia (kinetic energy that resists frequency change in the first seconds after a disturbance). Both the gas turbine and the steam turbine generators are synchronous. ENTSO-E (16 Dec 2020), 'Inertia and Rate of Change of Frequency (RoCoF)', Version 17, System Protection & Dynamics Inertia TF, 2020 · primary source | 9.4 avian fatalities per GWh (fossil-fuelled power stations) Sovacool, B.K. (2012), 'The avian and wildlife costs of fossil fuels and nuclear power', Journal of Integrative Environmental Sciences 9(4): 255-278, doi 10.1080/1943815X.2012.746993 (abstract via Tethys), 2012 · not re-verified | Small plant footprint (UNECE 1.0 m2/MWh) but a dispersed upstream landscape of well pads, pipelines and gas fields (Lovering 2022: footprint 410, spacing 1,900 ha/TWh/y). Lovering, J., Swain, M., Blomqvist, L., Hernandez, R.R. (2022), 'Land-use intensity of electricity production and tomorrow's energy landscape', PLOS ONE 17(7): e0270155, 2026 · reputable secondary | n/a |
| Gas, peaker | Dispatchable, peaking: simple-cycle gas turbines start within minutes and run only at times of high demand; US fleet capacity factor 13.9% (2024); Lazard assumes 10-15% CF. US EIA, Electric Power Annual, Table 4.08.A Capacity factors for utility scale generators primarily using fossil fuels, 2026 · primary source | Synchronous generator: the rotating turbine-generator mass inherently provides system inertia (kinetic energy that resists frequency change in the first seconds after a disturbance). Open-cycle gas turbines provide inertia only while running (low capacity factor). ENTSO-E (16 Dec 2020), 'Inertia and Rate of Change of Frequency (RoCoF)', Version 17, System Protection & Dynamics Inertia TF, 2020 · primary source | n/a | As for gas CCGT: compact plant, often sited at substations or industrial sites; run only a few hundred hours a year (EIA 2024 CF 13.9%), so visual/noise impact is intermittent. US EIA, Electric Power Annual, Table 4.08.A Capacity factors for utility scale generators primarily using fossil fuels, 2026 · reputable secondary | n/a |
| Hydrogen | Dispatchable as a storage medium: electrolysers absorb surplus power and hydrogen can be stored for weeks/months and re-converted in turbines or fuel cells on demand, at ~35% round-trip efficiency (PNNL). PNNL, Energy Storage Grand Challenge Cost and Performance Assessment 2020, Hydrogen chapter, 2026 · primary source | Depends on the re-conversion technology: hydrogen burned in a gas turbine drives a synchronous generator (inherent inertia); a fuel cell is inverter-connected (no inherent inertia). Electrolysers are flexible loads that can provide demand-side frequency response. ENTSO-E (16 Dec 2020), 'Inertia and Rate of Change of Frequency (RoCoF)', Version 17, System Protection & Dynamics Inertia TF, 2020 · reputable secondary | n/a | Industrial: electrolyser halls, compressors and either salt-cavern storage (underground, little surface change) or above-ground tanks; PNNL costs cavern storage at 2-10 USD/kWh. No published land-use intensity captured in this task. PNNL, Energy Storage Grand Challenge Cost and Performance Assessment 2020, Hydrogen chapter, 2026 · not re-verified | n/a |
| Hydro | Dispatchable where there is a reservoir (fast ramping, seconds to minutes, also pumped storage); run-of-river output is seasonal/variable. US fleet capacity factor 34.6% (2024), IRENA 2024 new-project average 48%. US EIA, Electric Power Annual, Table 4.08.B Capacity factors for utility scale generators primarily using non-fossil fuels, 2026 · reputable secondary | Synchronous generator: the rotating turbine-generator mass inherently provides system inertia (kinetic energy that resists frequency change in the first seconds after a disturbance). Hydro units can also run as synchronous condensers to supply inertia without generating. ENTSO-E (16 Dec 2020), 'Inertia and Rate of Change of Frequency (RoCoF)', Version 17, System Protection & Dynamics Inertia TF, 2020 · primary source | Dams and reservoirs are 'the leading contributors to the loss of river connectivity': only 37% of rivers longer than 1,000 km remain free-flowing over their entire length (Grill et al. 2019). Barriers block fish migration, alter flow and temperature regimes and flood terrestrial habitat. Grill, G. et al. (2019), 'Mapping the world's free-flowing rivers', Nature 569: 215-221, 2019 · primary source | Largest direct footprint of all sources: reservoirs flood valleys (UNECE land use 33.4 m2/MWh - the highest in the UNECE dataset; Lovering median 650 ha/TWh/y, mean 15,000), permanently transforming landscapes and displacing settlements; run-of-river plants are far smaller. OWID dataset 'Life-cycle impacts of energy sources (UNECE)' (updated 8 Jul 2022), based on UNECE (2021) Life Cycle Assessment of Electricity Generation Options, 2026 · primary source | n/a |
| Coal | Dispatchable, traditionally baseload but now cycling in the US (fleet capacity factor 42.6% in 2024); slow starts (hours) and limited ramping compared with gas. US EIA, Electric Power Annual, Table 4.08.A Capacity factors for utility scale generators primarily using fossil fuels, 2026 · primary source | Synchronous generator: the rotating turbine-generator mass inherently provides system inertia (kinetic energy that resists frequency change in the first seconds after a disturbance). ENTSO-E (16 Dec 2020), 'Inertia and Rate of Change of Frequency (RoCoF)', Version 17, System Protection & Dynamics Inertia TF, 2020 · primary source | 9.4 avian fatalities per GWh (fossil-fuelled power stations) Sovacool, B.K. (2012), 'The avian and wildlife costs of fossil fuels and nuclear power', Journal of Integrative Environmental Sciences 9(4): 255-278, doi 10.1080/1943815X.2012.746993 (abstract via Tethys), 2012 · not re-verified | Mining landscapes dominate: open-cast and mountaintop-removal mines plus ash ponds give coal a land-use intensity of 1,000 ha/TWh/y (Lovering 2022) and 14.9 m2/MWh (UNECE), far above the power station itself; plants have tall stacks and cooling towers. Lovering, J., Swain, M., Blomqvist, L., Hernandez, R.R. (2022), 'Land-use intensity of electricity production and tomorrow's energy landscape', PLOS ONE 17(7): e0270155, 2026 · reputable secondary | Coal ash concentrates heavy metals (arsenic, mercury, lead, selenium) and naturally occurring radionuclides (uranium, thorium, radium) from the coal; it is the largest TENORM waste stream and is mostly landfilled or ponded rather than isolated. World Nuclear Association, 'Radioactive Waste Management' (updated 22 Nov 2025), citing IAEA inventory estimates (published Jan 2022, end-2016), 2025 · not re-verified |
| Battery storage | Dispatchable within its energy limit: lithium-ion systems respond in milliseconds but typical utility systems hold 2-4 hours of output (Lazard: 100 MW / 4 h, 350 cycles/yr), so they shift energy within a day rather than across weather weeks. Lazard, Levelized Cost of Energy+ (LCOE v19.0, LCOS v11.0), July 2026, 2026 · primary source | Converter (inverter)-connected: 'delivers, therefore, in principle no additional inertia to the power system' (ENTSO-E); synthetic inertia or fast frequency response can be programmed but is not inherent. Batteries are the preferred fast-frequency-response resource (ENTSO-E: 'a good technical solution to provide fast frequency response'); grid-forming inverters can emulate inertia. ENTSO-E (16 Dec 2020), 'Inertia and Rate of Change of Frequency (RoCoF)', Version 17, System Protection & Dynamics Inertia TF, 2020 · primary source | n/a | Industrial-scale containers on a few hectares next to substations or solar/wind plants; landscape impact comparable to a small industrial site. No published land-use intensity captured in this task. Lazard, Levelized Cost of Energy+ (LCOE v19.0, LCOS v11.0), July 2026, 2026 · not re-verified | n/a |
What it takes to replace one 1 GW reactor
Energy equivalence only: the wind and solar fleets below produce the same annual electricity as the reactor at the 2024 US fleet capacity factors, not the same firm capacity. Storage is not sized. Every number is a derived row of the dataset; the arithmetic is shown on demand.
Show the arithmetic behind each figure
Nuclear, capacity factor: 7,954.1 GWh per year from 1 GW
Annual output of a 1 GW reactor at the 2024 US fleet capacity factor: 1,000 MW x 8,760 h x 90.8% = 7,954,080 MWh = 7,954 GWh/yr. Energy equivalence only; the replacement fleets below deliver the same annual energy but not the same firm capacity (no storage/firming included).Nuclear Atlas worked example 'replace one 1 GW reactor' (see docs/data/comparison_metrics.md) · USA, 2024Wind, onshore, capacity factor: 2,647.2 MW of onshore wind nameplate for the same annual energy
Nameplate needed = 7,954,080 MWh / (8,760 h x 34.3%) = 2,647 MW, i.e. 2.65 GW of wind per GW of nuclear (ratio = 90.8/34.3 = 2.65). Turbines at 3.4 MW each (average new US turbine 2023, DOE): 2,647 / 3.4 = 779 turbines.Nuclear Atlas worked example 'replace one 1 GW reactor' (see docs/data/comparison_metrics.md) · USA, 2024Wind, onshore, land use: 913.3 km2 total project area (Denholm 2009, 34.5 ha/MW)
2,647 MW x 34.5 ha/MW = 91,329 ha = 913 km2 of project area (std dev +/-22.4 ha/MW gives roughly 320-1,506 km2); direct physical footprint 1.0 ha/MW = 26 km2. Cross-check: NEI 2015 gives 260-360 sq mi = 673-932 km2 for 1,900-2,800 MW of wind.Nuclear Atlas worked example 'replace one 1 GW reactor' (see docs/data/comparison_metrics.md) · USA, 2024Solar PV, capacity factor: 3,913.8 MW(ac) of utility PV nameplate for the same annual energy
Nameplate needed = 7,954,080 MWh / (8,760 h x 23.2%) = 3,914 MWac, i.e. 3.91 GW of PV per GW of nuclear (ratio = 90.8/23.2 = 3.91).Nuclear Atlas worked example 'replace one 1 GW reactor' (see docs/data/comparison_metrics.md) · USA, 2024Solar PV, land use: 109.4 km2 total site area (Ong 2013, 3.4 acres/GWh/yr)
Generation basis: 7,954 GWh/yr x 3.4 acres/GWh/yr = 27,044 acres = 109 km2. Capacity basis: 3,914 MWac x 7.9 acres/MWac = 30,919 acres = 125 km2 (the Ong generation-weighted figure embeds a 2012 sunbelt capacity factor, hence the gap). 2019-built plants are 25-50% denser (Bolinger & Bolinger 2022), so ~82-88 km2 is a fairer modern estimate. Cross-check: NEI 2015 45-75 sq mi = 117-194 km2.Nuclear Atlas worked example 'replace one 1 GW reactor' (see docs/data/comparison_metrics.md) · USA, 2024Nuclear, land use: 3.4 km2 plant site (NEI 2015, 1.3 sq mi per 1,000 MW)
1.3 sq mi x 2.59 km2/sq mi = 3.4 km2 for the plant site (median of 59 US sites). Ratios in this example: wind project area / nuclear site = 271x (direct footprint 8x); PV site / nuclear site = 33x. Upstream uranium mining land is excluded here (Lovering 2022 total LUIE incl. mining: nuclear 7.1 vs wind footprint 130 / spacing 12,000 vs ground PV 2,000 ha/TWh/y).Nuclear Atlas worked example 'replace one 1 GW reactor' (see docs/data/comparison_metrics.md) · USA, 2024Battery storage, lcoe: n/a MWh
null by design: the worked example does not size storage. One day of the reactor's output is 21,792 MWh, i.e. 54 Lazard reference batteries (100 MW / 400 MWh) just to hold one day - and hours-scale lithium-ion does not cover multi-day wind lulls. Published firming method (Lazard): Levelized Firming Cost = Nameplate x (1 - ELCC) x Net CONE x 12 / (Nameplate x regional CF x 8,760); inputs p.34: ELCC solar/wind/storage MISO 39/26/95%, CAISO 14/27/96%, SPP 51/20/56%, PJM 12/38/55%, ERCOT 17/24/93%, NYISO 12/17/81%; Net CONE USD/kW-month MISO 6.49, CAISO 7.34, SPP 7.13, PJM 5.50, ERCOT 11.67, NYISO 5.80 (ELCC data as of May 2026). Lazard's resulting USD/MWh firming adders could not be read reliably from the PDF text, so they are not stored.Nuclear Atlas worked example 'replace one 1 GW reactor' (see docs/data/comparison_metrics.md) · USA, 2024Known gaps of the comparison dataset (11)
- EIA Electric Power Monthly Table 6.07.A/B (2025 full year, 2026 YTD) is only published as xlsx and could not be read in this environment; capacity factors are from the Electric Power Annual (final 2024 data; 2023 in notes).
- IEA/NEA 2020 LCOE figures are taken from the NucNet summary (confidence low) because the IEA PDF text extraction did not yield the numeric tables.
- Lazard's per-region firming costs (USD/MWh) could not be read reliably from the PDF text; only the method (Net CONE x (1-ELCC)) and ELCC/Net CONE inputs are stored (battery_storage derived row and docs). Lazard v19 contains no hydrogen (LCOH) section.
- IPCC AR5 Annex III PDF could not be fetched; Table A.III.2 values come from the Wikipedia reproduction (confidence medium).
- UNECE metal-use columns in the OWID CSV carry no unit in the OWID datapackage and the UNECE PDF could not be fetched; UNECE copper rows are confidence low. Use DOE QTR 2015 (t/TWh) and IEA 2021 (t/MW) for material intensity.
- IEA 2021 critical-mineral intensities are stored as technology totals (t/MW) via the WNA table; the element-level split (copper, nickel, rare earths ...) is only in the IEA chart image/CSV, not captured.
- Lovering et al. 2022 Table 1 is an image; medians were taken from the article text and the authors' Breakthrough Institute summary (interquartile ranges not captured).
- Observed nuclear construction times (median months for reactors grid-connected in 2024) not captured: WNA Performance Report 2025 PDF unreachable. Lazard construction-time rows are modelling assumptions.
- No ramp-rate / start-up-time table by technology was captured; dispatchability rows are qualitative with capacity-factor evidence.
- Battery and hydrogen rows have no land-use, materials or waste statistics; bat mortality from wind and fish mortality from hydro are qualitative only.
- Solar-panel composition statement (waste_toxicity_note) and coal-ash heavy-metal list are from general literature, not re-fetched today (confidence low).