Nuclear Power

Nuclear power produced about 2,800 TWh in 2025, close to 9% of the world’s electricity and, after hydropower, the second-largest source of low-carbon generation. The United States, China and France account for well over half of it. The IAEA’s Power Reactor Information System lists roughly 410–440 operable reactors with a combined net capacity of around 375–400 GWe; the exact count depends on how idled Japanese units are treated. Once built, a reactor is mature, very reliable technology. The trouble lies elsewhere: cost, construction risk, the politics of waste and public fear of accidents.

How it works

Think of a nuclear plant as a thermal power station with an unusual heat source. In the core, nuclear fission splits uranium-235 (and, as fuel burns, plutonium-239) nuclei after they absorb a slow neutron. Each fission releases about 200 MeV, mostly as kinetic energy of the fragments, which becomes heat in the fuel, along with two or three new neutrons. A moderator (ordinary water, heavy water or graphite) slows those neutrons so the chain reaction keeps going. Control rods and dissolved boron absorb neutrons to hold the reaction exactly critical.

A coolant carries heat out of the core and makes steam, which drives a turbine. A condenser then rejects the remaining heat to a river, the sea or a cooling tower. Because light-water reactors run at modest steam conditions (about 280–330 °C), thermal efficiency is lower than in a modern coal or gas plant:

What really sets nuclear apart is the energy packed into its fuel. One kilogram of low-enriched uranium burned to 50 GWd/t releases roughly 4 million MJ of heat, compared with about 25–30 MJ for a kilogram of hard coal. The fuel energy content calculator shows the gap against fossil fuels.

The fuel cycle

Natural uranium contains 0.7% U-235. Most reactors need it enriched to 3–5% U-235 in gas centrifuges. Some advanced designs need HALEU at up to 20%, a fuel Russia has dominated until recently. The chain goes from mining and milling to conversion to UF₆, enrichment, fuel fabrication, irradiation for 4–6 years to a burnup of about 45–60 GWd/t, then storage of spent fuel. France, Russia and Japan reprocess some spent fuel to recover plutonium for MOX fuel. Most countries, including the US, use a once-through cycle.

Types of reactors

Pressurised water reactors keep primary water at about 15.5 MPa so it does not boil. They pass heat through steam generators to a separate secondary loop. They dominate because the design came out of naval propulsion and has the longest licensing record. Boiling water reactors simplify the plant by boiling water in the core and sending that steam straight to the turbine. CANDU-type PHWRs use heavy water as moderator, so they can burn natural uranium and refuel at full power. Canada, India and Romania (Cernavodă) picked them for exactly that reason.

The UK’s AGRs reach higher steam temperatures with CO₂ coolant, but graphite cracking is closing them through the late 2020s. RBMK reactors, the Chernobyl design, have a positive void coefficient at low power, and post-1986 modifications reduced it. Russia is replacing the units still running. Fast reactors run without a moderator and can breed fissile plutonium from U-238, using uranium far more completely. Russia’s BN-800 is the largest operating example and China is building CFR-600 units.

Small modular reactors

SMRs are reactors of up to about 300 MWe, intended to move construction from site to factory. Instead of a few giant megaprojects, the bet is on many repeatable units. Operating examples include Russia’s floating Akademik Lomonosov (2 × 35 MWe, since 2020) and China’s HTR-PM high-temperature gas-cooled demonstrator (commercial operation in late 2023). Ontario’s Darlington BWRX-300 and China’s ACP100 (Linglong One) are under construction. The NuScale-UAMPS project in Idaho was cancelled in 2023 after its estimated power price rose to about $89/MWh even after subsidies. Small does not automatically mean cheap.

Efficiency and capacity factor

A nuclear plant turns only a third of its heat into electricity, but it hardly ever stops. Refuelling outages every 18–24 months take a few weeks. The US fleet has posted a capacity factor of about 90–93% every year for two decades (EIA), the highest of any technology. The global average is closer to 80%. France deliberately runs its reactors in load-following mode, and its fleet suffered a stress-corrosion outage crisis in 2022 that cut output to a 30-year low. Many plants have also gained 5–20% of output through capacity uprates. You can compare these figures with wind and solar in the capacity factor calculator.

Costs

Nuclear economics come down to capital cost and financing. Fuel is only about 10–20% of generation cost, and uranium price spikes barely move the cost of electricity. Published estimates are far apart:

• IEA/NEA Projected Costs of Generating Electricity 2020: new-build nuclear median LCOE of about $69/MWh at a 7% discount rate. Costs rise sharply at 10% and fall to roughly $40/MWh at 3%. Long-term operation of existing reactors after refurbishment comes in at about $30–50/MWh, among the cheapest low-carbon power available.
• Lazard LCOE+ 2024 (US, unsubsidised): $142–222/MWh, anchored on Vogtle’s actual costs.
• NREL ATB and EIA AEO sit between these, with overnight costs of roughly $7,000–9,000/kW for large US reactors.

The spread reflects what actually happened on construction sites. Recent large projects:

The cheaper builds share four things: a continuous build programme, a standardised design, an experienced supply chain and a stable regulator. South Korea and China have had all four, and so did France in the 1980s. First-of-a-kind builds in countries that hadn’t built a reactor in decades had none. Because so much capital is tied up before any revenue, the discount rate and WACC assumed largely decide the LCOE. The LCOE calculator shows that sensitivity directly.

Land, water and materials

Nuclear has the smallest direct land footprint of any major source, often under 1 km² per gigawatt for the plant itself. Mining and exclusion zones add a little. Water is the tighter limit. Using NREL’s survey (Macknick et al., 2012), plants with cooling towers consume roughly 600–800 US gallons per MWh, more than coal or gas because of their lower efficiency. Once-through cooled plants withdraw tens of thousands of gallons per MWh but return almost all of it, slightly warmer. During heatwaves, river-cooled plants in France and elsewhere have had to curtail output to respect discharge temperature limits.

Material intensity per MWh is low over a 60–80-year life, though plants use large amounts of concrete and steel. A typical light-water reactor generates roughly 20–30 tonnes of spent fuel per gigawatt-year. The US Department of Energy notes that all the commercial spent fuel the US has produced since the 1950s would fit on a single football field at a depth of less than 10 yards. Spent fuel sits in pools, then in dry casks. Finland’s Onkalo repository at Olkiluoto is the first deep geological disposal facility for spent fuel to approach operation, with Sweden’s Forsmark site and France’s Cigéo behind it.

Lifecycle CO2 emissions

Nuclear plants emit no CO₂ during operation. Their lifecycle emissions come from mining, enrichment (now mostly centrifuge-based and much less energy-intensive than the old gaseous diffusion plants), construction and decommissioning.

• IPCC AR5 WGIII Annex III (2014): median 12 gCO2e/kWh, range 3.7–110.
• UNECE 2021: 5.1–6.4 gCO2e/kWh for a European PWR, the lowest of all options assessed.

These figures are comparable with onshore wind power and below solar power. They are roughly 1–2% of the figure for coal.

Safety and health impacts

Per unit of electricity, nuclear ranks among the safest sources. Markandya & Wilkinson (Lancet, 2007) estimated about 0.07 deaths per TWh for European nuclear, including occupational and public health effects. Our World in Data’s updated synthesis puts it at about 0.03 deaths per TWh, including Chernobyl and Fukushima. Coal comes in at roughly 24.6, oil at 18.4 and gas at 2.8, mostly from air pollution.

The two major accidents need context:

• Chernobyl (1986): an RBMK reactor during a badly run test. About 30 workers and firefighters died of acute radiation syndrome or the explosion. WHO and UNSCEAR projected up to about 4,000 eventual cancer deaths among the most exposed populations. Wider estimates using linear no-threshold extrapolation are higher and contested. Thousands of thyroid cancers in children were linked to contaminated milk, and almost all were treatable.
• Fukushima Daiichi (2011): three BWR core meltdowns after the tsunami knocked out backup power. Japan has attributed one worker death to radiation. More than 2,000 elderly and ill residents died in connection with the evacuation. A badly calibrated emergency response can do more harm than the radiation itself.

Post-Fukushima stress tests added portable backup equipment, filtered venting and flood protection across fleets. Newer Generation III+ designs (AP1000, EPR, APR-1400, VVER-1200, Hualong One) add passive cooling and core catchers.

Outlook

Nuclear’s share of world electricity has fallen from about 17% in the late 1990s to under 9%. Germany closed its last reactors in 2023, and many older Western units have retired for economic reasons. That trend has partly turned. China has around 30 reactors under construction, more than the rest of the world combined, and builds them in about 5–7 years. At COP28 more than 20 countries pledged to triple global nuclear capacity by 2050. The US, Japan, the Netherlands, Sweden, Poland and Belgium have moved to extend lifetimes or plan new build. Data-centre operators have signed long-term power purchase deals for existing reactors and future SMRs.

What holds nuclear back now is construction cost and schedule outside China, Korea and Russia, the HALEU fuel supply chain for advanced reactors, and the long-unresolved question of waste repositories outside Finland and Sweden. In decarbonised grids heavy on variable renewables, nuclear supplies firm, dispatchable low-carbon capacity. Whether that is worth the price depends on local alternatives such as battery storage, interconnectors and hydro. The technology comparison page sets those trade-offs side by side, and the power plant map shows where the world’s reactors are.