Without the hype
The Carbon Cost of Stopping Nuclear Power
Britain dropped three planned nuclear reactors between 1989 and 1995, after investors refused to take them on and a government review found them uneconomic. Greenpeace had funded one of the main objectors at the public inquiry into the first. Germany shut its last reactor in April 2023, 23 years after a government that included the Greens agreed to end nuclear power, and 12 years after the Fukushima accident brought the date forward. This dashboard counts the extra CO2 from replacing the output of 48 reactors in eight countries that closed early or never opened, in decisions to which organised opposition contributed. It assumes that coal and gas plants made the power instead. The reactors’ figures come from the International Atomic Energy Agency’s (IAEA) reactor records, and the CO2 from each country’s fossil power from Ember. It then compares the total with the CO2 from burning the oil and gas of the world’s biggest producers, from the Carbon Majors database.
Each decision had other causes too, such as costs, investors, courts or accidents, and its description below gives them. The percentage in the sentence below sets how much of each decision you put down to campaigners. At 100%, the starting setting, the total is the most the campaigns could account for at these reactors, not a measure of their effect. What the figures leave out lists the other limits.
Loading the reactor data…
Give each reactor a life of 60 years at its average output, and put of each decision down to campaigners. The reactors' power came instead from:
Drag an underlined number left or right, or click it and use the arrow keys; a double click puts it back. The charts and headline numbers follow all three choices. The figures for Britain and Germany in the text follow the life and the fuel, but give all the extra CO2, whatever the share.
Where the extra CO2 comes from
Germany accounts for of the total. Each click opens a bar into its parts, from country to decision to reactor to single years. Back undoes one click, and the names above the chart take you straight back to an earlier step.
Extra CO2, million tonnes (Mt)Extra CO2 each year
A reactor that closed part of the way through a year counts only for the rest of that year.
Extra CO2, million tonnes (Mt)Extra CO2 compared with burning each producer’s oil and gas
Carbon Majors estimates how much oil and gas each big producer takes out of the ground each year. By Carbon Majors’ own factors, burning the oil and gas of the 82 companies it covers gave off 15.7 billion tonnes of CO2 in 2022. Replacing the stopped reactors’ output gave off an estimated 170 million tonnes of extra CO2 that year: about as much as burning Sinopec‘s oil and gas (171 million tonnes), and of the CO2 from all the companies’ oil and gas.
A company’s figure is the CO2 from its customers burning its fuel; What the figures leave out explains why you can’t add the two together.
CO2, million tonnes (Mt)The chart ranks companies by the CO2 from burning their oil and gas in the period chosen above. 2022 is the latest year in the Carbon Majors data published openly. "Investor-owned" companies belong to their shareholders; "state-owned" ones to governments. Source: Carbon Majors (InfluenceMap), production by company, times its combustion factors.
How Britain and Germany stopped their reactors
Britain’s three cancelled reactors
The Central Electricity Generating Board (CEGB) planned Sizewell B, Britain’s only pressurised water reactor (PWR), as the first of four identical stations. Investors would not take the nuclear stations when the government privatised the electricity industry, so in November 1989 the government kept them out of the sale and stopped new nuclear building until a review in 1994. Hinkley Point C still won planning consent in 1990, after a 14-month inquiry at which the Stop Hinkley Expansion campaign, funded by Greenpeace, was among the main objectors and 22,000 people lodged objections. The review, published in 1995, found Hinkley Point C, Wylfa B and Sizewell C uneconomic.
Had the three gone ahead as copies of Sizewell B, starting in 1999, 2001 and 2003, they would have made 659 TWh of electricity by 2025 and saved 413 million tonnes of CO2 from coal and gas plants. The new Hinkley Point C, two 1,630 MW reactors that Électricité de France (EDF) is building on the same site, is due to start around 2030.
Germany’s phase-out
Germany stopped its reactors in three steps: a court closed Mülheim-Kärlich in 1988 on a case brought by neighbours; the Social Democrat (SPD) and Green government agreed with the operators in 2000 to end nuclear power; and after Fukushima in 2011 Angela Merkel’s government reversed its own extension of the reactors’ lives and closed the rest by April 2023. Together they account for 1.3 billion tonnes of extra CO2 by 2025.
For 2012 to 2017 this method gives 55 TWh of lost output and 46 million tonnes of extra CO2 a year from the 2011 decision. Jarvis, Deschenes and Jha, using hourly records from every German power station, found 53 TWh and 36 million tonnes a year, with more than 1,100 extra deaths a year from air pollution. Their CO2 counts only German power stations’ own emissions: imports filled part of the gap, and Ember’s figures used here include the CO2 from mining coal and producing gas.
What the figures leave out
- The lost output is an estimate. The model assumes how long each reactor would have run. Those never finished have no record at all, so their start dates and output are my assumptions, and those that did run might have closed early anyway for repairs, costs or an accident. Set a 40-year life in the sentence at the top and the total falls by about a fifth.
- Campaigners were one cause among several. Each decision’s description gives the others.
- The model assumes fossil plants filled the whole gap. It takes the average CO2 of each country’s fossil plants, not the particular plants that ran longer, and assumes the reactors would not have pushed out any wind or solar. Imports, lower demand or renewables built in their place would all make the extra CO2 smaller.
- The list leaves cases out. It covers reactors in eight countries where opposition clearly helped stop them, and leaves out the cases listed under How it works. It doesn’t count the cost of delays, of stricter rules that made later reactors dearer, or of reactors never proposed because of opposition, so it leaves out some of the extra CO2 that opposition may have added.
- It counts only CO2. Campaigners opposed nuclear power over accidents, waste, cost and weapons. This page weighs none of those, nor the air pollution from the coal that ran instead.
- Don’t add the company figures to the reactor figures. The two sets of figures count different things: Ember’s CO2 per kWh includes mining, drilling and gas leaks, while the companies’ figures count only burning their oil and gas. Some of the gas burnt in place of the reactors came from these same companies, so adding the two together would count some CO2 twice.
Figures for every reactor
All 48 reactors, with their figures for the choices above. Click a row for the reactor’s details, the decision behind it and its sources.
How it works
What counts. A reactor is in if organised opposition clearly contributed to the decision that closed it early or stopped it before it ran. The opposition might be campaign groups, a referendum, a lawsuit by neighbours, or a party that grew out of the anti-nuclear movement. Left out:
- Wyhl in Germany, where protesters occupied the building site in 1975: its reactor parts went into Philippsburg 2, so Germany lost no output.
- Kalkar, a prototype fast-breeder reactor (one designed to make more fuel than it burns) whose output is too uncertain to estimate.
- Switzerland’s Kaiseraugst, whose power would have replaced imports.
- Spain’s 1984 moratorium, where the Basque separatist group ETA’s attacks on Lemóniz make the campaigners’ part impossible to separate.
- Japan’s reactors idle since 2011, which the regulator’s new safety reviews, court orders and local objections all held back.
- Diablo Canyon in California, whose closure, agreed with campaign groups in 2016, California’s legislature reversed in 2022.
- American reactors closed on cost alone.
- Delays, such as Sizewell B’s long public inquiry or Greenpeace’s 2007 High Court win against the government’s consultation on new nuclear power.
Lost output. Each reactor’s net capacity, start and closing dates and lifetime load factor (its output as a share of what it would make running flat out) come from the IAEA’s Power Reactor Information System (PRIS). The model runs each reactor from its start for the number of years chosen above, so a reactor that closed after 36 years of a 60-year life loses 24 years of output, less whatever falls after 2025. A reactor that never ran, or ran for only months, takes an 80% load factor; the British PWRs take Sizewell B’s own, 84%. Start dates for those never finished are my assumptions: Hinkley Point C in 1999 (it won consent in September 1990, and Sizewell B took eight years from consent to the grid), Wylfa B and Sizewell C two and four years later, Montalto di Castro in 1991 and 1992, Lungmen in 2016 and 2018 (a year after the dates that Taipower, Taiwan’s state power company, gave in January 2014), Shoreham in 1987 (two years after its low-power tests) and Zwentendorf in 1979. Some reactors stood idle for years before their formal closure (Caorso from 1986, Chinshan 1 from 2014); I count their lost output only from the formal date, at a lifetime load factor that includes the idle years, so their figures are on the low side.
What it replaced. I multiply each year’s lost output by the CO2 per kWh of the country’s fossil power that year, from Ember’s yearly data, less 5 g per kWh for the reactor’s own construction and fuel (the UN Economic Commission for Europe’s 2021 life-cycle estimate for nuclear power is 5 to 6 g). Ember’s figures are lifecycle ones: they include mining, drilling and gas leaks as well as the chimney. Where Ember’s data starts later than the loss (1985 for Britain and Taiwan, 1990 for most of Europe, 2000 for the United States), its first year stands in for the years before. Three exceptions: Barsebäck takes Denmark’s fossil mix, since it faced Copenhagen and imports from Danish and German plants replaced most of its output; the American reactors take the US figure for gas, since New York and California made up the gap with gas (and Long Island, in the 1980s, with oil); and I leave out Germany’s 1985 to 1989 data, which covers only gas. The choices “Gas” and “Coal” put every country on its gas or coal figure instead, to show the low and high ends. The model assumes the reactors’ output would have replaced fossil power one for one: that countries would have built the same wind and solar farms anyway, and that demand would have been the same.
Campaigners’ share. You choose it: it scales the CO2 on the charts and in the headline numbers, not the lost output.
The companies. Carbon Majors (InfluenceMap) records how much oil, natural gas liquids and gas each producer took out of the ground each year. The CO2 from burning it is that production times Carbon Majors’ combustion factors, which already allow for oil and gas made into plastics and other products: 0.371 tonnes of CO2 per barrel of oil and 53,400 tonnes per billion cubic feet of gas. Company climate reports call this CO2 from customers using the products “scope 3, category 11” emissions; Carbon Majors counts it by what each company produces, not what it sells. It leaves out flaring, venting and methane leaks at the wells, which by Carbon Majors’ factors add about 7% for oil and 36% for gas. The open copy of the data, republished by TidyTuesday in 2024, runs to 2022; carbonmajors.org now asks for a login for its newer files. The comparison uses the years both sets of data cover, 1979 to 2022.
Sources. IAEA PRIS (reactors); Ember, yearly electricity data, CC BY 4.0 (CO2 per kWh); Carbon Majors (company production); Jarvis, Deschenes and Jha (2022), The private and external costs of Germany’s nuclear phase-out, Journal of the European Economic Association. Each decision’s own sources are in its description. Data downloaded 14 November 2026.