Capacity factor and the economics of a reactor

The capacity factor, or load factor, is the energy a reactor delivered over a period divided by what it would have delivered at rated power the whole time. A 1,000 MWe reactor that produces 8 TWh in a year has a capacity factor of 91%. The only losses are refueling and maintenance outages, 18 to 35 days a year, and unplanned trips. The world average was 81.5% in 2023; the US fleet has exceeded 90% for years, and the best units top 95%. Compare that with 35 to 42% for coal in the United States, 25 to 45% for wind and 10 to 29% for solar.

That number is no accident but economics. The cost of a nuclear plant is almost all in construction: capital is at least 60% of the levelized cost, and fuel only 15 to 20%. Once built, each additional kilowatt-hour costs very little, so the rational thing is to produce as many as possible: that is what running baseload means. With 60-year design lives, every extra year of operation spreads the initial investment over more energy, which is why life extensions are the cheapest electricity in the system.

The levelized cost of electricity, LCOE, adds up all the investment, operation and fuel over the plant's life and divides it by all the energy it produces, discounting money over time. Since the money is spent up front and the energy comes later, the discount rate dominates: in the OECD comparison, the same nuclear plant costs between 27 and 61 dollars per MWh at a 3% rate and between 56 and 146 at 10%. The other two factors are the construction cost per kilowatt, from 2,157 dollars in South Korea to 7,821 in the United States in 2023, and construction time: every year of delay is interest on billions that are not yet producing.

Construction time is where projects are won or lost. A gas plant is built in about two years; a nuclear one in over five at best. Korea, China and the Emirates build in five to seven years because they repeat the same design with crews that have done it before; the first-of-a-kind units in the West, Olkiluoto 3, Flamanville 3, Vogtle 3 and 4, took between 10 and 18 years and doubled or tripled their budgets. The SMR bet is precisely to shorten and standardize the build.