In the next 50 yr, the world will need to develop hundreds of gigawatts of non-fossil-fuel energy sources for production of electricity and fuels. Nuclear fusion can probably provide much of the required energy economically, if large single-unit power plants are acceptable. Large power plants are more common than most people realize: There are already many multiple-unit power plants producing 2 to 5 GW(electric) at a single site. The cost of electricity (COE) from fusion energy is predicted to scale as COE ≈ COE0(P/P0)−n, where P is the electrical power, the subscript zero denotes reference values, and the exponent n ≈ 0.36 to 0.7 in various designs. The validity ranges of these scalings are limited and need to be extended by future work. The fusion power economy of scale derives from four inter-related effects: improved operations and maintenance costs; scaling of equipment unit costs; a geometric effect that increases the mass power density; and reduction of the recirculating power fraction. Increased plasma size also relaxes the required confinement parameters: For the same neutron wall loading, larger tokamaks can use lower magnetic fields. Fossil-fuel power plants have a weaker economy of scale than fusion because the fuel costs constitute much of their COE. Solar and wind power plants consist of many small units, so they have little economy of scale. Fission power plants have a strong economy of scale but are unable to exploit it because the maximum unit size is limited by safety concerns. Large, steady-state fusion reactors generating 3 to 6 G W(electric) may be able to produce electricity for 4 to 5 ¢/kW·h, which would be competitive with other future energy sources.