The hydrolysis of ATP that accompanies actin polymerization occurs on the F-actin subsequent to the elongation step. For Mg-actin, the rate of ATP hydrolysis is similar to the rate of elongation at low concentrations of G-actin but increases more slowly as the G-actin concentration is increased. This behavior can be quantitatively modeled by assuming that ATP hydrolysis occurs predominantly, but not exclusively, on a single subunit of Mg-F-actin at the interface between an ATP-subunit cap and an ADP-subunit core. The rates of elongation of Ca-actin and Mg-actin are similar but the rate of ATP hydrolysis on Ca-F-actin is appreciably slower than the rate of elongation at all concentrations of Ca-G-actin. The data for Ca-actin can be modeled by assuming that ATP hydrolysis occurs essentially randomly on Ca-F-actin within a large ATP cap which can be as long as 2,000 subunits in a 10,000-subunit long filament.