High-frequency voltages across the steel ball bearings and the corresponding currents can cause premature bearing failures in electric machines driven by PWM converters. The bearing voltage, one of the most commonly-used failure indicators, depends heavily on the bearing capacitance. This paper presents a combined numerical and experimental approach for the calculation of ball bearing capacitances to address the uncertainty introduced by lubricant property, lubrication status and other metal parts, such as seals and ball retainers. Based on the obtained capacitance breakdown, the influences of temperature, speed and bearing load (radial, axial or combined) on the capacitance are studied. Measurements and associated results of bearing capacitances are provided to validate the proposed method.