Performance of a digital current controller for a PWM inverter is degraded by changes of load impedances and delay necessary for calculation of the controller.This paper presents a digital current controller for a PWM inverter using a neural network. The current controller automatically tunes its gains to the optimal ones for varying load impedances and compensates the calculation delay. Stability condition for the learning rate of the neural network is analyzed. The feasibility of the new controller is verified by simulations and experiments. Stable region of the current controller with deviations of the load impedances is obtained by simulations.