Nine-Switch Inverter (NSI) is composed of two conventional inverters with three common switches. Two sets of three phase ac loads can be connected to the outputs of NSI and independently controlled without any undesirable interaction. In conventional multi-step Model Predictive Control (MPC) of a nine-switch inverter, the prediction steps of both load stages must be equal. Unfortunately, this results in losing the freedom of selecting independent prediction horizon for individual loads. To overcome this problem, a novel asymmetrical multi-step direct model predictive control method is presented in this paper. The proposed method finds two independent optimum solutions for each load to match with their utilization profile. It is assumed that two individual loads are controlled by two separate virtual inverters, and two separate model predictive control problems with their own prediction steps are solved to identify optimum control actions. The control calculations are performed in a Cyclone IV Field Programmable Gate Array (FPGA) by using a pipelined architecture. The system stability is analyzed using Lyapunov stability method. To highlight the effectiveness of introduced strategy, mathematical proof for controlling two separate loads with an asymmetrical prediction step is validated in experimentally. INDEX TERMS Model predictive control, nine-switch inverter, FPGA, digital control, dual-output inverter systems.