This paper presents an experimental implementation of a predictive control strategy for a system made up of two induction machines fed by a six-leg indirect matrix converter. The objective is to evaluate the practical feasibility of this method, considering the large computational cost, and to incorporate available market technologies. This strategy achieves tracking of torque, speed, and flux references on both machines, minimizing the instantaneous input reactive power present in the system. A theoretical explanation of the main concepts that are used in the predictive control strategy is presented, along with comments on the experimental implementation, experimental setup, and results obtained under both balanced and unbalanced grid conditions.