In this paper, a novel control scheme, known as multi-step model predictive control (MPC), is presented for multi-input systems. To reduce computational complexity, only one or several control inputs are optimized at each time interval. A multi-step MPC scheme for nominal multi-input systems is presented. A set invariance condition for polytopic uncertain systems is identified, and the invariant set is determined by solving a linear matrix inequality optimisation problem. Based on the set invariance condition, a min-max multi-step robust MPC algorithm is proposed for the polytopic uncertain systems. Two examples are carried out to demonstrate the effectiveness of the proposed algorithms.