We study a predictive control formulation for uncertain discrete-time non-linear uniformly continuous plant models where controller output data is transmitted over an unreliable communication channel. The channel introduces Markovian data-loss and does not provide acknowledgments of receipt. To achieve robustness with respect to dropouts, at every sampling instant the controller transmits packets of data. These contain possible control inputs for a finite number of future time instants, and minimize a finite horizon cost function. At the actuator side, received packets are buffered, providing the plant inputs. Within this context, we adopt a stochastic Lyapunov function approach to establish stability results of the networked control system. A distinguishing aspect of this work is that it considers situations where the maximum number of consecutive packet dropouts has unbounded support. (C) 2012 Elsevier Ltd. All rights reserved.