The system studied consists of two parallel factories subject to production rate-dependent failure rates. The factories are subject to random non-operational periods considered herein as governed by a failure/repair process, and respond to a single product type demand. The main objective here is to propose production policies that will minimize the total cost (inventory and backlog costs), over an infinite planning horizon. The failure rate of the first factory depends on its production rate, while that of the second factory is constant. The proposed model is based on a non-homogeneous Markov decision process, and the stochastic dynamic programming approach is used to obtain optimality conditions.