When the system states are unmeasurable, the problem of decentralized robust predictive control with output feedback is studied for the uncertain singular large-scale system with interconnected subsystems. By constructing the Lyapunov function, and with the method of LMIs, the “min-max” optimization problems are converted into convex programming problems. Then the sufficient conditions for the existence of the decentralized robust output feedback predictive controller are derived. The stability of the closed-loop singular large-scale systems is guaranteed by the initial feasible solutions of the optimization problems. The regular and the impulse-free of singular systems are also hold. Simulation results illustrate the effectiveness of the proposed method.