This paper proposes a partially decentralized state observer that can be implemented on network-based parallel computers (multicomputers). The state estimation is decentralized in the sense of processing only local measurements, but is centralized in the sense that the interaction terms are kept in the model simulation. The state estimator is coupled with a fully decentralized state feedback to give rise to the overall model-based controller. The proposed implementation involves parallel simulation of the process model using a Jacobi-like iteration scheme. This is important in reducing the computational effort in model-based control. The proposed scheme is based on partitioning of the overall dynamic system into a number of loosely-coupled interconnected subsystems of smaller dimension and then designing a local observer for each subsystem, taking into account the interaction effects among the subsystems. A new state observer design methodology is addressed that accounts for the parallel nature of the implementation and also guarantees stability and optimal performance of the parallel observer. Important issues that are studied include the effect of observer gains on convergence of the state estimation and the computational speedup that can be attained with a different number of computer nodes (processors). Simulation results on a message-passing multicomputer for a class of chemical engineering applications demonstrate the potential of parallel processing for state estimation in the context of model-based control.