<p>This paper is about the decentralized control for discrete-time Markov jump linear system (MJLS) with asymmetric information, in which only controller 2 transmits state information to controller 1. Because of the asymmetry in MJLS, the estimation error covariance (EEC) couples with controller 1, which means that the separation principle fails. The backward and forward stochastic difference equation is obtained by Pontryagin’s maximum principle. According to the linear optimal controllers and the EEC, controller 1 decouples from the EEC. This operation can make the control gain and the estimation gain independent of each other. To simultaneously solve the backward Riccati equation for the controller gain and the forward Riccati equation for the estimator gain, iterative solutions to the Riccati equations are proposed, and a suboptimal solution for this decentralized system in MJLS is provided. It is demonstrated that the MJLS under decentralized control is bounded in the mean-square sense as time approaches infinity.</p>