This paper studies the stability of state estimation for a discrete-time linear stochastic system, the states of which are measured by multiple sensors and transmitted over multiple wireless channels. Random packet loss process introduced by each wireless channel is modeled by an independent and identically distributed (i.i.d.) Bernoulli process. The estimation strategy designed in this paper is based on Covariance Intersection fusion of local state estimates of the observable subsystem of each sensor. The sufficient conditions, imposing constraint on the packet success probability of each channel, are established by taking into account each observable subsystem structure to guarantee the expectation of the trace of estimation error covariance matrices is exponentially bounded, and the upper bound is given. Simulation examples are provided to demonstrate the effectiveness of the results.