For switched stochastic nonlinear systems with time delay, an adaptive output-feedback tracking control problem is studied by using the dynamic surface control (DSC) method associated with a switched state observer. The time-varying delay terms are handled by constructing an appropriate stochastic Lyapunov–Krasovskii functional. The unknown nonlinearities in the considered systems are adaptively approximated by the function approximation approach—neural networks (NNs). It is shown that, under the average dwell time (ADT) method, the semiglobal uniformly ultimately boundedness (SGUUB) is guaranteed for all closed-loop signals by the developed switched controllers of different subsystems even in the presence of time delay, meanwhile, the tracking error is confined to a small residual set. A practical system example and a numerical example are provided to demonstrate the efficacy of the proposed scheme.