<p>For stochastic nonlinear systems with input delay and unmeasured states, the existing control schemes cannot be directly applied to their control problems. Proposed here is a brand-new adaptive output feedback control scheme based on the multi-dimensional Taylor network (MTN) that approximates the unknown nonlinear function. The MTN-based state observer is constructed to estimate the unmeasured states, and an appropriate auxiliary system with the same order as that of the controlled system is introduced to compensate for the influence of the input delay. Using the designed state observer, the MTN-based adaptive output feedback control strategy is proposed via the backstepping technique to ensure that all the signals of the closed-loop system remain bounded in probability, the output signal tracks the reference signal successfully, and the tracking error is bounded by the expected bound. Two simulation examples are provided to verify the validity of the developed controller.</p>