The paper presents a speed adaptive observer for the induction motor (IM) drive that can be used in a sensorless implementation. The observer is based on the model of the IM in the stationary reference frame: it uses linear feedback terms to estimate the fluxes and an adaptive law to estimate the speed. The feedback gains are designed using Lyapunov's nonlinear stability method. The investigation finds that the initial design only performs in limited speed range - as the speed increases, the speed estimate of the observer becomes inaccurate. Based on the initial design, the paper develops a hybrid observer: this uses a speed signal as an additional input (this input speed signal is relatively inaccurate - it can be an initial estimate of the motor speed or an estimate of the synchronous speed). It is shown that the hybrid design can be tuned to perform well in wide speed range: it yields relatively accurate estimates of the fluxes and speed. The theoretical developments are supported with simulations and experiments.