In this paper, a hybrid adaptive observer is designed for a class of nonlinear sampled-data systems with constant unknown parameters. The proposed observer uses a predictor of the output between the sampling times. This predictor is re-initialized at each sampling time. This observer is very simple to implement and converges exponentially under some sufficient conditions. An explicit relation between the bound of the maximum allowable sampling time (τMASP) and the parameters of the observer is also given. Copyright © 2012 John Wiley & Sons, Ltd.