The data analysis problem of step-stress accelerated life testing with exponential distribution is discussed. At the step-stress accelerated life testing conditions, an approximate E-Bayesian parameter estimation of step-stress accelerated life testing with exponential distribution is given by considering the prior distributions of the hyperparameters and using Gibbs sampling method. Finally, a simulation example is given, the results show that the Gibbs sampling method is simple and the convergence is better. E-Bayesian parameter estimation is more effective than the maximum likelihood estimation.