In this article we consider the analysis of progressively censored competing risks data obtained from a simple step‐stress experiment. It is assumed that there are only two competing causes of failures at each stress level and the lifetime distribution of each one of them is one parameter exponential distribution. Based on the cumulative exposure model assumption, the conditional maximum likelihood estimators (MLEs) of the unknown parameters can be obtained in explicit forms. Confidence intervals of the unknown parameters based on the exact distributions of the conditional MLEs and percentile bootstrap method, are constructed. Further we obtain Bayes estimates and the associated credible intervals based on a very flexible Beta‐gamma prior on the unknown parameters. A simulation experiment has been performed to observe the performances of the different estimators.