This article investigates the design of accelerated life testing plans under progressive censoring when test units experience competing failure modes and are subjected to either single or multiple stress types. The optimal test plan results in failure data at accelerated conditions that can then be used to obtain accurate reliability prediction at normal conditions. The new test plan criterion is based on the minimization of the asymptotic variance of the mean time of first failure and meets practical constraints. Numerical examples based on parameters from a real test are presented to illustrate the application of the proposed method. © 2013 Taylor & Francis Group, LLC.