This work proves that inferences on parameter vectors based on moment inequalities typically used in linear models with outcome censoring are sharp, i.e., they exhaust all the information in the data and the model. This holds for fixed and randomly censored linear models under median...