In order to accurately evaluate the reliability of the servo drive unit of a computer numerical control (CNC) system, a reliability modeling method with multi-stage degradation data fusion was proposed. Firstly, by analyzing multiple degradation process models, random effects were introduced, and a reliability model building scheme considering individual differences was given for the characteristics of individual differences existing between different servo drive units. Then, a Bayesian approach was adopted to fuse different degradation data by considering multiple degradation processes, and a reliability model of servo drive units with multi-stage degradation data fusion was established. Moreover, a Markov chain Monte Carlo (MCMC) method was used to complete the model parameter estimation. Finally, a servo drive unit loading test platform was built in the laboratory environment to collect experimental data and verify the validity of the model.