The electromagnetic bearing is a core component of integrated compressor units, and its reliability directly affects the operational stability. To address the challenges of insufficient failure samples, difficulties in reliability assessment and inadequate risk quantification for electromagnetic bearings of the high-power integrated compressor unit, this study proposes a state analysis model for electromagnetic bearings based on failure mode and effects analysis (FMEA) and Bayesian network (BN). By integrating prior knowledge with observed data, the model can effectively handle the incomplete and uncertain information, identifying 21 potential failure modes of electromagnetic bearings and their impacts. Based on the assessment results of severity, occurrence and detectability, the failure modes were ranked by risk, and the high-risk failures such as bearing collapse, frictional wear, power supply faults and bearing corrosion were analyzed. The results show that risk values of auxiliary bearing collapse, friction loss, power failure, bearing corrosion, main control board failure, signal transmission failure, etc. are relatively high. The corresponding optimization and improvement measures proposed based on the characteristics of failure modes can reduce the probability and impact of failures, and lower the research and development costs.