A method of determining power losses in magnetic journal bearings is presented that uses temperature measurements to solve the inverse heat transfer problem. ntis is different from conventional methods, which are based on torque or speed measurements, and is the first such application to magnetic bearings. The method is benchmarked against available theoretical results for windage losses and ohmic losses in the magnet wire. An error analysis based on Monte Carlo simulations is presented. Power loss results are presented for two different bearing designs, one constructed of silicon iron material and another constructed of cobalt iron material. Magnetic journal core losses are determined for surface speeds up to 3.1 million DN (mm-rpm) and magnetic flux densities approaching saturation for each bearing. To the authors' knowledge, these are the first power loss results for cobalt iron magnetic bearings at these test conditions.