Axial flux permanent magnet (AFPM) machines are promising for hybrid electric vehicles (HEVs) due to the compactness, high torque density and high efficiency. However, poor thermal characterization leads to an over-sizing of these machines which ultimately compromises overall system efficiency. In this paper, the transient thermal behaviour of all the components in the single sided AFPM machine are characterized in an accurate but computationally efficient lumped parameter thermal model (LPTM). For the first time, contact measurements on the rotor have been used in AFPM machines to demonstrate the ability of the model to predict all component temperatures to within 4 °C for steady state. The mean temperature error over a load step transient was less than 5°C with a maximum error less than 13.5 °C which was for the winding. The model has a running time of approximately 1000 times faster than real time on a desktop machine and is suitable for integration into system simulation tools and predictive control strategies to avoid over-sizing of the motor and improve the usage of the electric machine in dynamic duty cycles.