A power device with self-health status prediction function and a self-health status prediction method thereof and a cloud server suitable for a plurality of power devices are provided. The power device is configured to obtain a health index prediction model and a health prediction index threshold of the power device from the cloud server through a communication module thereof, and is configured to put a plurality of obtained sensing data and basic information of the power device into the health index prediction model to obtain a health prediction index. The power device is also configured to obtain a health status prediction data of the power device according to a comparison result between the health prediction index and the health prediction index threshold, so as to perform a follow-up process. Wherein the health index prediction model is constructed and trained by the cloud server by way of machine learning.