Mechanical fault is one of the main faults occurring during the life cycle of high-voltage circuit breakers (HVCBs). In order to enhance the reliability of HVCBs and the power system, it is important to assess and predict the mechanical condition of HVCBs. In this paper, the mechanical prediction algorithm for HVCBs based on support vector machine (SVM) was studied. SVM is a statistical learning algorithm which minimizes the structural risk for training purposes and can solve the problems of traditional machine learning methods (e.g. over-fitting, dimension disaster, local optimum, et al.). For the implement of algorithm, the historic data of contact travel and coil current were used to predict the future values. In order to predict the mechanical condition, the process of fault diagnosis for HVCBs can be applied. The methods, such as data scale, cross validation and grid search, were adopted to obtain the presetting parameters of algorithm and improve the performance. In the end, the mechanical life experiment data of a HVCB was applied to validate the feasibility of the algorithm. The results showed that the proposed algorithm could predict the mechanical condition of HVCBs successfully.