This letter proposes a handover-aware content replication mechanism for mobile content delivery networks. This mechanism allows us to select the most optimal surrogate servers and assign these to the mobile clients through built-in tables. The optimal surrogate selection process begins by predicting the mobility of clients through Gauss-Markov model. Then, it continues with server modeling based on G/G/1 queuing system for load estimation. We also use the k -order autoregressive model to observe the future values of the estimated loads. The proposed strategy reduces the content replication delays and packet losses by 53% and 42%, respectively.