The atomic force microscopy (AFM)-based repeated nanomachining of nanochannels on silicon oxide surfaces is investigated both theoretically and experimentally. The relationships of the initial nanochannel depth vs. final nanochannel depth at a normal force are systematically studied. Using the derived theory and simulation results, the final nanochannel depth can be predicted easily. Meanwhile, if a nanochannel with an expected depth needs to be machined, a right normal force can be selected simply and easily in order to decrease the wear of the AFM tip. The theoretical analysis and simulation results can be effectively used for AFM-based fabrication of nanochannels.