Cancellation of clutter and multi-path is one of the key steps in passive radar target information extraction. Extensive Cancellation Algorithm Batches (ECA-B) is an effective time-domain clutter suppression algorithm, but with high time and space complexity, and even higher with multi-channel (or multi-beam) data processing. Combining high memory throughput and tremendous computational horsepower of GPU graphics processor, this paper proposes a multi-channel ECA-B algorithm which is suitable for parallel implementation on GPUs. Firstly, the principle of multi-channel ECA-B algorithm is derived, avoiding the redundancy of processing each channel singly. Then an iterative calculation method is presented for reducing the biggest time-consuming calculation of the correlation matrix, so that time and space complexity are both reduced to 1/K (K is clutter’s degree of freedom) of the conventional method. Finally, the full GPU parallel implementation of the algorithm is given. The simulation and experimental results verify the accuracy and effectiveness of the proposed algorithm.