The successive cancellation list (SCL) decoding is the most powerful decoding algorithm for polar codes. However, the long latency makes it unsuitable for practical applications, and thus many algorithms have been presented to resolve the problem. While most of the previous algorithms have reduced the latency by pruning unnecessary operations in the middle of the process, this paper proposes an early stopping criterion. The proposed method can determine the minimum metric path without having to traverse the tree to the end if a special condition is satisfied at a certain point. The remaining bits located after the stopped bit position are decoded directly from the soft information. The simulation results show that the criterion reduces the decoding latency effectively for various code rates and list sizes without any performance loss. We also propose a hardware architecture that is efficient in realizing the proposed method. Compared to previous SCL decoding algorithms, the proposed decoder achieves 16% higher throughput and 12% improved area efficiency at an SNR of 5.0 dB when the list size is 4.