Improving the scalability of keyword spotting systems remains a significant challenge. Recent methods based on few-shot class-incremental learning often suffer from catastrophic forgetting. To tackle this issue, we propose Neural Collapse Keyword Spotting (NC-KWS), an incremental keyword spotting framework inspired by the Neural Collapse phenomenon. Specifically, NC-KWS employs a classifier constructed with an Equiangular Tight Frame formulation, which enforces maximal pairwise separation between classes and enhances its discriminative capacity. In addition, an optimal feature-classifier alignment is predefined as a fixed target throughout incremental learning, maintaining a theoretically optimal solution for class-incremental models. Experimental results show that, compared with the baseline models, NC-KWS achieves higher accuracy and greater robustness in later sessions with more incremental categories.