Communities in the network have closer ties than general alliances, which contribute to share of knowledge and collaborative research and development. Current studies are less concerned with clique communities and their impact on communities’ innovation. Based on the data of 9 high-tech industries in China, the clique-degrees are calculated in this paper. The more complex the cliques are, the larger the clique-degrees are. The clique percolation algorithm and the negative binomial regression model are applied to study the effect of the basic 3-cliques communities on the innovation capability. The result shows that there is an inverted U-type between clique-degrees and the innovative ability of communities, while the external enterprise number will inhibit the process. The study can be used as a reference for the choice of cliques’ communities in the enterprise alliance.