With the explosive growth in R&D investments and patent applications in recent decades, has China truly achieved improved innovation quality? To answer this question, it is necessary to correctly estimate China’s innovation efficiency. However, when measuring innovation efficiency, the dynamic and network features of the innovation process are seldom considered simultaneously. Therefore, this paper employs the method of dynamic network data envelopment analysis to estimate the overall, period, and sub-stage innovation efficiency of China’s 30 provinces between 2012 and 2016. We conclude that: (1) There is a regional imbalance in the overall scores, for example, developed provinces are more efficient than less developed areas. (2) The period and sub-stage values are not high in each period and represent a gap among the various provinces. (3) For most provinces, scores in the R&D stage are higher than those in the commercialization phase, indicating an uneven distribution of the innovation structure. Accordingly, policymakers should focus on innovation efficiency indicators, encourage innovation according to local conditions, and facilitate the long-run enhancement of both R&D and commercialization.