Abstract: A comparative static analysis of technological innovation was done in this paper and a peaks-shaped structural model between agglomeration and growth was established, which unified conflicting conclusions such as" inverted U-shaped" curve,"U-shaped" curve hypothesis and so on. Then, on the basis of measuring the spatial agglomeration degree of China's 25 provinces from 1990 to 2013 using spatial HHI and CR2 index, the theoretical propositions put forward in the paper were tested empirically using panel IV model capable of solving the problems of individual fixed effect and? endogenous variable simultaneously. The results show that when other factors remain constant,"inverted U-shaped" curve holds between agglomeration and growth; when innovation and other related factors improve, the inverted "U-shaped" curve will move to the upper right, forming a cyclical peaks-shaped structure, expanding the upper limit of regional growth level; while the agglomeration degree affects the growth level within the limit, the optimal agglomeration degree that achieves the highest level of economic growth is positively related to innovation level, and going beyond the limit is as bad as falling short.