Based on the panel data of high-tech industries of 29 provinces, autonomous regions and municipalities in China from 2014 to 2018, the super-efficiency DEA model was first used to calculate the production efficiency of each province, and then the DEA-Malmquist model was constructed to calculate and decompose the comprehensive productivity index of each province. The results show that: The overall level of innovation efficiency of China's high-tech industries is relatively high, but more than half of the provinces and municipalities have unreasonable factor input structure, which is mainly distributed in the central and western regions, and the innovation efficiency has great regional differences. The average total factor productivity index of China's high-tech industries is greater than 1. Many provinces suffer from low scale efficiency, resulting in n-type callback vibration of innovation efficiency. Technological progress and technical efficiency are the main driving forces for total factor productivity improvement. Therefore, in order to further improve the innovation efficiency of China's high-tech industry, attention should be paid to maintain the technical level and efficiency of high-tech industry, especially to optimize the allocation of high-tech industry's scientific and technological resources, and to improve the management level of high-tech industry's innovative resources.