The long process of political development of human society is full of phenomena of power production and transfer. With the advent of the era of artificial intelligence, data and algorithm, as a key component of artificial intelligence technology, have gradually become a new element of power production. Whoever has more sufficient data and more advanced algorithm may have more or greater power, and the empowerment effect of technology can be demonstrated.Technological empowerment may give rise to new power subjects, and the power transfer phenomenon may occur again.Artificial intelligence enterprises not only have some power because of their unique technological advantages in data and algorithm, but also gain some power because of providing technical assistance for state governance,so that artificial intelligence enterprises may become new power center comparable to the state.The growth of the power of artificial intelligence enterprises may promote the transformation of state governance power structure from traditional single-centered to multi-centered. At the same time, due to the increasing reliance of state governance in the era of artificial intelligence on algorithm support provided by artificial intelligence enterprises, which may lead to errors in algorithms, combined with the existence of “algorithm black box” and the chain of avoiding responsibility, the power growth of artificial intelligence enterprises may also eliminate the basic political rules of power and responsibility, causing the phenomenon of separation of power and responsibility or the problem of state governance accountability.The probability of the emergence of new power center depends not only on the technological strength comparison of the state and artificial intelligence enterprises, but also on the degree of technological dependence of state governance on artificial intelligence enterprises. In response to the phenomenon of separation of power and responsibility or the problem of state governance accountability,in addition to promoting the high self-discipline of artificial intelligence enterprises, it is also necessary to design systems such as the state matching record filing and pilot system, random sampling inspection system, service points system.