The productivity effect of infrastructure investment is controversial in the traditional literature using aggregate production function estimation due to reverse causality. This paper develops a new approach, using a structural model of firm-level production function, and matching Chinese firm-level production data with province-level infrastructure data. The estimated rates of return are about 6 percent averaged from 1999 to 2007. The returns triple if national-level spillover effects are taken into account. Controlling for the demand effect of public expenditure leads to smaller but still positive returns. The effect of infrastructure investment on firm-level productivity is heterogenous. With an increase in infrastructure investment, lower productivity firms are more likely to exit and higher productivity firms gain more market share.