Since China’s reform and opening up, its infrastructure has been improved dramatically, creating the so-called infrastructure miracle. As we have already known that intra-region trade costs and inter-region trade cost, which are determined by the quality of infrastructure, have crucial effects on the spatial distribution of industries across regions within a country. Adopting the footloose capital model, we investigate the effects of trade cost differences due to infrastructure quality on regional industrial agglomeration, regional gap and social welfare. The main findings of the paper are listed as follows. （1) The residents of the less developed areas will benefit both from the improvement of the intra-region infrastructure of the less developed areas and/or of the inter-region infrastructure between the developed and the less developed ones. However, the mechanisms of the two policies are different. The policy of improving the infrastructure in the less developed areas has direct and indirect effects, both of which benefit the residents within it. The direct effect is more clearly visible in the rapid improvements in the infrastructure, while the indirect effect takes more time to be fully realized. This finding explains the common phenomenon of the low utilization of the infrastructure in the backward areas. The policy of improving inter-region infrastructure benefits the residents of the less developed areas through decreasing import cost. However, when the infrastructure in developed areas is better than that in less developed areas, this kind of policy will not only widen the regional gap as industries agglomerate around developed areas, but also widen regional income inequity. A new perspective for us to consider the relationship between industrial agglomeration and regional gap is that the industrial agglomeration caused by the infrastructure differences between developed areas and less developed ones may to some extent account for the widening income gap. （2) The residents of the less developed areas will benefit from the improvement of the infrastructure as industries spread to these areas, while the residents of the developed ones may enjoy fewer welfare benefits. There is high probability that this kind of policy may cut down the total social welfare benefits when both the industrial-size gap and the infrastructure gap between regions are significant. This finding has significant implications for the central government in making policies for regional coordination. If coordinated regional development is a long term target of the central government, early investment in the infrastructure in the less developed areas is vital. If the investment is delayed until considerable gap develops between regions, an overall efficiency loss may occur and the economic development in the earlier stages will be greatly hampered. （3) The main policy implication of this paper is that the optimal policy of the central government to improve the infrastructure depends on the initial levels of industrial sizes and infrastructure of the two regions. When the size difference between regional economies is small, when the trade barriers between regions are high, and when the less developed areas have significant investment cost advantage in improving the intra-region infrastructure, the central government should choose to invest in the infrastructure in less developed areas to maximize the total social welfare. One limitation of this study is that we only assume the public expenditure on infrastructure to be exogenous. In more general and realistic situations, local governments actually rely on tax revenue to finance infrastructure investments. One interesting extension would be to introduce tax policy into our model in order to better explain the strategic behaviors of local governments and their impact on industrial agglomeration and economic development.