Including tax management in supply chain planning is mandatory, especially in Brazil, where each state practices its own tax policy. This dissertation studies the importance of tax costs upon deciding the location of distribution centers. Was developed a Mixed Integer Linear Programming (MILP) that includes the restrictions and opportunities created by the Brazilian tax policies along with the traditional restrictions where deployment, operating and transportation costs are included. Specifically, we focus on distribution systems that operate within an interstate and inter-municipal environment and where a Goods Circulation Tax (ICMS) applies. The optimization model minimizes the total costs given by the sum of distribution costs between factory - distribution centers - clients, facility fixed and variable costs and non-recovered cumulative tax credits. Thus, the model considers and calculates for each case and transaction the costs or credits resulting from the application of ICMS. The presented model is by nature non-linear and a linearization form of the model is presented. The model was applied to a Brazilian company that deals with non-durable goods and operates in the whole country. The solution found proposed the opening of three new distribution centers, besides the existing one, corresponding to a distribution system where all cumulative tax credits deriving from ICMS transactions are used. The annual saving amount to 29 percent of the distribution system present costs.