A mixed-integer nonlinear programming (MINLP) algorithm for the synthesis of cooling networks is presented. As opposed to previous models in which the minimization of utility consumption has been formulated, the algorithm considers simultaneously the capital cost for the coolers and the utility costs. The strategy is based on a superstructure that allows bypass and splitting of utility streams. Also, the superstructure considers a combination of arrangements in series and in parallel that allows for an efficient use of cooling streams. The typical arrangement of cooling networks in parallel usually conducts to higher total costs. Also, it is shown that the use of minimum amounts of cold utilities does not necessarily conduct to the minimum total cost. Three case studies are used to show the application of the proposed method.