Abstract: To control the competition risk of a project of virtual enterprise in dynamic alliance, of which the time to complete its subprojects is stochastic, the problem of partner selection is studied to maximize the completion probability of the project that is constrained by investment and the precedence of its subprojects, which is described on the basis of program evaluation and review technique (PERT). Then, a nonlinear integer-programming model is developed and a PERT-embedded genetic algorithm is designed to solve the problem. A case study is carried out to compare the proposed algorithm with enumeration algorithm. The results suggest that the proposed method is effective in both running time and performance, as a useful management instrument to select partner for a virtual enterprise in dynamic alliance under random conditions.