The dominating majority of previous computer virus epidemic models assume a bilinear infection rate. This assumption, however, ignores the fact that, due to reasons such as overcrowded infected nodes and active protection measures taken at a high level of viral prevalence, the infection rate typically rises in a nonlinear fashion. This paper is devoted to understanding the impact of nonlinear infection rate on the propagation of computer infections. For that purpose, a new computer virus epidemic model is proposed by introducing a generic nonlinear infection rate into a traditional SLBS model. Theoretical analysis shows that, under moderate conditions, the proposed model admits a (viral) globally asymptotically stable equilibrium, fully demonstrating the robustness of stability of the equilibrium to the details of infections. The new model is justified through simulation experiments. We also determine the influence of some model parameters on the viral equilibrium. Our results extend some previously known results.