OBJECTIVE: Three recent randomized trials have shown that male circumcision (circumcision) reduces HIV incidence in heterosexual men by about 60%. Mathematical models are needed to assess the historical role of circumcision in the observed disparate levels of prevalence in sub-Saharan Africa and to translate these findings into estimates of the population-level impact of circumcision on HIV prevalence. METHODS AND FINDINGS: A deterministic compartmental model of HIV dynamics with circumcision was parameterized by empirical data from the Rakai, Masaka, and Four-City studies. Circumcision was found to account for about two-thirds of the differential HIV prevalence between West Africa and East and Southern Africa. We found that in Kisumu, Kenya, and in Rakai, Uganda, universal circumcision implemented in 2008 would reduce HIV prevalence by 19% and 14%, respectively, by 2020. In Kisumu, a setting with high HIV prevalence, about 6 circumcisions would be needed for each infection averted while in Rakai, 11 circumcisions would be needed. Females will also benefit from circumcision with a substantial reduction in prevalence of about 8% in Kisumu and 4% in Rakai within a few years of universal circumcision. The beneficial impact of circumcision for both males and females will not be undermined by risk behavior compensation unless the increase in risk behavior is in excess of 30%. The effectiveness of circumcision as an intervention is maximized by universal circumcision within 2-3 years. CONCLUSIONS: In West Africa, circumcision may have "quarantined" the spread of HIV by limiting sustainable transmission to within high risk groups and bridge populations. Our findings indicate that circumcision is an effective intervention in both high and intermediate HIV prevalence settings. Circumcision coverage should be expanded as soon as possible to optimize the epidemiological impact.