In this paper, we identify and quantify the role of international migration in the propagation of HIV across sub-Saharan African countries. We use a panel database on bilateral migration flows and HIV prevalence rates covering 44 countries over the nineties. Controlling for unobserved heterogeneity, spatial autocorrelation, reverse causality and reflection issues, and incorrect treatment of country fixed effects, we regress the log-change of HIV prevalence rates on the average levels of prevalence at destination and origin of migrants. We find evidence of a very robust emigration-induced propagation mechanism. On the contrary, immigration has no significant effect. Numerical experiments reveal that the long-run effect of emigration accounts for more than 5 percent of HIV prevalence rates in 18 countries (resp. 20 percent in 9 countries).