Summary It remains unclear why the global distribution of human immunodeficiency virus (HIV), between and within continents, is so heterogeneous. This ecologic study of 34 populations of the Americas explored the hypothesis that populations differ in their intrinsic, biological susceptibility to HIV which, together with exposure, might determine the ultimate ‘mature’ prevalence. If true, national HIV prevalence in populations of the Americas should be predictable from each country's ethnic mosaic, inter-racial admixture and HIV prevalence in regions of Africa, Europe and Asia from where their ancestors migrated. For each country, the adult population (15–49 years) was multiplied by the proportion corresponding to each ethnic group by HIV prevalence in the country/region of origin of each group, yielding the predicted prevalences, which were then compared with observed prevalences documented by UNAIDS for 2001. Predicted and observed HIV prevalences were highly correlated (r = 0.70, P < 0.001). In North America, predicted prevalences were within 0.5% of the observed values, except for African–Americans and African–Canadians. In Central and South America, differences between predicted and observed prevalences were <1.0% except in Honduras and Guyana. Some Caribbean countries had a predicted prevalence identical to the observed one, but there were outliers. Overall, predicted prevalence was 0.93% and observed prevalence 0.64%; two-thirds of this difference was attributed to Brazil. Although it was not possible to adjust to the confounding effects of sexual behaviour and cofactors of transmission (such as sexually transmitted infections) because of the lack of nationally representative data for each and every country, a number of arguments reviewed in the paper suggest that confounders cannot explain all this association and that differential susceptibility might be an important determinant of steady-state HIV prevalence.