The lack of biomarker information in population-based, nationally-representative data sets such as the Demographic and Health Surveys (DHS) has, until recently, limited analyzing the relative importance of individual and contextual factors in explaining the apparent reversal of the poverty-HIV relationship observed in African settings. In this paper, we fill this gap by taking advantage of the DHS data collected in Burkina Faso, Cameroon, Ghana, Kenya and Tanzania between 2003 and 2004. We choose these surveys because they include biomarkers for HIV that can be linked to individual-level survey data, and because the DHS cluster sampling strategy allows us to control for unobserved community-level factors by estimating cluster-level fixed effects models. We find that the observed positive correlation between wealth and HIV infection in African countries is spurious because of the concentration of wealthy individuals in communities (not necessarily urban) with higher HIV risk. This is particularly the case for Burkina Faso, Cameroon and Ghana, where the AIDS epidemic is at an earlier stage. We also find that, in all countries, HIV does not disproportionately affect the poor even in wealthier communities where the risk of HIV is higher. Policies should thus target inequalities across communities, rather than across income groups within communities, in order to reduce the spread of AIDS in African countries.