OBJECTIVES: To analyse the association between socio-economic status (SES) and HIV in Manhica, a district of Southern Mozambique with one of the highest HIV prevalences in the world. METHODS: Data were gathered from two cross-sectional surveys performed in 2010 and 2012 among 1511 adults and from the household census of the district's population. Fractional polynomial logit models were used to analyse the association between HIV and SES, controlling for age and sex and taking into account the nonlinearity of covariates. The inequality of the distribution of HIV infection with regard to SES was computed through a concentration index. RESULTS: Fourth and fifth wealth quintiles, the least poor, were associated with a reduced probability of HIV infection compared to the first quintile (OR = 0.595, P-value = 0.009 and OR = 0.474, P-value < 0.001, respectively). Probability of HIV infection peaked at 36 years and then fell, and was always higher for women regardless of age and SES. HIV infection was unequally distributed across the SES strata. CONCLUSIONS: Despite the high HIV prevalence across the entire population of Manhica, the poorest are at greatest risk of being HIV infected. While women have a higher probability of being HIV positive than men, both sexes showed the same infection reduction at higher levels of SES. HIV interventions in the area should particularly focus on the poorest and on women without neglecting anyone else, as the HIV risk is high for everyone.