AIM: We sought to study the survival of newborn children according to HIV status of the mother, that of the child and the timing of infection. METHODS: This is a prospective cohort study of 883 mothers (665 HIV-positive and 218 HIV-negative) and their infants. Data were collected using semi-structured questionnaires during home visits between the antenatal period and 36 weeks post-delivery. Infant HIV status was determined at 3, 24 and 36 weeks by HIV DNA PCR. RESULTS: The majority (81.3%) of infected infants who died were infected by 3 weeks of age. Of the HIV-exposed infants who died, 19 (28.4%) died before 6 weeks and 38 (56.7%) died by 12 weeks. The hazard ratio (HR) of mortality at 36 weeks of age in HIV-infected infants compared with exposed but negative infants was 8.9 (95% CI: 6.7-11.8). There was no significant difference in 36 week survival rates between HIV-non-exposed and HIV-exposed but negative infants (HR: 0.7; 95% CI: 0.3-1.5). The infant being HIV-positive at age 3 weeks (HR: 32 95% CI: 14.0-73.1) and rural site (HR: 4.4 95% CI: 1.2-23.4) were the two independent risk factors for infant death amongst HIV-exposed infants. CONCLUSION: The prognosis for infants with early HIV infection was very poor in this cohort. A greater focus on prevention of early infection, earlier screening for HIV infection and access to antiretrovirals for eligible infants is recommended.