Early initiation of antiretroviral therapy reduces HIV-related infant mortality. The early peak of pediatric HIV-related deaths in South Africa occurs at 3 months of age, coinciding with the earliest age at which treatment is initiated following PCR testing at 6 weeks of age. Earlier diagnosis is necessary to reduce infant mortality. The performances of the Amplicor DNA PCR, COBAS AmpliPrep/COBAS TaqMan (CAP/CTM), and Aptima assays for detecting early HIV infection (acquired in utero and intrapartum) up to 6 weeks of age were compared. Dried blood spots (DBS) were collected at birth and at 2, 4, and 6 weeks from HIV-exposed infants enrolled in an observational cohort study in Johannesburg, South Africa. HIV status was determined at 6 weeks by DNA PCR on whole blood. Serial DBS samples from all HIV-infected infants and two HIV-uninfected, age-matched controls were tested with the 3 assays. Of 710 infants of known HIV status, 38 (5.4%) had in utero (n = 29) or intrapartum (n = 9) infections. By 14 weeks, when treatment should have been initiated, 13 (45%) in utero-infected and 2 (22%) intrapartum-infected infants had died or were lost to follow-up. The CAP/CTM and Aptima assays identified 76.3% of all infants with early HIV infections at birth and by 4 weeks were 96% sensitive. DNA PCR demonstrated lower sensitivities at birth and 4 weeks of 68.4% and 87.5%, respectively. All assays had the lowest sensitivity at 2 weeks of age. CAP/CTM was the only assay with 100% specificity at all ages. Testing at birth versus 6 weeks of age identifies a higher total number of HIV-infected infants, irrespective of the assay.