The objective of this study is to determine whether detection of HIV infection was delayed in infants exposed to antiretroviral prophylaxis to prevent HIV transmission during breastfeeding. The Breastfeeding, Antiretrovirals and Nutrition (BAN) study was a randomized trial of 2369 mother-infant pairs conducted from 2004 to 2010. In addition to an intrapartum regimen, all mother-infant pairs were randomly assigned to three antiretroviral intervention arms during 28 weeks of breastfeeding: no further antiretroviral prophylaxis (control arm); infant-daily nevirapine (nevirapine arm); and maternal zidovudine, lamivudine and either nevirapine, nelfinavir or lopinavir-ritonavir (maternal arm). After breastfeeding cessation counselling and stopping the antiretroviral interventions by 28 weeks, 28 infant HIV infections occurred. To determine whether these infections occurred during the breastfeeding and antiretroviral intervention phase but had delayed detection on the antiretroviral arms, we performed ultrasensitive (droplet digital PCR) HIV testing on infants with stored peripheral blood mononuclear cell (PBMC) specimens at 24 weeks (n = 9). Of the nine infants, all three on the infant nevirapine arm had detectable HIV DNA at 24 weeks, compared with two of four on the maternal antiretroviral arm and one of two on the control arm. For infants with detectable HIV at 24 weeks, the median delay in detection between the ultrasensitive and standard assays was 18.3 weeks for the nevirapine arm, 15.4 weeks for the maternal arm and 9.4 weeks for the control arm. The prolonged inability to detect HIV with standard assays in the context of postnatal antiretroviral prophylaxis suggests that early antiretrovirals may restrict HIV replication sufficiently to lead to missed diagnosis among infected infants. Therefore, repeat virologic testing is warranted beyond the WHO-recommended point of testing at 6 weeks after breastfeeding cessation.