All nucleoside analogues for treating HIV infection, due to their capacity to integrate into and alter human DNA, are experimentally genotoxic to some extent. The long‐term oncogenic risk after in utero exposure remains to be determined. Cancer incidence in uninfected children exposed to nucleos(t)ide reverse transcriptase inhibitors (NRTIs) was evaluated, by cross‐checking against the National Cancer Registry, in the French perinatal study of children born to HIV+ mothers. Twenty‐one cancers were identified in 15,163 children (median age: 9.9 years [interquartile range (IQR): 5.8–14.2]) exposed to at least one NRTI in utero between 1990 and 2014. Five of these children were exposed to zidovudine monotherapy, and 15 to various combinations, seven of which included didanosine. Overall, the total number of cases was not significantly different from that expected for the general population (SIR = 0.8[0.47–1.24]), but the number of cases after didanosine exposure was twice that expected (SIR = 2.5 [1.01–5.19]). Didanosine accounted for only 10% of prescriptions but was associated with one‐third of cancers. In multivariate analysis, didanosine exposure was significantly associated with higher risk (HR = 3.0 [0.9–9.8]). This risk was specifically linked to first‐trimester exposure (HR = 5.5 [2.1–14.4]). Three cases of pineoblastoma, a very rare cancer, were observed, whereas 0.03 were expected. Two were associated with didanosine exposure. Despite reassuring data overall, there is strong evidence to suggest that didanosine displays transplacental oncogenicity. These findings cannot be extrapolated to other NRTIs, but they highlight the need for comprehensive evaluations of the transplacental genotoxicity of this antiretroviral class. Environ. Mol. Mutagen., 60:404–409, 2019. © 2017 Wiley Periodicals, Inc.