Growth failure is a common problem in HIV-infected children. The extent to which this growth failure could be reversed after the children receive antiretroviral therapy (ART) is not known. This study assessed the incidence of growth failure in HIV-infected Thai children, impact of ART on growth, and the predictors of growth reversal after initiating ART. Growth parameters and other characteristics were extracted from the database of a prospective cohort of HIV-infected children (age <or=15 years) who were enrolled to initiate non-nucleoside reverse transcriptase inhibitor-based ART between August 2002 and May 2007. Body weight and height measurements, CD4 cell counts, plasma HIV RNA levels were collected at baseline and 24-week intervals. A total of 225 HIV-infected children were included, 116 (51%) were males. The median age at baseline was 7.4 years (interquartile range [IQR] 5.2-9.8). Fifty-three percent were in Centers for Disease Control and Prevention (CDC) category C and 54% had CD4 percentage 5 or less. The mean (standard deviation [SD]) of baseline weight-for-age (WAZ) and height-for-age (HAZ) z-scores were -2.02 (1.17) and -2.22 (1.51). The median follow-up time was 216 weeks (IQR 120-240). The cumulative probability of growth reversal among the 179 subjects with growth failure at entry was 58% (95% confidence interval [CI] 49-67) at week 240. In a multivariate Cox regression model, higher entry WAZ (p < 0.001) and HAZ (p < 0.001), use of a nevirapine-based regimen (compared to efavirenz, p = 0.027) and larger CD4% gains to week 48 (p < 0.001) were significant predictors of growth reversal after initiating ART. NNRTI-based ART leads to a substantial improvement in growth of HIV-infected children. Initiation of ART before the children developed growth failure should be encouraged.