Abstract Objectives Maternal factors shape the risk of infant undernutrition, however the contributions of age at marriage versus age at pregnancy are rarely disentangled. We explore these issues in a population from lowland rural Nepal, where median ages at marriage and first pregnancy are 15 and 17 years respectively and marriage almost always precedes pregnancy. Methods We analyzed data on first‐time mothers (n = 3002) from a cluster‐randomized trial (2012–2015). Exposures were ages at marriage and pregnancy, categorized into groups. Outcomes were z‐scores for weight (WAZ), length (LAZ), head circumference (HCAZ), and weight‐for‐length (WLZ), and prevalence of wasting and stunting, for neonates (<8 days) and infants (6–12 months). Mixed linear and logistic regression models tested associations of marriage and pregnancy ages with outcomes, adjusting for parental education, household assets, caste, landholding, seasonality, child sex, intervention arm, randomization strata and cluster. Results For neonates, pregnancy <18 years predicted lower LAZ, and <19 years predicted lower WAZ and HCAZ. Results were largely null for marriage age, however early pregnancy and marriage at 10–13 years independently predicted neonatal stunting. For infants, earlier pregnancy was associated with lower LAZ and HCAZ, with a trend to lower WAZ for marriage 10–13 years. Early pregnancy, but not early marriage, predicted infant stunting. Conclusions Early marriage and pregnancy were associated with poorer growth, mainly in terms of LAZ and HCAZ. Associations were stronger for neonatal than infant outcomes, suggesting pregnancy is more susceptible to these stresses. Early marriage and pregnancy may index different social and biological factors predicting child undernutrition.