This paper investigates the impact of dams on infant mortality in Africa. It exploits variation in child exposure to dams at birth across both regions and birth cohorts. The paper uses a compiled sample of 912,080 children from DHS surveys in 17 African countries. Only children of non-migrant mothers from this sample are used in the main analysis to minimise sample selection bias from endogenous parental migration. We account for the fact that dam impacts are determined by relative locations of households and dams along the river network. Our results indicate that each dam in the neighbouring upstream river basin from a child reduces the probability of her death during age 0-12 months by 3.84-4.60% on average. This is due to households being close enough to access dam irrigation services that are provided immediately downriver from the dam, and which reduce vulnerability to rainfall shocks in agricultural production. In contrast, upstream dams in river basins farther upriver from the child increase infant mortality risk by 2.18-2.36%. This is because dams reduce water levels downriver, and households cannot access compensating irrigation services from dams that are too far upstream. The decline in water level is very detrimental to oodplain recession agriculture, on which much of Africa depends. Regulated discharges of water from these upstream dams to reduce amplitude of rainfall shocks downriver also do not compensate for the detriments of declining water levels. Within-basin dams marginally increase infant mortality by 0.74-0.79% for children born in oodplain regions, by disrupting the river inundation cycle and possibly increasing incidence of waterborne diseases in the same river basin. Downstream dams have no apparent impact on infant mortality.