Air pollution has been shown to have adverse effects on many health outcomes including respiratory effects, cardiovascular effects, and mortality. However, evidence on the effects of prenatal exposure is still limited. We investigate the causal impact of prenatal exposure to air pollution on neonatal health in Italy in the 2000s. We exploit variation in rainfall shocks to instrument for non-random air pollution exposure. Our empirical setting combines detailed information on mother's residential location from birth certificates with PM10 concentrations from air pollution monitors. Ten additional units in the average PM10 level (approximately one standard deviation) would decrease birth weight by about 0.5% and gestational age by 0.16%; it would increase the prevalence of low birth weight by 22% and of preterm birth by 16%. The effects are stronger in magnitude for third trimester exposure and for less educated mothers. These findings suggest that the health impacts of air pollution on newborns are unequally distributed in the population.