Importance: Many premature infants are born without exposure to antenatal steroids (ANS) or with incomplete courses. This study evaluates the dose-dependent effect of ANS on rates of neonatal morbidities and early childhood neurodevelopmental outcomes of extremely premature infants. Objective: To compare rates of neonatal morbidities and 18- to 22-month neurodevelopmental outcomes of extremely premature infants exposed to no ANS or partial or complete courses of ANS. Design, Setting, and Participants: In this observational cohort study, participants were extremely premature infants (birth weight range, 401-1000 g; gestational age, 22-27 weeks) who were born at participating centers of the National Institute of Child Health and Human Development Neonatal Research Network between January 2006 and December 2011. Data were analyzed between October 2013 and May 2016. Main Outcomes and Measures: Rates of death or neurodevelopmental impairment at 18 to 22 months' corrected age. Neurodevelopmental impairment was defined as the presence of any of the following: moderate to severe cerebral palsy, a cognitive score less than 85 on the Bayley Scales of Infant and Toddler Development III, blindness, or deafness. Results: There were 848 infants in the no ANS group, 1581 in the partial ANS group, and 3692 in the complete ANS group; the mean (SD) birth weights were 725 (169), 760 (173), and 753 (170) g, respectively, and the mean (SD) gestational ages were 24.5 (1.4), 24.9 (2), and 25.1 (1.1) weeks. Of 6121 eligible infants, 4284 (70.0%) survived to 18- to 22-month follow-up, and data were available for 3892 of 4284 infants (90.8%). Among the no, partial, and complete ANS groups, there were significant differences in the rates of mortality (43.1%, 29.6%, and 25.2%, respectively), severe intracranial hemorrhage among survivors (23.3%, 19.1%, and 11.7%), death or necrotizing enterocolitis (48.1%, 37.1%, and 32.5%), and death or bronchopulmonary dysplasia (74.9%, 68.9%, and 65.5%). Additionally, death or neurodevelopmental impairment occurred in 68.1%, 54.4%, and 48.1% of patients in the no, partial, and complete ANS groups, respectively. Logistic regression analysis revealed that complete (odds ratio, 0.63; 95% CI, 0.53-0.76) and partial (odds ratio, 0.77; 95% CI, 0.63-0.95) ANS courses were associated with lower rates of death or neurodevelopmental impairment compared with the no ANS group. The reduction in the rate of death or neurodevelopmental impairment associated with exposure to a complete ANS course may be mediated through a reduction in rates of severe intracranial hemorrhage and/or cystic periventricular leukomalacia in the neonatal period. Conclusions and Relevance: Antenatal steroid exposure was associated with a dose-dependent protective effect against death or neurodevelopmental impairment in extremely preterm infants. The effect was partly mediated by ANS-associated reductions in rates of severe intracranial hemorrhage and/or cystic periventricular leukomalacia. These results support prompt administration of ANS, with the goal of a complete course prior to delivery.