Objective. To explain differences in weight growth velocity of extremely premature infants among 6 level III neonatal intensive care units (NICUs). Methods. In 6 NICUs, we studied 564 infants, stratified by gestational age (GA), who were first admissions, survivors, <30 weeks’ GA at birth, and in the NICU at least 16 days. Case mix (eg, birth weight, GA, race, illness severity, prenatal steroids), exposure to medical practices/complications (eg, respiratory support, postnatal steroids, necrotizing enterocolitis, infection), and nutritional intake (kcal/kg/d and protein in g/kg/d) were collected and used to predict weight growth velocity between day 3 and day 28 (or discharge, if transferred early) in multiple linear regression models. Results. Weight growth velocities varied significantly among the 6 NICUs. Adjustment for case mix and medical factors explained little of this variability, but additional control for calorie and especially protein intake accounted for much of the intersite variability. For the average infant, adjusted growth velocity ranged from 10.4 to 14.3 g/kg/d among the sites studied. The final predictive model, including case mix and medical and nutritional factors, explained 53% of the overall variance in growth velocity. Prolonged (≥15 days) exposure to postnatal steroids and greater severity of illness both decreased growth velocity. The model predicted that adding 1 g/kg/d protein to the mean intake for our sample would increase growth by 4.1 g/kg/d. Conclusions. Variation in nutrition explained much of the difference in growth among the NICUs studied. Mean intake of calories and protein failed to meet recommended levels, and the average growth in only 1 NICU approximated intrauterine growth standards. Increasing nutritional intake into the recommended ranges, in particular of protein, may increase growth of extremely premature infants up to or above intrauterine rates.