Background: Decisions regarding provision of intensive care and post-discharge follow-up for infants born extremely preterm (<28 weeks' gestation) are based on the risks of mortality and neurodevelopmental disability. We aimed to elucidate the changes in probability of three outcomes (death, survival with major disability, and survival without major disability) with postnatal age in extremely preterm infants offered intensive care, and the effect of postnatal events on the probability of survival without major disability. Methods: In this prospective observational study, we used data from three geographical cohorts composed of all extremely preterm livebirths offered intensive care at birth during three distinct periods (1991–92, 1997, and 2005) in Victoria, Australia. Participants were assessed at 8 years' corrected age for major neurodevelopmental disability, defined as moderate or severe cerebral palsy, general intelligence more than 2 SDs below term-born control means, blindness, or deafness. Probabilities of outcomes conditional on survival to different postnatal ages were calculated by logistic regression. Multivariable logistic regression was used to assess factors predictive of survival with major disability. Findings: 751 (82%) of 915 extremely preterm livebirths free of lethal anomalies were offered intensive care, of whom 546 (73%) survived to age 8 years. Of the 499 survivors assessed, 86 (17%) had a major disability. With increasing gestational age at birth or days of postnatal survival, the probability of death decreased and of survival without major disability increased. By contrast, the probability of survival with major disability varied little with gestational age or postnatal survival. In survivors, major disability was associated with the occurrence of four important postnatal events: grade 3 or 4 intraventricular haemorrhage (odds ratio 2·61 [95% CI 1·11–6·15]), cystic periventricular leukomalacia (9·17 [3·57–23·53]), postnatal corticosteroid use (1·99 [1·03–3·85]), and surgery (2·78 [1·51–5·13]). 241 survivors (48%) had no major postnatal events during the newborn period, and had the lowest prevalence of major disability (17 participants [7%]). The probability of survival without major disability decreased with increasing number of major events (0·93 [0·89–0·96] for no events vs 0·31 [0·11–0·59] for three or more events). Interpretation: Long-term prognosis in terms of death and major neurodevelopmental disability changes rapidly after birth for extremely preterm infants. Counselling of families and post-discharge planning should be individualised to changing circumstances following birth. Funding: National Health and Medical Research Council of Australia.