Abstract Study objectives We aimed to investigate the prospective associations of sleep phenotypes with severe intentional self-harm (ISH) in middle-aged and older adults. Methods A total of 499,159 participants (mean age: 56.55 ± 8.09 years; female: 54.4%) were recruited from the UK Biobank between 2006 and 2010 with follow-up until February 2016 in this population-based prospective study. Severe ISH was based on hospital inpatient records or a death cause of ICD-10 codes X60-X84. Patients with hospitalized diagnosis of severe ISH before the initial assessment were excluded. Sleep phenotypes, including sleep duration, chronotype, insomnia, sleepiness, and napping, were assessed at the initial assessments. Cox regression analysis was used to estimate temporal associations between sleep phenotypes and future risk of severe ISH. Results During a follow-up period of 7.04 years (SD: 0.88), 1,219 participants experienced the first hospitalization or death related to severe ISH. After adjusting for demographics, substance use, medical diseases, mental disorders, and other sleep phenotypes, short sleep duration (HR: 1.50, 95% CI: 1.23–1.83, p < .001), long sleep duration (HR: 1.56, 95% CI: 1.15–2.12, p = .004), and insomnia (usually: HR: 1.57, 95% CI: 1.31–1.89, p < .001) were significantly associated with severe ISH. Sensitivity analyses excluding participants with mental disorders preceding severe ISH yielded similar results. Conclusion The current study provides the empirical evidence of the independent prediction of sleep phenotypes, mainly insomnia, short- and long-sleep duration, for the future risk of severe ISH among middle-aged and older adults.