BACKGROUND: There is now strong evidence that cannabis use increases the risk of psychoses including schizophrenia, but the relationship between cannabis and different psychotic disorders, as well as the mechanisms, are poorly known. We aimed to assess types of psychotic outcomes after use of cannabis in adolescence and variation in risk over time. METHOD: A cohort of 50 087 military conscripts with data on cannabis use in late adolescence was followed up during 35 years with regard to in-patient care for psychotic diagnoses. RESULTS: Odds ratios for psychotic outcomes among frequent cannabis users compared with non-users were 3.7 [95% confidence interval (CI) 2.3-5.8] for schizophrenia, 2.2 (95% CI 1.0-4.7) for brief psychosis and 2.0 (95% CI 0.8-4.7) for other non-affective psychoses. Risk of schizophrenia declined over the decades in moderate users but much less so in frequent users. The presence of a brief psychosis did not increase risk of later schizophrenia more in cannabis users compared with non-users. CONCLUSIONS: Our results confirm an increased risk of schizophrenia in a long-term perspective, although the risk declined over time in moderate users.