Object. The aim of this study was to investigate the clinical appearance of untreated pediatric brainstem cavernous malformations (CMs) and to identify the hemorrhage risks and functional outcomes. Methods. All pediatric patients with a diagnosis of brainstem CM between 1985 and 2012 were registered. The clinical chart and radiographs were recorded, and follow-up evaluations were obtained prospectively. Results. A total of 85 patients (69.4% male) were included with a mean age of 12.7 years. Sixty-seven patients (78.8%) had prior hemorrhage, and 6 patients (7.1%) were asymptomatic. There were 15 midbrain lesions, 53 pons lesions, and 17 medulla lesions. The mean lesion size was 1.9 cm. During a total of 401.6 patient-years of follow-up, 47 hemorrhages occurred in 37 patients, and the annual hemorrhage rate was 11.7% per patient-year. The mean hemorrhage interval was 47.8 months. The hemorrhage risk declined over time, especially after the first 2 years. Both a lesion size >= 2 cm (hazard ratio [HR] 2.122, p = 0.037) and the presence of perilesional edema (RR 2.192, p = 0.039) predicted future hemorrhage and were associated with a high annual hemorrhage rate. The hemorrhage-free survival at 6 months was 85.7%, and at 1, 5, 10, and 15 years was 71.5%, 49.4%, 27.5%, and 13.7%, respectively. At the most recent functional evaluation, 33 patients (38.8%) had improved, 32 (37.6%) had stabilized, and 20 (23.5%) had worsened, without any deaths. Twenty-two patients (25.9%) obtained a full recovery. Prospective hemorrhage (HR 0.191, p = 0.003) was the adverse predictor for full recovery. Full recovery primarily occurred within the first 12 months, after which the chance of full recovery decreased. The cumulative percentage of complete recovery at 6 months was 32.7%, and at 1, 3, and 5 years was 40.8%, 43.6%, and 49.2%, respectively. Conclusions. In this study the hemorrhage rate was relatively high in pediatric brainstem CMs, although the functional outcome was acceptable. The decline in hemorrhage risk and the identified adverse predictors in this study were helpful for clinicians and patients when deciding on treatment. Referral bias and the insufficient follow-up period of the study were highlighted as limitations.