The authors report the long‐term survival of 458 young breast cancer patients with follow‐up times ranging from 20 to 51 years. Patients with distant metastases at diagnosis were not included in the study. An excess breast cancer mortality was observed for at least 40 years after diagnosis. At 40 years, the actuarial breast cancer survival for all patients was 32% ± 3% (standard error). For patients with localized and regional disease, it was 53% ± 6% and 19% ± 3%, respectively. A persistent excess mortality after 40 years, however, could not be ruled out because too few patients remained at risk. It is concluded that a cured fraction of patients defined as a group subject only to normal mortality risks, in practice, can only be identified in young patients with a low expected mortality, since older patients give only minor contributions to long‐term survival results. In an unselected series of breast cancer patients, an excess breast cancer mortality will probably persist during a period that is longer than the life expectancy of the average patient. A parametric survival model, the log‐normal model, was found to provide a good fit to the observed survival data. The cured fraction of patients predicted by the model was similar to the actuarial breast cancer survival at 40 years. These results suggest that the model could be used as a tool for analyzing observed survival patterns in breast cancer. It might thus provide an alternative to the conventional 5‐ or 10‐year rates. The good fit of the model also suggests that late recurrences represent one end of a continuous, broad spectrum of behavior in disseminated breast cancer.