A model for estimating the failure probability of axially cracked steam generator tubes is proposed. The model compares observed crack length distribution with critical crack length distribution by means of probabilistic fracture mechanics. The observed crack length is influenced by measured data, measurement reliability, sizing accuracy, and predicted crack growth rate. The critical crack length is defined by a deterministic mechanical model. All cracks are conservatively assumed to extend through the tube wall. The effect of the plugging limit is studied along with the number of cracked tubes to perform risk-based lifetime optimization of steam generators. A numerical example presented considers hypothetical accidental operating conditions during a feedwater line break.