This work presents a machine learning–based uncertainty quantification framework for condition monitoring and generation risk analysis in nuclear power plants. The framework integrates machine learning models that estimate equipment reliability from condition monitoring data—while also explicitly accounting for associated uncertainties—in combination with techniques for synthetically generating data on electricity prices and plant output such that the statistical characteristics (i.e., data uncertainty) of the original data are preserved. These uncertainty sources are then propagated through a dynamic nuclear power plant generation risk model to quantify their combined impact on generation risk. The proposed approach improves predictions of future plant economic lost and supports cost-effective, risk-informed decision-making for project prioritization and long-term equipment reliability planning.