In general, probabilistic fatigue life estimation were performed using Weibull analysis and end-of-life fatigue data. This is basically using traditional strain/stress ∼ life (S∼N) data. Although it is always better to estimate the reliability of a component based on probabilistic evaluation of end-of-life data, it may not be always possible to conduct hundreds of costly and time-consuming fatigue experiments for each and every different loading case. The strain/stress ∼ life data are mostly related to push-pull type symmetric (R = −1) loading cases and the traditional Weibull probabilistic model are based on the associated end-of-life data. However, the Weibull approach doesn’t depend on the time evolution of damage but is just based on end-of-life data. It is our assumption that for a given loading and environment, the time-evolution based probabilistic risk assessment (PRA) and reliability model would produce more accurate results compared to a PRA model simply based on the end-of-life data obtained under a different loading conditions. However, the time-evolution based PRA for a particular loading and environment requires hundreds of fatigue tests, which might not always be possible to perform due to the high cost and time requirements. To overcome these issues, we propose the use of Markov-Chain-Monte-Carlo (MCMC) techniques for time series or time-evolution prediction of damage states under a particular loading and environment (e.g. in this case, high-temperature PWR coolant-water condition) condition. Then, the probabilistic fatigue life can be estimated on the basis of the simulated scatter band in damage states for a given failure criteria. In this paper, we discuss the MCMC model based probabilistic damage state estimation and associated probabilistic fatigue life of 316 stainless steel subjected to cyclic loading at 300 °C in-air and PWR-coolant-water conditions.