In power system portfolio decisions, cost risk (uncertainty over cost) should be an important consideration, in addition to central estimates of cost. This study quantifies the uncertainty in each of the cost components for nuclear power, and combines them using a Monte Carlo analysis, allowing a direct assessment of cost risk for this technology. This can be used as an input to sophisticated portfolio optimisation modelling tools that take cost risk into account. Levelised cost of energy (LCOE) estimates are also provided here to allow an indicative comparison of the scale of cost risk, compared with other technologies. The most important contributors to cost risk for nuclear power generation are identified to be overnight capital cost (OCC) estimates, the degree of cost escalation over the construction and pre-construction periods and the duration of those periods. In the absence of cost escalation, the mean LCOE of nuclear power is found to be AU$145/MWh in jurisdictions excluding Asia (or AU$130/MWh if Asian plant costs are included in the distributions), with a standard deviation of AU$62/MWh. However, when cost escalation over construction and pre-construction periods is included at rates observed during nuclear build programs in France and the USA, the mean LCOE increases to AU$515/MWh, with a standard deviation of AU$2646/MWh. These results indicate that nuclear power costs have an 80% probability of exceeding AU$170/MWh, and a 50% probability of exceeding AU$278/MWh, based upon Monte Carlo analysis. This suggests that for OECD jurisdictions without an established nuclear industry (such as Australia), nuclear power may be comparable in cost with 'dispatchable', synchronous renewable alternatives such as concentrating solar thermal. Furthermore, the considerable cost risk associated with nuclear power is a significant disadvantage. Copyright (C) 2016 John Wiley & Sons, Ltd.