The goal of this study was to characterize the uncertainty associated with the cost of drilling and completion of geothermal wells. Previous research and publications have produced correlations for the average cost of geothermal wells as a function of well depth. This project develops this concept further by using a probabilistic approach to evaluate the distribution of geothermal well costs for a range of well depths. The well cost uncertainty was characterized by identifying the main cost components of geothermal wells and quantifying the probability distributions of the key variables controlling these costs. These probability distributions were determined based on the detailed cost records of U.S. geothermal wells drilled or designed from 2009 to 2013 as well as cost data from drilling equipment manufacturers and vendors. Probability distributions of the key variables were examined to find statistically significant correlations between them. Lastly, the previously determined probability distributions of individual cost components and the correlations between them were input into WellCost Lite, a predictive geothermal drilling cost model, using the Monte Carlo method. This approach allowed us to generate the overall well cost probability distributions for 8000-15,000 ft. (2400-4600 m) geothermal wells. We have shown that the median geothermal well cost increases exponentially with depth. Deep wells typically have higher cost uncertainty and more positively-skewed cost probability distributions. The correlations presented in this paper can be used to determine the economic feasibility of geothermal energy systems, assess the project risk, and facilitate investment decisions. (C) 2016 Elsevier Ltd. All rights reserved.