Power generation megaprojects are central in energy planning and policy. However, many megaprojects fail to deliver the scale and efficiency aspirations initially expected. The purpose of this paper is to estimate the probability distribution function of cost overruns and delays in the construction of power generation projects, with a particular focus on mega hydroelectric dams recently built in Brazil. Results show that construction costs were, on average, 97.53% above the initial estimates. The distribution that best fits the hydroelectric power plants costs overruns is the gamma distribution. For the delays, the construction completion time had an average increase of 74.28%, or 3.5 years. The distribution that best fits the hydroelectric plants delays is the lognormal distribution. The essential statistical message obtained in this paper is that megaprojects fail to deliver the economies of scale embedded in large projects because the exposure to risk is disproportionate to the financial economies they can generate. Decision makers should carefully evaluate whether "bigger is better".