Peak cooling loads are usually calculated at early stages of the building project, when large uncertainties affect the input data. Uncertainties arise from a variety of sources like the lack of information, random components and the approximate nature of the building mathematical model. Unfortunately, these uncertainties are normally large enough to make the result of the calculation very dependent on starting assumptions about the value of input data. HVAC engineers deal with uncertainties through worst-case scenarios and/or safety factors. In this paper, a new approach is proposed based on stochastic simulation methods. Uncertainty bands are applied to the input data and propagated through the building model in order to determine their impact on the peak cooling load. The result of this calculation is a probability distribution that quantifies the whole range of possible peak loads and the probability of each interval. The stochastic solution is compared with the conventional one, and a global sensitivity analysis is undertaken to identify the most important uncertainties. (C) 2010 Elsevier B.V. All rights reserved.