The paper presents a simplified model of the form often used to project long-term emissions of carbon dioxide from energy production and use. It then considers three uncertainties in parameters of the model concerning (a) the rare of improvement in energy efficiency, (b) the costs of environmental damage, and (c) the rate of technical progress in the development and use of technologies for abating CO2 emissions. Monte Carlo analysis is used to estimate the frequency distributions of costs and benefits under different policy assumptions. The results suggest that the possibilities of an economic surprise cannot be ruled out and that 'technology policies' to support the development of non-carbon technologies directly are robust under uncertainty. Such policies can also be defended by reference to their option value, to their environmental benefits, and to the positive externalities of innovation.