The recent literature discusses optimal paths towards stabilization of atmospheric concentrations of greenhouse gases, The differences in the impact of climate change between alternative scenarios have largely been ignored, however, This paper analyses these differences, Using a selection of plausible, yet speculative, models of climate change impact and a range of sensitivity analyses, it is shown that no stabilization path is unambiguously preferred to its alternative, The ambiguity originates both from the limited scientific knowledge and from ethical choices (eg time discount rate), However, over most of the assumption space explored here, there appears to be a preference for an earlier reduction of greenhouse gas emissions, The best estimate of the additional impact caused by a delay of emission reduction is smaller than the cost savings, but uncertainties are too large to draw this conclusion with any certainty,