Previous versions of the FUND model assumed, like many integrated assessment models, that the carbon cycle is independent of climate change. I here introduce a feedback through which warming leads to higher net emissions. This increases the atmospheric concentration of carbon dioxide in the year 2100 by 100 (20-200) ppm. This leads to a higher estimate of the Pigou tax. The benefit of emission reduction now includes the direct benefits of lower emissions as well as the indirect benefits of a smaller feedback, but the latter effect is small. For any given stabilization target, abatement costs are substantially higher with the climate feedback than without. Abatement costs become sensitive to assumptions about climate change. Non-CO2 emission reduction becomes essential for meeting CO2 concentration targets. For pessimistic assumptions about the strength of the feedback, model results (for the 21st century) become sensitive to small variations in parameters.