This paper reveals a stark inequality in the effect of ambient temperatures on death in human populations. Using district-level daily weather and annual mortality data from 1957 to 2000, we find that hot days lead to substantial increases in mortality in rural but not urban India. Despite being far poorer, the mortality response in urban India is not dissimilar to that in the US over the same period. Looking into potential mechanisms we find that the rural death effects are driven by hot days in the growing season which reduce productivity and wages in agriculture. Consistent with a model of endogenous survival in the face of credit constraints, we also find that the expansion of bank branches into rural India helped to mitigate these effects. When coupled with a climatological model that predicts many more hot days in a typical year by the end of this century, these estimates imply considerable reductions in rural Indian, but not urban Indian or US, life expectancy ceteris paribus. Joint with Robin Burgess, Dave Donaldson and Michael Greenstone.