Ecosystem carbon cycle feedbacks to climate change comprise one of the largest remaining sources of uncertainty in global model predictions of future climate. Both direct climate effects on carbon cycling and indirect effects via climate-induced shifts in species composition may alter ecosystem carbon balance over the long term. In the short term, climate effects on carbon cycling may be mediated by ecosystem species composition. We used an elevational climate and tree species composition gradient in Rocky Mountain subalpine forest to quantify the sensitivity of all major ecosystem carbon stocks and fluxes to these factors. The climate sensitivities of carbon fluxes were species-specific in the cases of relative aboveground productivity and litter decomposition, whereas the climate sensitivity of dead wood decay did not differ between species, and total annual soil CO2 flux showed no strong climate trend. Lodgepole pine relative productivity increased with warmer temperatures and earlier snowmelt, while Engelmann spruce relative productivity was insensitive to climate variables. Engelmann spruce needle decomposition decreased linearly with increasing temperature (decreasing litter moisture), while lodgepole pine and subalpine fir needle decay showed a hump-shaped temperature response. We also found that total ecosystem carbon declined by ∼50% with a 2.8°C increase in mean annual temperature and a concurrent 63% decrease in growing season soil moisture, primarily due to large declines in mineral soil and dead wood carbon. We detected no independent effect of species composition on ecosystem C stocks. Overall, our carbon flux results suggest that, in the short term, any change in subalpine forest net carbon balance will depend on the specific climate scenario and spatial distribution of tree species. Over the long term, our carbon stock results suggest that with regional warming and drying, Rocky Mountain subalpine forest will be a net source of carbon to the atmosphere.