Over the long term, soil carbon (C) storage is partly determined by decomposition rate of carbon that is slow to decompose (i.e., recalcitrant C). According to thermodynamic theory, decomposition rates of recalcitrant C might differ from those of non-recalcitrant C in their sensitivities to global warming. We decomposed leaf litter in a warming experiment in Alaskan boreal forest, and measured mass loss of recalcitrant C (lignin) vs. non-recalcitrant C (cellulose, hemicellulose, and sugars) throughout 16 months. We found that these C fractions responded differently to warming. Specifically, after one year of decomposition, the ratio of recalcitrant C to non-recalcitrant C remaining in litter declined in the warmed plots compared to control. Consistent with this pattern, potential activities of enzymes targeting recalcitrant C increased with warming, relative to those targeting non-recalcitrant C. Even so, mass loss of individual C fractions showed that non-recalcitrant C is preferentially decomposed under control conditions whereas recalcitrant C losses remain unchanged between control and warmed plots. Moreover, overall mass loss was greater under control conditions. Our results imply that direct warming effects, as well as indirect warming effects (e.g. drying), may serve to maintain decomposition rates of recalcitrant C compared to non-recalcitrant C despite negative effects on overall decomposition.