Shifting cultivation dominates many tropical forest regions. It is expanding into old‐growth forests, and fallow period duration is rapidly decreasing, limiting secondary forest recovery. Shifting cultivation is thus a major driver of carbon emissions through deforestation and forest degradation, and of biodiversity loss. The impacts of shifting cultivation on carbon stocks have rarely been quantified, and the potential for carbon‐based payments for ecosystem services (PES ), such as REDD +, to protect carbon in shifting cultivation landscapes is unknown. We present empirical data on aboveground carbon stocks in old‐growth forest and shifting cultivation landscapes in northeast India, a hotspot of threatened biodiversity. We then model landscape‐level carbon stocks under business‐as‐usual scenarios, via expansion into the old‐growth forest or decreasing fallow periods, and intervention scenarios in which REDD + is used to either reduce deforestation of primary or secondary forest or increase fallow period duration. We found substantial recovery of carbon stocks as secondary forest regenerates, with a 30‐yr fallow storing about one‐half the carbon of an old‐growth forest. Business‐as‐usual scenarios led to substantial carbon loss, with an 80% reduction following conversion of old‐growth forest to a 30‐yr shifting cultivation cycle and, relative to a 30‐yr cultivation landscape, a 70% reduction when switching to a 5‐yr cultivation cycle. Sparing old‐growth forests from deforestation using protected areas and intensifying cropping in the remaining area of shifting cultivation is the most optimal strategy for carbon storage. In areas lacking old‐growth forest, substantial carbon stocks accumulate over time by sparing fallows for permanent forest regeneration. Successful implementation of REDD + in shifting cultivation landscapes can help avert global climate change by protecting forest carbon, with likely co‐benefits for biodiversity.