Reducing Emissions from Deforestation and Degradation (REDD+) in developing countries is based on the premise that conserving tropical forests is a cost-effective way to reduce carbon emissions and therefore can be fully funded by international actors with obligations or interests in reducing emissions. However, concerns have repeatedly been raised about whether stakeholders in REDD+ host countries will actually end up bearing the costs of REDD+. Most prior analyses of the costs of REDD+ have focused on the opportunity costs of foregone alternative uses of forest land. We draw on a pan-tropical study of 22 subnational REDD+ initiatives in five countries to explore patterns in implementation costs, including which types of organizations are involved and which are sharing the costs of implementing REDD+. We find that many organizations involved in the implementation of REDD+, particularly at the subnational level and in the public sector, are bearing implementation costs not covered by the budgets of the REDD+ initiatives. To sustain this level of cost-sharing, REDD+ must be designed to deliver local as well as global forest benefits.