We study sequential contributions to public goods in a decentralized environment in which commitment to a contribution schedule is not feasible. A natural (partial) solution to dynamic free-riding incentives in this context is for agents to alternate in making small steps towards completion of the project, dividing the larger project into smaller parts. In this paper, we consider a model in which agents with different valuation for the good are selected at random to contribute in each period. We show that if the project is sufficiently large, the unique equilibrium of the model displays endogenous contribution cycles, in which agents of different types alternate making gradual contributions towards the completion of the project. We characterize these cycles in terms of the primitives of the model, and study the efficiency of equilibrium outcomes.