This paper builds a model determining optimal capacities of diversion dams or water transfer projects. The model incorporates stochastic inflows to the dams and the role of the dam capacity in reducing overflows, and gives a closed-form expression of the marginal benefit of capacities. Comparative static analysis suggests that larger water projects could be required by 1) improvements in water management efficiency, 2) upward shifts in the marginal overflow-caused loss, or 3) more abundant inflows. The result provides important policy implications about the impact of integrated water reforms, rising concern about food security, and climate change on optimal water project capacities. The model is also applied to analyze the relation between water project capacities and conservation technologies, showing 1) that too large or too small water projects could discourage adopting conservation technologies, 2) that the impact of conservation technologies on optimal capacities is ambiguous, and 3) that if designers of water projects take water users' potential adoption of conservation technologies into account, the first-order condition of the capacity determination model could have multiple solutions.