Water markets may help societies adapt to rising water scarcity and variability, but their setup costs can be substantial and their benefits uncertain. I estimate the gains available from strengthening the wholesale surface water market in California, where conveyance infrastructure is well-developed yet transaction volume remains low. To do so, I develop a new empirical framework to analyze welfare in water markets that uses transactions data. First, I recover marginal valuations of water in the presence of unobserved transaction costs, by using particular price comparisons to find the incidence of both known and unknown cost determinants. Second, I estimate demand using yearly water endowments, which have rich variation driven by weather and amplified by historical rules. Then, I combine this demand model with a hydrological network model to simulate counterfactual outcomes. I find that efficient trading across regions and sectors would achieve benefits of only $86 to $278 million per year, without accounting for any environmental costs. These results suggest that promoting large-scale water markets may not achieve large gains without also reforming the policies and institutions that govern local water allocation.