In many regions, groundwater resources play a pivotal role in sustaining irrigation for agricultural crop production. However, the future behavior, trajectory, and sustainability of these coupled agricultural-groundwater systems is subject to a host of hydrologic, geologic, and socioeconomic uncertainties. Here, we introduce a new large-scale coupled farm-groundwater modeling approach for exploring future outcomes in agricultural systems that rely on groundwater for irrigation. The model combines an agent-based positive mathematical programming representation of farmer cropping and irrigation decisions with a new cost-curve based method for simulating the evolution of groundwater production costs and availability, the latter adopting a Theis-based analytic formulation of groundwater drawdown evolution that is designed for computational efficiency and simplicity. The new model is deployed for simulating >50k representative farm-groundwater systems covering the entire continental United States for a wide range of potential hydrologic and economic futures. We further introduce a classification scheme to characterize patterns of response in these agricultural-groundwater systems based on groundwater sustainability and agricultural productivity outcomes.