The last century has seen a dramatic simplification of global landscapes, driven largely by the expansion and intensification of agriculture. Landscape simplification has known negative impacts on ecosystem health and function; however, less is known about how landscape simplification affects agricultural production. There is mounting field-scale evidence that simplification can reduce agricultural production by eroding the ecosystem processes on which agricultural systems depend; however, many of these processes emerge not at the field scale, but from complex interactions between land use, biophysical context, and human activity at the landscape scale. This research uses hierarchical Bayesian models to estimate the relationship between landscape-scale agricultural diversity and the yields of corn, soy, and winter wheat in the coterminous United States. We find that the yields of corn and winter wheat increase by as much as 20% in highly diversified agricultural systems. Our findings also indicate that (1) crop production is more responsive to the number of distinct crop types cultivated on a landscape than their cultivated extent and that (2) increasing diversity in agricultural systems that are already diverse brings the highest yield gains. Our models provide strong evidence at national and regional scales that agricultural diversification—an intervention with known ecosystem benefits—can increase crop production.