This paper examines a well known empirical puzzle in the literature on technology adoption: despite the potential of technologies to increase returns dramatically, a significant fraction of households do not use these technologies. I study the use of hybrid maize and fertilizer in Kenya, where there are persistent cross-sectional differences in aggregate adoption rates with a large fraction of households switching in and out of adoption. By allowing for selection of farmers into technology use via comparative advantage differences, I examine whether the yield returns to adopting hybrid maize vary across farmers. If so, high average returns can coexist with low returns for the marginal farmer. My findings indicate the existence of two interesting subgroups in the population. A small group of farmers has potentially high returns from adopting the technologies. Yet, they do not adopt. This lack of adoption appears to stem from supply and infrastructure constraints, such as the distance to fertilizer distributors. In addition, a larger group of farmers faces very low returns to adopting hybrid maize, but chooses to adopt. This latter group might benefit substantially from the development of newer hybrid strains to increase yields. On the whole, the stagnation in hybrid adoption does not appear to be due to constraints or irrationalities.