Technology adoption in follower countries can be accomplished by local R&D activities, but it can also be achieved without formal R&D, for example, by foreign direct investment. Empirical evidence suggests that current R&D activities often expand local knowledge for future R&D, while adoption without R&D does not seem to have this effect. We formalize this idea in a quality-ladder growth model and find that this biased externality results in multiple steady states: In the long run, countries with sufficient initial knowledge and human capital converge to a state in which R&D is locally undertaken and thus become relatively rich, while other countries fully rely on technology adoption without R&D and stay poor. Switching regression using cross-country data supports the presence of multiple steady states in R&D expenditures.