This paper documents large and persistent differences in R&D returns across listed US firms, with firms at the 75th percentile earning twice the median return. Further- more, returns are consistently larger for highly innovative firms with a large inventor workforce. Systematic R&D return differences are surprising as workhorse endogenous growth models predict that R&D resources flow from low to high returns firms until return equalization. I show that R&D return dispersion can reflect heterogeneity in firms’ market power over inventors in theory and provide evidence in favor of this hypothesis. I estimate that firms with high returns and those with a large inventor workforce face less elastic inventor supply, suggesting that they have more inventor market power. Calibrating a Schumpeterian growth model to match this evidence, I find that inventor monopsony can account for 1/3 of the documented R&D return dispersion and slows down growth by 4%, a welfare reduction of 2.1%.