We present an endogenous growth model with two sectors: a real sector where the final good is produced, and a banking sector that intermediates between savers and firms. Banking concentration exerts two opposite effects on growth. On the one hand, it induces economies of specialization, which is beneficial to growth. On the other hand, it results in duplication of banks' investment in fixed capital, which is detrimental to growth. The trade-off between the two opposing effects is ambiguous and can vary along the process of economic development. Hence, there is a potential nonlinear and nonmonotonic relationship between concentration and growth. We test this implication, using cross-country data on income and industry growth. We find that banking concentration is negatively associated with per-capita income growth and industrial growth only in low-income countries. This suggests that reducing concentration is more likely to promote growth in low-income countries than in high-income ones.