There is growing empirical evidence that individuals differ in the returns that they make on their investments, and that this is importantly driving inequality in capital income. Return heterogeneity might originate from underlying ability heterogeneity — more able people are better in investing — or from underlying wealth heterogeneity — higher wealth `buys’ higher returns through access to better investment vehicles. We determine the implications of such return heterogeneity for optimal labor- and capital-income taxes. We show that marginal tax rates on capital income are optimally positive when returns are heterogeneous and increasing with individuals’ income. The reason for this is twofold. (i) First, if high-ability individuals are simply more able investors, capital income provides information on individuals’ underlying ability even conditional on labor income. The tax on capital income then provides redistributive benefits over and above those of a tax on labor income. (ii) Second, even if ability does not directly feed into returns, but if higher wealth buys higher returns, the government would optimally want to redistribute income from rich to poor late in life rather than early in life. This would encourage savings of high-return rich people and discourage savings of low-return poor people and thereby achieve a more efficient allocation of savings. As capital income is typically earned later in life than labor income, a capital-income tax is more effective in redistributing late in life than the labor-income tax. Numerical simulations show that moderate degrees of return heterogeneity result in significant optimal capital-income taxes.