How do poor entrepreneurs trade off investments in business enterprises versus children's human capital, and how do these choices influence intergenerational socio-economic mobility? To examine this, we exploit experimental variation in household income resulting from a one-time relaxation of household liquidity constraints (Field et al., 2013), and track schooling and business outcomes over the subsequent 11 years. On average, treatment households, who were made wealthier through the experiment, increase human capital investment such that their children are 35% more likely to attend college. However, schooling gains only accrue to children with literate parents, among whom college attendance nearly doubles. In contrast, treatment effects on investment among the illiterate accrue only on the business margin and are accompanied by adverse educational outcomes for children. As a result, treatment lowers relative educational mobility. In a forecasting exercise, we find that earnings gains for literate households are four times larger than the earnings gains for illiterate households, raising earnings inequality. Our findings highlight how parental investment choices can contribute to a growth in intergenerational earnings inequality despite reductions in urban poverty. [Additional funding was provided by PEDL, IPA SME, and WAPP Harvard.]