The effect of HIV treatment on hospitalization rates for HIV-infected people has never been established. We quantified this effect in a rural South African community for the period 2009-13. We linked clinical data on HIV treatment start dates for more than 2,000 patients receiving care in the public-sector treatment program with five years of longitudinal data on self-reported hospitalizations from a community-based population cohort of more than 100,000 adults. Hospitalization rates peaked during the first year of treatment and were about five times higher, compared to hospitalization rates after four years on treatment. Earlier treatment initiation could save more than US$300,000 per 1,000 patients over the first four years of HIV treatment, freeing up scarce resources. Future studies on the cost-effectiveness of HIV treatment should include these effects.