MHC polymorphism is explained by natural selection driven by the MHC-dependent impact of certain infections, inflammatory conditions, autoimmune diseases, and cancers. However, examples of human disease driving this process are rare. We evaluated the impact of HIV-1 in altering HLA-I frequencies in KwaZulu-Natal, South Africa, and the influence of antiretroviral therapy (ART) on this process. In a historical mother-child cohort in the pre-ART era (1998–2005), HIV-1 survival and vertical transmission were both strongly HLA-B dependent: “disease-susceptible” HLA-B alleles (HLA-B*18/B*45:01/B*58:02) increased adult AIDS progression and vertical transmission (OR 1.6, P = 0.01), whereas “protective” HLA-B alleles (HLA-B*57/B*58:01/B*81:01) slowed AIDS progression, and decreased vertical transmission (OR 0.57, P = 0.002). By contrast, in contemporary antenatal KwaZulu-Natal cohorts in the ART era (2015–2025) the impact of HLA-B on HIV-1 disease outcome and vertical transmission is dramatically reduced. Using these and reported data, we constructed a model to estimate the impact of HIV-1 on HLA-B frequencies in KwaZulu-Natal, both in the prevailing setting of ART and in a hypothetical counterfactual scenario where ART was never rolled out. Over the 45-y period 1990–2035, in the absence of ART, the proportion of the population possessing any “protective” HLA-B allele was projected to increase from 23 to 42% (allele frequencies increasing from 0.12 to 0.24), and the proportion of the population possessing any “disease-susceptible” HLA-B allele was projected to decrease from 28 to 18% (allele frequencies declining from 0.15 to 0.092). The introduction of ART radically slows HLA-B frequency change. These data therefore demonstrate the potential for natural selection from an infectious disease to alter human population genetics within decades, and for the successful roll-out of therapy to halt this process.