The epidemic in South Africa is characterized by high genetic diversity driven by multiple independent introductions. The bulk of these introductions occurred between 1985-2000 during which time HIV prevalence increased exponentially. Epidemic growth has stabilized in recent years with the implementation of several interventions. Here we identified distinct HIV clades from a large sequence dataset of southern African HIV sequences (n = 15,332). Each clade was characterized using phylodynamic and phylogeographic methods to infer their growth through time and space. The estimated date of origin for the 18 clades that were found, fell between 1979-1992 with strong growth during the 1990's. Phylogeographic reconstruction revealed wide dispersal of clades throughout the country with the city of Johannesburg as the focal point of viral dispersal. We found clear signs of decreasing growth rate in four of the clades since the advent of interventions, while other clades have continued to growth and expand. Our results demonstrate that interventions do not affect the HIV epidemic universally with major difference between different clades over time and space. Here we demonstrate the utility and flexibility of molecular epidemiological methods and demonstrate how they can potentially be a powerful tool in HIV epidemic monitoring in South Africa.