Short and medium term projections of AIDS incidence and HIV back-projections can be estimated using different techniques. The simplest technique for projection of AIDS cases is extrapolation by curve fitting applied to past AIDS incidence data, corrected for reporting delays. A severe criticism of this method is the total dependency of the projections on the chosen function for the curve. A second technique which incorporates more knowledge of the transmission mechanisms of the epidemic, uses the incubation time distribution, sometimes including data on early treatment with anti-viral drugs, to back-project the HIV incidence from the delay corrected AIDS incidence, and projects forwards the minimum expected AIDS incidence. Projections depend on the form of the chosen incubation time distribution, and moreover, the interplay of behavioural change, early treatment and other interventions all make the use of the back-projection method more difficult in the future. If still more information is used on the mechanism of the epidemic, i.e. not only incubation time distribution but also assumptions on the spread of infection through the population, more detailed projections can be made by the use of deterministic or stochastic models on restricted risk groups and relatively small populations e.g. Switzerland, San Francisco or Amsterdam. In this paper, a model developed earlier for the risk group of homo/bisexual men in Amsterdam (Heisterkamp et al. 1992) is adapted for the inclusion of early treatment effects and used for the projection of AIDS and HIV incidence for the same transmission group although for a risk group on a larger scale.