OBJECTIVES: Modelling and forecasting the consequences of AIDS on economic growth in Africa. METHODS: Our model is based on two crucial hypothesis: AIDS has a short-term impact on a flow variable (the flow of labour available and capable of working at a moment t in the economy); AIDS has a long-term impact on stock variables (human capital, i.e. the stock of health or the stock of education and competence incorporated in the workers; and physical capital). The first effect is generally taken into account in the existing literature but the second is missing. Data from Ivory Coast (data given by UNAIDS and WHO) about macroeconomic variables and epidemiologic situations are used to calibrate our model. RESULTS: Integrating these two impacts in a model of growth with multiple factors of accumulation is sufficient to reverse the standard impact-evaluations based on classical tools. We show that, under a realistic range of epidemiological shocks, an involution trap can appear, corresponding to a modification of the long-term growth regime of the economy. CONCLUSIONS: When the long term impact of AIDS is taken into account, the impact of AIDS on economic growth in Africa is far worse than predicted by UNAIDS for instance. In Ivory Coast, the GDP loss is about 15% of the no-AIDS scenario in 2010 (against -8% for the UNAIDS forecast).