This paper presents a mathematical model of the HIV/AIDS epidemic in Kunming, the provincial capital of Yunnan, China. The population is divided into several groups, with individuals possibly changing group. Two transmission routes of HIV are considered: needle sharing between injecting drug users (IDUs) and commercial sex between female sex worker (FSWs) and clients. The model includes male IDUs who are also clients and female IDUs who are also FSWs. Groups are split in two--risky and safe--according to condom use and needle sharing. A system of partial differential equations is derived to describe the spread of the disease. For the simulation, parameters are chosen to fit as much as possible data publicly available for Kunming. Some mathematical properties of the model--in particular the epidemic threshold R0 which determines the goal of public health interventions--are also presented. Though the model couples two transmission routes of HIV, the approximation R0 approximately = max[R0(IDU), R0(sex)], with closed formulas for R0(IDU) and R0(sex), appears to be quite good. The critical levels of condom use and clean needle use necessary to stop both the sexual transmission and the transmission among IDUs can therefore be determined independently.