A clear understanding of geographic distribution of HIV-infected people and maintaining up-to-date lists and locations of facilities providing HIV-related services are essential for monitoring the epidemic and for providing treatment, care, and support services to the infected and their families. In this study, we model and map human immunodeficiency virus (HIV) prevalence in Kenya in relation to its spatial and behavioral determinants, using data from the 2003 Kenya Demographic and Health Survey (DHS). The 2003 Kenya DHS is one of the first population-based national surveys to link individual HIV test results for both males (age 15 - 54) and females (age 15 - 49) with the full set of behavioral, social, and demographic indicators included in the survey. The survey also collected spatial coordinates of the communities where survey respondents lived. These coordinates have been used to estimate spatial indicators such as distance to roads, distance to Lake Victoria, and population density. Using these spatial, social, demographic, and behavioral indicators, we developed a model to predict HIV prevalence. We apply this model to map HIV concentration areas at sub-provincial level, and we assess the existing HIV service coverage in relation to the spatial distribution of HIV prevalence. The study finds large subregional variations in the prevalence of HIV in Kenya. Areas of high concentration of HIV-infected people have a disproportionately low density of HIV-related services.