Highly active antiretroviral therapy (HAART) prolongs short-term survival in patients with HIV/AIDS. HAART has only been available since 1996; thus, no long-term survival data are available. Computer simulation models extrapolating short-term survival data can provide estimates of long-term survival. These survival estimates may assist patients and clinicians in HAART treatment planning. The authors construct a computer simulation model based on observational data to estimate long-term survival in a cohort of HIV/AIDS patients undergoing treatment with HAART. The authors use data from the Collaboration in HIV Outcomes Research-US (CHORUS) observational cohort (N = 4791), the published literature, and US Life Tables to specify a computer simulation model of expected survival accounting for baseline CD4 cell count, progressive HAART treatment failure, progressive risk of HAART on treatment mortality, and age-associated mortality. Time to treatment failure for each of three rounds of HAART and risk of mortality on-treatment were estimated using parametric survival models with censoring of follow-up fit to CHORUS data. Off-treatment survival after HAART failure was estimated from the pre-HAART literature. Age-associated mortality was taken from US Life Tables. Median projected survivals stratified by baseline CD4 cell count subgroups were CD4 > 200 cells/mm3, 15.4 years; CD4 < or = 200 cells/mm3, 8.5 years; and CD4 < or = 50 cells/mm3, 5.5 years. These values are 4 to 6 years longer than pre-HAART cohorts. The sensitivity analyses showed that the model survival predictions were most sensitive to the treatment failure rate, the on-treatment mortality rate, and the number of treatment rounds. Computer simulation modeling of long-term survival of patients with HIV/AIDS on HAART--accounting for differential treatment failure and death rates stratified by CD4 cell count and age-associated mortality--suggests a relatively consistent 4- to 6-year survival benefit over pre-HAART therapies.