BACKGROUND: Routine national incidence testing with enzyme immunoassay for recent HIV-1 infections (EIA-RI) has been done in France since January, 2003. From the reported number of HIV infections diagnosed as recent, and accounting for testing patterns and under-reporting, we aimed to estimate the incidence of HIV infection in France in 2003-08. METHODS: We analysed reports from the French National Institute for Public Health Surveillance for patients who were newly diagnosed with HIV between January, 2003, and December, 2008. Missing data were imputed with multiple imputation. Patients were classified with non-recent or recent infection on the basis of an EIA-RI test, which was calibrated with serial measurements from HIV seroconverters from the French ANRS-PRIMO cohort. We used an adapted stratified extrapolation approach to calculate the number of new HIV infections in men who have sex with men (MSM), injecting drug users (IDUs), and heterosexual men and women by nationality. Population sizes were obtained from the national census and national behavioural studies. FINDINGS: After accounting for under-reporting, there were 6480 (95% CI 6190-6780) new diagnoses of HIV infection in France in 2008. We estimate that there were 6940 (6200-7690) new HIV infections in 2008, suggesting an HIV incidence of 17 per 100000 person-years. In 2008, there were 3550 (3040-4050) new infections in heterosexuals (incidence of 9 per 100000 person-years), 3320 (2830-3810) in MSM (incidence of 1006 per 100000 person-years), and 70 (0-190) in IDUs (incidence of 86 per 100000 person-years). Overall HIV incidence decreased between 2003 and 2008 (p