Prevention of cardiovascular disease is a major issue in the current management of people living with HIV. Concern is growing about the metabolic impact of integrase strand transfer inhibitors (INSTIs), which could lead to an increased risk of diabetes, but the data are conflicting. This is an updated version of our previous analysis, with longer follow-up and new molecules. We retrospectively evaluated the incidence of new-onset diabetes in people living with HIV starting combined antiretroviral therapy with an INSTI compared with non-nucleoside reverse transcriptase inhibitors and protease inhibitors. Data were collected from the Dat'AIDS cohort study, a collaboration of 30 HIV treatment centres in France. We used a propensity score-based inverse probability of treatment weighting approach to adjust for baseline characteristics between the two groups (INSTI and non-INSTI). Between 2009 and 2021, a total of 12 150 people living with HIV were included. The incidence of diabetes was higher in the INSTI group than in the non-INSTI group (hazard ratio 1.38; 95% confidence interval 1.07-1.77; p = 0.012). Regardless of the third drug, but to a greater extent for INSTIs, we observed a peak of new-onset diabetes in the year following initiation of combined antiretroviral therapy. The incidence of diabetes was higher in people treated with integrase inhibitors than in those receiving other third agents. This increased risk occurred both during the first year of treatment and in the longer term.