Abstract Objectives Most infections in patients with RA are treated in primary care with antibiotics. A small fraction require hospitalization. Only a few studies exist regarding the overall risk of infection (i.e. prescription of antibiotics or hospitalization due to infection) in patients initiating non-TNF-inhibitor therapy. In Danish RA patients initiating abatacept, rituximab and tocilizumab treatment in routine care, the aims were to compare adjusted incidence rates (IR) of infections and to estimate relative risk of infections across the drugs during 0–12 and 0–24 months. Methods This was an observational cohort study including all RA patients in the DANBIO registry starting a non-TNF-inhibitor from 2010 to 2017. Infections were defined as a prescription of antibiotics or hospitalization due to infection. Prescriptions, comorbidities and infections were captured through linkage to national registries. IRs of infections (age, gender adjusted) and rate ratios (as estimates of RR (relative risk)), adjusted for additional covariates) (Poisson regression) were calculated. Results We identified 3696 treatment episodes (abatacept 1115, rituximab 1017, tocilizumab 1564). At baseline, rituximab users were older and had more previous cancer. During 0–12 months, 1747 infections occurred. Age and gender-adjusted IRs per 100 person-years were as follows: abatacept: 76 (95% CI: 69, 84); rituximab: 87 (95% CI: 79, 96); tocilizumab: 77 (95% CI: 71, 84). Adjusted RRs were 0.94 (95% CI: 0.81, 1.08) for abatacept and 0.94 (95% CI: 0.81, 1.03) for tocilizumab compared with rituximab and 1.00 (95% CI: 0.88, 1.14) for abatacept compared with tocilizumab. RRs around 1 were observed after 24 months. Switchers and ever smokers had higher risk compared with biologic-naïve and never smokers, respectively. Conclusion Overall infections were common in non-TNF-inhibitor-treated RA patients, with a tendency towards rituximab having the highest risk, but CIs were wide in all analyses. Confounding by indication may at least partly explain any differences.