Abstract Background and Aims With an epidemic of drug overdoses, contemporary research is needed to examine drug overdose deaths among homeless populations. This study measured rates, types and correlates of drug overdose deaths occurring over a 5‐year study period among veterans experiencing homelessness (VEH) and non‐homeless veterans (NHV) in the US Department of Veterans Affairs (VA) healthcare system. Design Retrospective cohort study. Setting USA. Participants A total of 6 128 921 veterans. We followed 399 125 VEH and 5 729 796 NHV between 2017 and 2021 using linked administrative VA and National Death Index data. Measurements Multivariable Cox regression models were constructed to estimate hazard ratios (HRs) for homelessness as a predictor of time‐to‐occurrence of overdose deaths with 95% confidence interval (CIs), sequentially controlling for demographic, medical, substance use and mental health characteristics. Findings Among overdose deaths, 8653 [93.7%, 95% confidence interval (CI) = 93.2–94.2%] were unintentional and 5378 (57.8%, 95% CI = 56.8–58.8%) involved opioids. The overdose‐specific mortality rate (per 100 000 person‐years) was 227.3 among VEH and 23.2 among NHV (HR = 9.76, 95% CI = 9.36, 10.16), with rates 7–14 times higher among VEH than NHV, depending on the drug involved. In fully adjusted models, homelessness was associated with greater risk of drug overdose death (HR = 3.33, 95% CI = 3.18, 3.49), with the greatest risk involving psychostimulants (HR = 4.19), followed by antiepileptic/sedative/hypnotic drugs (HR = 3.69), synthetic opioids (HR = 3.50) and natural and semi‐synthetic opioids (HR = 2.79). Conclusions US veterans experiencing homelessness appear to have three times the risk for drug overdose deaths than non‐homeless veterans. There may be specific risks associated with psychostimulant, antiepileptic, sedative and hypnotic drugs in this population that deserve greater attention.