Objective: Quantify incidence and determine predictors of Gram-negative bacilli (GNB) in people who inject drugs (PWID) with injection-drug use (IDU)-related infections. Design: Retrospective cohort of hospitalized PWID from 1/2017-12/2019. Methods: Inclusion criteria: age ≥18 years, active IDU, treated IDU-attributable infection, organism growth from microbiology cultures. Infection types: infective endocarditis (IE), acute bacterial skin/skin structure infection (ABSSSI), osteoarticular infection (OAI), other bloodstream infections (BSI). Primary outcome was GNB identification from microbiologic culture; descriptive statistics were used to describe the cohort. Multivariable regression was used to identify variables associated with GNB infection. Results: 230 PWID included; 65 (28%) GNB infections, 165 (72%) Gram-positive infections. The median (IQR) population age was 38 (31-45) years. Most patients were women (56%); 37% had no insurance. Infection types were: IE (41%), ABSSSI (37%), OAI (20%), other BSI (2%). 278 organisms were isolated from 230 patients; most common organisms were methicillin-resistant Staphylococcus aureus (43%), Streptococcus spp. (19%), methicillin-susceptible S. aureus (17%), Serratia marcescens (8%); 10% were mixed GNB and Gram-positive infections. 80% of patients received empiric Pseudomonas aeruginosa coverage; only 7% had P. aeruginosa infections. In multivariable regression, age >50 years (adjOR, 2.9; 95%CI; 1.2-7.2), prior hospitalization within 90-days (adjOR, 2.2; 95%CI; 1.2-4.3), and OAI (adjOR, 3.2; 95%CI; 1.5-6.6) were associated with GNB infection. Conclusions: GNB in PWID with IDU-attributed infections were more frequently observed in recently hospitalized, older patients with OAI. The majority of patients received empiric anti-pseudomonal antibiotic coverage, but P. aeruginosa was infrequent. PWID are a potential population to target improved empiric antibiotic use.