Vancomycin pharmacokinetic (PK) and pharmacodynamic (PD) data in neonates are based on total concentrations. However, only unbound vancomycin is pharmacologically active. The objective was to determine vancomycin protein binding and the covariates impacting unbound vancomycin concentration in neonates and young infants. In neonates and young infants to whom vancomycin was administered intermittently for medical indications, total and unbound vancomycin plasma concentrations were determined using LC-MS/MS. Sampling occurred randomly during vancomycin exposure, covering a broad range of concentrations. Impact of covariates on unbound vancomycin concentration was determined using linear regression. Significant results of the univariate regressions were entered in a stepwise multiple regression. Passing-Bablok regression and Bland-Altman were used to assess the difference between measured and calculated unbound vancomycin concentration. Thirty-seven samples in 33 patients (median (interquartile range) gestational age 35 (29–39) weeks) were collected. Median total and unbound vancomycin concentrations were 14.2 (7.4–20.6) and 13.6 (7.2–22.5) mg/L, respectively. Median unbound fraction was 0.90 (0.77–0.98). Multiple regression revealed total vancomycin concentration (β = 0.884, p < 0.001) and albumin (β = − 0.323, p = 0.007) as most important covariates of unbound vancomycin concentrations, with an R2 adjusted of 0.953 (p < 0.0001). Mean absolute difference between calculated and measured unbound vancomycin was − 0.008 (95% CI − 0.92–0.91) mg/L. The unbound vancomycin fraction in neonates is higher compared to that in children and adults, and total vancomycin concentration and albumin were the most important covariates of unbound vancomycin concentration. Integration of protein binding in future PK/PD analyses is appropriate to optimize vancomycin dosing and to determine population-specific vancomycin PD targets for neonates.