Death from fentanyl-related overdose is now a leading cause of mortality among US adults. We sought to characterize fentanyl preference among street-based people who use drugs (PWUD).Cross-sectional surveys were administered to PWUD (N = 308) who illicitly used heroin or prescription opioids in the prior six months. Recruitment occurred in 2017 in three US east coast cities with high overdose mortality: Baltimore, Boston, and Providence. Our main outcome was preference for fentanyl (yes/no); exposures included sociodemographics, drug use, and overdose history. Pearson's χ2, Shapiro-Wilk-Mann rank-sum tests, and tiered log-binomial regression determined sociodemographic and exposure-related factors associated with fentanyl preference.Preference for nonmedical use of fentanyl was reported by 27% (n = 83) of the sample. Fentanyl preference was associated with non-Hispanic white race (adjusted risk ratio (ARR) = 1.68, 95% confidence interval (CI):1.18-2.40), daily illicit drug use (aRR = 2.2, CI:1.71-2.87), and overdose ≥1 year ago (aRR = 1.33, CI:1.18-1.50). Age (in decades; aRR = 0.77, CI:0.61-0.98) and overdose <1 year ago (aRR = 0.92, CI:0.87-0.97) were associated with a decreased likelihood of preference. In our model excluding sociodemographics, initiating opioid use with non-prescribed opioids was associated with fentanyl preference (aRR = 1.48, CI:1.26-1.73).In three cities with high levels of opioid use and overdose, a quarter of street based PWUD reported preferring fentanyl. An opioid use age cohort effect and disproportionate access to prescription opioids by race could be contributing to preference. Frequency of opioid use, not route of administration, was associated with preference. Our data demonstrate the need to consider preferences for fentanyl when targeting services and interventions for PWUD.