Background: Checkpoint inhibitors are transforming cancer care. However, the high prices of these medicines raise concerns over their affordability and disparity in use. The objective of this study is to describe the disparity in initiating checkpoint inhibitors and examine patient- and area-level factors associated with delayed initiation. Methods: This study is a retrospective cohort study using Optum data. We identified commercially insured patients newly diagnosed with metastatic lung cancer and melanoma since the introduction of checkpoint inhibitors in these cancers (lung cancer cohort: diagnosed between January 2015 and December 2020; melanoma cohort: diagnosed between January 2011 and December 2020). Time from metastatic cancer diagnosis to initiating checkpoint inhibitors was analyzed using Cox proportional hazard models. Independent variables included county-level measures (percentage of black population, percentage of Hispanic population, percentage of other minority, percentage of population living below poverty line, rurality, number of medical oncologist per population, and having a National Cancer Institute designated cancer center) and patient-level characteristics (age, sex, Charlson comorbidity index, any dual eligibility, Medicare Advantage, and year of diagnosis). We clustered standard errors at the county level. Results: The percentage of metastatic lung cancer and metastatic melanoma patients on checkpoint inhibitors increased from 23% to 52% from 2015 to 2020 and from 22% to 58% from 2011 to 2020. Counties with greater percentage of black, Hispanics, and other minorities were high urban with greater density of medical oncologists and NCI-designated cancer centers. However, greater percentage of Hispanic population in a county was associated with significantly slower initiation of checkpoint inhibitors for both the lung cancer and the melanoma cohorts (hazard ratios [HR]: 0.937 and 0.946, respectively; p-values: < 0.001 and 0.014, respectively). Percentage of other minority population in a county was associated with slower initiation for metastatic lung cancer (HR: 0.983; p-value: < 0.001). No other county-level factors had a significant coefficient from the multivariate Cox models. In terms of patient-level characteristics, older age, female, more comorbidities, any dual eligibility, and Medicare Advantage were associated with significantly slower initiation for the lung cancer cohort and older age and female were associated with significantly slower initiation for the melanoma cohort. Conclusions: Commercially insured metastatic lung cancer and melanoma patients who lived in counties with greater percentage of Hispanic population had slower initiation of checkpoint inhibitors after their cancer diagnosis, despite the fact that those counties had greater density of medical oncologists and NCI-designated cancer centers.