Background Urban last-mile delivery has grown rapidly with urbanization and e-commerce, and is among the least environmentally sustainable stages of logistics, contributing disproportionately to greenhouse gas emissions, air pollution, and urban congestion. Electric vehicles are widely proposed as a solution, yet research has concentrated on passenger vehicles. How commercial carriers adopt them remains poorly understood, and adoption rates among major carriers differ sharply despite comparable access to the same vehicle technology. Purpose This study explains why major logistics carriers, all publicly committed to fleet electrification and operating at substantial scale, differ markedly in the pace and pattern of their EV adoption. It applies the Technology–Organization–Environment framework consistently across five carriers to examine how technological, organizational, and environmental conditions interact to constrain or enable adoption. Methods The study uses qualitative comparative document analysis. The corpus comprises 62 verified documents covering DHL, Amazon, UPS, FedEx, and Royal Mail from 2023 to early 2026: 55 company-produced disclosures, including annual reports, sustainability and ESG reports, and press releases, together with seven independent regulatory and industry benchmarking sources used for triangulation. Documents were verified against their originating publishers, and 353 extracts were coded against the TOE dimensions, with source and location recorded for each. Conclusion Adoption is not determined by technological, organizational, or environmental conditions independently, but by how they combine. Deployment scale does not track firm size or parcel volume, and 269 of 353 coded extracts span more than one TOE dimension. Regulatory pressure and public funding operate together rather than as alternatives, while organizational capacity can partly compensate for weaker environmental support. The study contributes a firm-level comparative application of TOE to commercial fleet electrification and shows that the binding constraint on adoption shifts with the surrounding conditions rather than remaining fixed.