Despite the widespread distribution of horses and their close interface with human environments, the equine antibiotic resistance gene (ARG) reservoir remains insufficiently characterized. This study aimed to comprehensively elucidate the equine resistome by analyzing 104 equine fecal metagenomes sampled across China, thereby establishing a critical baseline for antimicrobial stewardship in equine husbandry. Through a non-redundant gene catalog approach, we identified 3,264 unique ARG-related coding sequences (CDSs). To detect distant homologs, hidden Markov model (HMM) profiling was applied, revealing 2,420 putatively novel ARG-like sequences. Functional classification indicated that over 90% of these sequences mediate resistance through antibiotic inactivation or target protection, whereas efflux and impaired penetration accounted for less than 10%. Taxonomic profiling attributed these predicted sequences to 23 phyla and 146 genera, with Firmicutes, Bacteroidetes, and Proteobacteria identified as the detected taxa. Furthermore, we identified 731 plasmid-associated ARG-like candidates, of which 446 is novel variants, predominantly conferring resistance to macrolides, streptogramins, aminoglycosides, and β-lactams. Experimental validation confirmed that the novel gene abc-f(9) significantly reduced macrolide susceptibility in vitro, elevating the tilmicosin minimum inhibitory concentration (MIC) of Escherichia coli from 64 to 256 μg/mL. These findings provide the first systematic characterization of the equine ARG repertoire, significantly expanding the known global resistome. Our results offer a foundational resource for assessing the risks of antimicrobial resistance (AMR) transmission and implementing informed surveillance strategies in horse farming.