Topiramate reduces alcohol consumption in individuals who drink heavily. Candidate gene studies aimed at identifying genetic variants that predict topiramate's effects on drinking have yielded inconsistent findings. To identify genetic variation associated with treatment response, we conducted a genome‐wide association study (GWAS) among participants in the Million Veteran Program (MVP) who initiated topiramate treatment. Using electronic health records, we identified individuals who were dispensed topiramate for at least 60 days for any indication (i.e., the index event). Alcohol consumption was assessed using Alcohol Use Disorders Identification Test‐Consumption (AUDIT‐C) scores during the year prior to topiramate exposure (pre index) and at least 60 days after initiating topiramate but within 6 months of drug discontinuation (post index). The final GWAS sample included 8386 individuals who reported alcohol consumption (i.e., AUDIT‐C score > 0) during the preceding year. We calculated polygenic scores (PGS) for topiramate treatment response in the Yale‐Penn sample (n = 10,275) and examined associations with 692 phenotypes using a phenome‐wide association study. In the cross‐ancestry GWAS meta‐analysis, 35 loci had suggestive associations, though none reached genome‐wide significance. Topiramate response PGS had nominally significant associations with lower rates of alcohol‐related liver disease, older age at alcohol use disorder diagnosis, and higher frequency of alcohol use. Conclusions: Although no loci reached genome‐wide significance, the suggestive variants identified in the cross‐ancestry meta‐analysis are promising candidates for future investigation. Larger studies are needed to identify significant genetic predictors of topiramate response and advance precision medicine strategies for treating AUD.