Background: Bladder carcinomas (BCs) is featured by high incidence of recurrence which remains to be a big clinical challenge. This study aimed to compare the genomic profiles of primary and recurrent BCs in advanced stage. Methods: Primary and recurrent tissues (and control blood) were collected and somatic variants including short variants, copy number variations, gene rearrangements and TMB (Tumor Mutation Burden) were analyzed following the CAP/CLIA-certificated workflows. Chi-Square Test/Fisher Test was applied to compare the difference between two variables. Results: To eliminate the impact of stage, only patients in advanced stages were included (stage III-IV) in this study. 53 patients were investigated, including 43 primary and 10 recurrent (evidence of tumor observed after radical treatment). It should be noted that matched primary and recurrent samples were not obtained from the same patient. Median age of the patients was 64 years old. The proportion of male patients was 86% (n = 37) in the primary cohort and 60% (n = 6) in the recurrent cohort. Among the primary cases, the most frequent alterations were detected in TP53 (58.1%), TERT (51.2%), KMT2D (27.9%), KDM6A (23.3%) and RB1 (23.3%). Among the recurrent cases, the most frequent alternations were showed in TERT (70%), KMT2D (50%), PIK3CA (50%), TP53 (50%) and KDM6A (40%). PIK3CA substitution and KRAS substitution showed significant difference when comparing between the two cohorts. PIK3CA substitution preferentially occurred in recurrent samples (50% recurrent vs 14% primary, p-value = 0.01), mostly occurred in E545K; while KRAS substitution showed 4.7% in primary and 30% in recurrent with p-value = 0.04, all presented as G12V. TMB showed no significant difference between the two cohorts (6.2 vs 6.3 muts/Mb, p-value = 0.72). Conclusions: By NGS, we observed that substitutions in PIK3CA and KRAS showed significant difference between primary and recurrent samples from advanced BCs patients. PIK3CA and KRAS mutations were reported to be associated with local recurrence in CRC, but no association was observed in early stage non-muscle-invasive BCs. Our findings may provide a different insight into the treatment selections for advanced BCs, although the sample size is small.