We analyzed the predictive potential of pretreatment soluble carbonic anhydrase IX levels (sCAIX) for the efficacy of bevacizumab in the phase III neoadjuvant GeparQuinto trial. sCAIX was determined by enzyme‐linked immunosorbent assay (ELISA). Correlations between sCAIX and pathological complete response (pCR), disease‐free and overall survival (DFS, OS) were assessed with logistic and Cox proportional hazard regression models using bootstrapping for robust estimates and internal validation. 1,160 HER2‐negative patient sera were analyzed, of whom 577 received bevacizumab. Patients with low pretreatment sCAIX had decreased pCR rates (12.1 vs. 20.1%, p = 0.012) and poorer DFS (adjusted 5‐year DFS 71.4 vs. 80.5 months, p = 0.010) compared to patients with high sCAIX when treated with neoadjuvant chemotherapy (NCT). For patients with low sCAIX, pCR rates significantly improved upon addition of bevacizumab to NCT (12.1 vs. 20.4%; p = 0.017), which was not the case in patients with high sCAIX (20.1% for NCT vs. 17.0% for NCT‐B, p = 0.913). When analyzing DFS we found that bevacizumab improved 5‐year DFS for patients with low sCAIX numerically but not significantly (71.4 vs. 78.5 months; log rank 0.234). In contrast, addition of bevacizumab worsened 5‐year DFS for patients with high sCAIX (81 vs. 73.6 months, log‐rank 0.025). By assessing sCAIX levels we identified a patient cohort in breast cancer that is potentially undertreated with NCT alone. Bevacizumab improved pCR rates in this group, suggesting sCAIX is a predictive biomarker for bevacizumab with regards to treatment response. Our data also show that bevacizumab is not beneficial in patients with high sCAIX.