The androgen receptor (AR) has been shown to be of potential prognostic importance in retrospective cohorts. We evaluated immunohistochemical AR expression on a tissue microarray of 673 core biopsies from primary breast cancer patients treated with neoadjuvant docetaxel/doxorubicin/cyclophosphamide (TAC) chemotherapy in the prospective GeparTrio phase-III trial. AR was detected in 53.2% of tumours. Lowest AR expression was detected in triple-negative breast cancers (TNBC) with 21.2%. Highest AR expression was observed in Luminal A-like tumours with 67%. In AR-positive tumours, pathological complete response (pCR) rate was 12.8% compared to 25.4% in AR-negative tumours (P < 0.0001). In multivariate analysis, AR independently predicted pCR (OR 1.86; 95% CI [1.16-2.79] P = 0.0086). Overall patients with an AR-positive tumour had a significant better disease-free (DFS) (AR-positive 78.9% vs. AR-negative 72.5%; log-rank P = 0.0329) and overall survival (OS) (88.8% vs. 82.7%; log-rank P = 0.0234) than those with AR-negative tumours. Stratified analysis revealed that in the TNBC subgroup, but not in the other subgroups defined by ER, PgR and HER2, AR expression predicted a better DFS (AR-positive 85.7% vs. AR-negative 65.5% log-rank P = 0.0544) and OS (95.2% vs. 76.2%; log-rank P = 0.0355). Within the non-pCR subgroup, AR positivity selected a group with a significant better DFS (P = 0.045) and OS (0.021) but not within the pCR group. Patients with an AR-negative tumour have a higher chance of achieving a pCR than those with an AR-positive one. But, patients with AR-positive tumours have a better survival especially if they did not achieve a pCR.