The purpose of the study was to elucidate the role of expertise on muscle synergies involved in bench press. Ten expert power lifters (EXP) and nine untrained participants (UNT) completed three sets of eight repetitions at 60% of three repetition maximum in bench press. Muscle synergies were extracted from surface electromyography data of 21 bench press cycles using non-negative matrix factorization algorithm. The synergy activation coefficient represents the relative contribution of the muscle synergy to the overall muscle activity pattern, while the muscle synergy vector represents the relative weighting of each muscle within each synergy. Describing more than 90% of the variability, two muscle synergies reflected the eccentric and concentric phase. The cross-correlations (ρmax) for synergy activation coefficient 2 (concentric phase) were 0.83 [0.71;0.88] and 0.59 [0.49;0.77] [Median ρmax (25th;75th percentile)] (P = 0.001) in UNT and EXP, respectively. Median correlation coefficient (ρ) for muscle synergy vector 2 was 0.15 [−0.08;0.46] and 0.48 [0.02;0.70] (P = 0.03) in UNT and EXP, respectively. Thus, EXP showed larger inter-subject variability than UNT in the synergy activation coefficient during the concentric phase, while the muscle synergy vectors were less variable in EXP. This points at the importance of a specialized neural strategy in elite bench press performance.