In this paper, we consider a multiple-intelligent reflecting surface (IRS)-aided multi-user multiple-input-multiple-output (MU-MIMO) down-link communication system, wherein a base station (BS), equipped with multiple antennas, serves massive single-antenna users. In order to mitigate high inter-user interference, we divide all the users into a single near-user (NU) group and multiple cell-edge user (CEU) groups based on their connection with BS. In particular, we assign a single IRS for each CEU group such that the BS communicates and serves the set of CEUs using their corresponding IRS. Subsequently, we formulate a problem of energy-efficient resource allocation design subject to optimal beamforming design at BS and multiple IRS under given power constraints. Due to the non-convexity of the formulated resource allocation problem, we proposed two different alternating optimization frameworks which solve the joint optimization problem based on 1) minimum mean square error (MMSE) and Riemannian conjugate gradient (RCG) and 2) fractional programming and successive convex approximation techniques for active beamforming vectors at the BS and passive beamforming vectors for IRSs, respectively. Simulation results confirm that the MMSE-RCG-based resource allocation design attains fast convergence when compared to sequential fractional programming (SFP) based solution. Moreover, the MMSE-RCG attains better energy efficiency whereas the SFP renders better spectral performance. Nevertheless, both the proposed solution achieve 70%–80% better performance than random phase-shift settings.