This paper proposes a computationally efficient beamforming scheme for massive multiple input multiple output (MIMO) systems by reducing the complexity of uplink-side power optimization. Unlike conventional fixed-point methods that require intensive iterations and matrix operations, we reformulate the max-min signal-to-interference-plus-noise ratio (SINR) problem using uplink-downlink duality and introduce a closed-form expression for the virtual uplink power allocation. The proposed method eliminates the need for iterative updates in the uplink power optimization, while the downlink power is optimized using the conventional eigenvalue decomposition (EVD) method. Simulation results confirm that the proposed scheme achieves near-optimal performance with reduced computational complexity, particularly in large-scale systems.