In this letter, an equivalent uplink problem formulation is developed in the multi-cell environment such that the downlink beamforming throughput is maximized under the assumption that base station (BS) can only use the estimated receive signal to interference plus noise power ratio (SINR) of mobile station (MS) in computing beam vectors because of channel estimation error and delay feedback. Numerical results show that the proposed downlink beamforming throughput maximization offers noticeable improvement of 5% outage and average throughput over conventional minimum mean squared error (MMSE) beamforming in the presence of channel uncertainties whereas the performance degradation in the environment of the nominal channel uncertainty is significant compared to the case of no uncertainty.