In this paper, robust transceiver designs for point-to-point MIMO-OFDM systems are investigated. A general performance metric of weighted mean-square-error (MSE) minimization is adopted as the optimization objective function. Exploiting the signal structure of point-to-point MIMO-OFDM systems, a simple and readily comprehensive derivation procedure for the optimal solution is given. It is revealed that for the optimal robust transceiver design the solution is an interesting variant of the famous water-filling solution, which is named as cluster water-filling. This solution is significantly different from that of the robust transceiver design over flat fading channels and its counterpart with perfect channel state information (CSI) as the water-filling bottoms are also variables to be optimized among different subcarriers. Finally, the simulation result demonstrates the performance advantages of the proposed robust design. © 1997-2012 IEEE.