Multiple-input multiple-output (MIMO) techniques can be used for reducing the energy consumption of wireless communications. Although some research has been reported on this topic, the rules by which the MIMO physical layer parameters should be chosen in order to achieve energy efficiency have not yet been formally established. In this paper, we analyze the case of MIMO singular value decomposition (SVD) technique. We present a model for the mean energy consumption of a MIMO SVD system per data bit transferred without error. We find that, for a given number of eigenchannels used with equal power allocation, exists a single optimal radiation power level at which the mean energy consumption is minimized. We also find that beamforming (only the best eigenchannel is used) is optimal in the energy consumption sense for long transmission distances, while the optimal number of eigenchannels to be used grows as transmission distance shortens. Using all the eigenchannels is optimal only for very short transmission distances.