Spatial modulation (SM) is a promising single frequency-radio (RF) large-scale multiple-input multiple-output (MIMO) technique which can strike an attractive tradeoff between spectral efficiency (SE) and energy efficiency (EE) of a wireless network. Recently, a novel technique termed as orthogonal frequency division multiplexing with index modulation (OFDM-IM) emerges, where the SM principle is applied to OFDM subcarriers. It is revealed that OFDM-IM has the potential to outperform classical OFDM in terms of uncoded bit error rate (BER). In this paper, the EE of OFDM-IM with finite-alphabet input is analyzed from an information-theoretic perspective. Based on the analytical expression, the optimal strategy in maximizing the EE of OFDM-IM is proposed, which involves the optimal subcarrier grouping approach as well as the optimal selection of the number of active subcarriers. Monte-Carlo simulations validate the analysis.