Offloading cellular traffic to device-to-device (D2D) communications has been proposed to improve the network capacity and to alleviate the traffic burden on base stations (BSs). However, as mobile devices are powered by limited battery energy, there is no obligation for D2D transmitters (D2D-Txs) to offload cellular traffic through D2D content sharing. In this paper, we model and analyze the wireless-powered D2D-assisted offloading (WPDO) in cellular networks, where the D2D-Txs can harvest radio frequency (RF) energy from nearby BSs and utilize the harvested energy to share popular contents with nearby user equipments (UEs). Stochastic geometry is used to characterize the intrinsic relationship between the wireless power transfer (WPT) and the information transmission. Based on the proposed model, we derive the average transmit power at D2D-Tx, the expected minimum transmit power at BS, the D2D outage probability, and the cellular downlink outage probability. We also investigate the energy efficiency of the WPDO network from a systemlevel perspective. Simulation and numerical results show that the energy efficiency of the WPDO network can be maximized by optimizing the fraction of time allocated for WPT and it can be further improved by using massive antenna arrays at each BS and by sharing more popular contents between devices. Index Terms-D2D communications, energy efficiency, traffic offloading, cellular networks, wireless power transfer. I. INTRODUCTION W ITH the upsurge of mobile data traffic and the explosively increase of mobile devices, cellular networks are facing technical challenges in supporting enormous data flows, high data rate, and large system capacity. In high user density areas, the base stations (BSs) are suffering heavy load burdens. To address the above issues, device-to-device (D2D) communications have been proposed to improve the network