Featured with high bandwidth, high reliability, and native IP compatibility, WiFi has been recommended for a wide range of Internet-of-Things (IoT) applications. However, WiFi is inherently energy-hungry and it may impose high energy consumption on not only IoT devices but also gateways. To reduce gateway’s WiFi energy consumption, many energy-efficient solutions for WiFi tethering services can be applied. However, these solutions mainly target the energy optimization of downlink data traffic in WLANs, and they are not suitable for the uplink data traffic of delivering massive IoT data from device to gateway (D2G), which is more common in IoT. When a gateway is powered by batteries, the high energy consumption caused by D2G communications may deplete the gateway quickly and renders the whole system dysfunctional as a result. Toward achieving energy-efficient D2G communications, we propose an innovative Green IoT Gateway (GIG) scheme, which aims at minimizing gateway energy consumption while ensuring the specific delay requirements of devices via cross-interface collaboration. Through utilizing the coexisting low-power ZigBee radios, GIG dynamically schedules the wakeup behaviors of high-power WiFi radios for energy-efficient and delay-bounded D2G communications. GIG has been implemented and evaluated in a prototype system, and the experiment results show that, under the moderate uplink data traffic and delay requirements, the energy consumption of GIG is 38.5% and 12.7% lower than those of a state-of-the-art WiFi tethering scheme and a simple version of the GIG scheme, respectively. Moreover, a great reduction of energy consumption at the device side can also be observed.