A voice over Internet protocol processing method and a related network device are disclosed. The method includes: detecting, by a calling access network element, an off-hook event; encapsulating, by the calling access network element, the off-hook event into first signaling; sending, by the calling access network element, the first signaling to a cloud server; receiving, by the calling access network element, second signaling sent by the cloud server, where the second signaling is used to establish a first transmission path between the calling access network element and the cloud server, and the first transmission path is used to transmit data between the calling access network element and the cloud server; and establishing, by the calling access network element, the first transmission path based on the second signaling. Overall efficiency of the voice over Internet protocol system can be improved, and the construction costs can be lowered.