Establishing reliable routing among highly mobile vehicles is a challenging problem in vehicular networks. Towards this issue, we present VehiCloud, a novel cloud computing architecture that leverages emerging cloud computing technologies to deal with unreliable inter-vehicle communications and extend the restricted computational capabilities of mobile devices. A way-point information framework (WIF) is devised within the VehiCloud architecture, aiming to provide routing service for vehicular network, where each vehicle serves as a mobile service node and predicts its future locations by generating way point messages, which describe the trajectory of the vehicle's movement. A decision module in VehiCloud collects vehicles' way points and makes routing decisions for inter-vehicle communication. Selected paths of the routing are globally optimized in terms of message delivery ratio by respecting the constraints of end-to-end delay and communication cost. Our implementation of VehiCloud and real-road experiments demonstrate that it is practical and efficient to address fundamental routing problems for vehicular networks.