In the context of vehicular delay-tolerant networks (VDTNs), routing is critical to the overall performance and functionality of the network. With no static connection, and with a highly temporal network behavior, conventional routing techniques seldom work in the context of VDTNs. Techniques that leverage context information, such as position and number of nearby vehicles, are becoming popular. However, the opportunity and time-window for the maximum exploitation of context information is often limited for several reasons, such as nature of information (private or public), and duration within which the information is valid for. Therefore, using available context information quickly is crucial to ensure optimal performance. Routing techniques that require heavy computations or unlimited replications of messages are rarely attractive. More specifically, as the routing takes place when a message carrying node (vehicle) encounters another vehicle, the decisive action has to be executed within a very short period of time. In this paper, we propose a novel multi-copy, speed-adaptive routing algorithm that purely relies on the relative speed of encountering nodes for the routing operation, without demanding expensive computations. We study the performance of the proposed algorithm under various motion configurations using a number of different performance metrics, such as probability of delivery, network overhead, message latency, number of created and dropped messages, and number of hops that messages pass through before being delivered. Our results show that the proposed speed-adaptive algorithm outperforms all conventional routing techniques across different performance metrics, and offers superior probability of delivery, negligible network overhead and message latency.