As the advances in network technologies are leading to huge bandwidth increases, users with high bandwidth connectivity have started to download all kinds of very large objects routinely. Our contributions in this paper are the following twofold. First, we find that, during peak times, even a small number of concurrent requests for large objects often cause significant degradation in web caching performance, on a high bandwidth network. Second, as a solution to the previously defined web cache server bottleneck problem, this paper proposes a hybrid web caching architecture which exploits the resource scalability of the peer-to-peer model and high processing power of dedicated server model. We evaluate the proposed model through trace-driven simulation and show that it performs the task of caching and delivery of large objects in an efficient and cost-effective manner, without imposing overheads on participating peers.