Sensor network nodes are often limited in battery capacity and processing power. Thus, it is imperative to develop solutions that are both energy and computationally efficient. In this work, we present a simple static multi-path routing approach that is optimal in the large system limit. In a network with energy replenishment, the largeness comes into play because the energy claimed by each packet is small compared to the battery capacity. Compared to the other routing algorithms in the literature, this static routing scheme exploits the knowledge on the patterns of traffic and energy replenishment, and does not need to collect instantaneous information on node energy. We also outline possible approaches for a distributed computation of the optimal policy, and propose heuristics to build the set of pre-computed paths. The simulations verify that the static scheme outperforms leading dynamic routing algorithms in the literature, and is close to optimal when the energy claimed by each packet is relatively small compared to the battery capacity.