Automakers keep introducing new functions to vehicles to entice customers, thus increasing the size/number of in-vehicle buses. As a result, adopting multi-buses with a central gateway is becoming a norm in current and future vehicles. Since adding buses increases production cost, knowing whether or not to add a bus at design time to meet the given requirements is essential to design a cost-optimized in-vehicle network. However, due to the lack of an optimal priority-assignment algorithm for multi-bus systems, it is difficult to determine whether additional buses are needed. To address this difficulty, we propose an optimal priority-assignment algorithm, called OPMB, for multiple CAN/CAN-FD buses with a central gateway. OPMB builds on backtracking, and is thus of exponential time complexity. To reduce the execution time effectively for industry-size problems, we identify several theory-proven search-space reduction conditions. Our in-depth simulation has demonstrated that OPMB outperforms the state-of-art priority-assignment algorithms for multi-bus systems, and is suitable for high-speed systems which represent future automotive systems. Also, OPMB is shown to be feasible for most realistic automotive message sets.