We address the optimal allocation of a given capacity in a multi-product inventory-production system with base stock. In the system, the base stock level is equal to the sum of the finished goods and work-in-process inventory. Each exogenous demand releases production of a new unit and increases the level of work-in-process. Unsatisfied demand is backordered. Using an open single-server queueing network, we model the behavior of the work-in-process in the manufacturing system in order to determine the cost-optimal base stock level. We focus on the utilizations in the network and compare balanced and unbalanced systems regarding their base stock level. In particular, we analyze the effect of capacity reallocation and show that a manufacturing system consisting of production stations with constant utilization rates, results in the lowest base stock level. With the combined application of queueing network theory and majorization theory, we are able to provide analytical explanations of the network's behavior.