A complexity measure for assembly supply chains has been proposed based on Shannon’s information entropy. This paper extends the definition of such a measure by incorporating the detailed information of the supply chain structure, the number of variants offered by each node in the supply chain, and the mix ratios of the variants at each node. The complexity measure is then applied to finding the optimal assembly supply chain configuration given the number of variants offered at the final assembler and the mix ratios of these variants. The optimal assembly supply chain configuration is theoretically studied in two special scenarios: (1) there is only one dominant variant among all the variants offered by the final assembler, and (2) demand shares are equal across all variants at the final assembler. It is shown that in the first scenario where one variant dominates the demand, the optimal assembly supply chain should be nonmodular; but in the scenario of equal demand shares, a modular supply chain is better than nonmodular one when the product variety is high. Finally a methodology is developed to find the optimal supply chain with/without assembly sequence constraints for general demands.