Drawing on an empirical study on automakers? management of supplier involvement in product development in Japan, this paper shows that when the design of a component is outsourced to a supplier, how much and what automakers know about the component matters for them to gain a better outcome. While the actual tasks of designing and manufacturing components could be outsourced, automakers should retain the relevant knowledge to obtain better component design quality. The paper argues that knowledge partitioning should be distinguished from task partitioning, and provides some implications for the knowledge-based theory of the firm. The results indicate that effective pattern of knowledge partitioning differs by the nature of component development project in terms of technological newness. For regular projects, it is more important for the automaker to have a higher level of architectural knowledge (how to coordinate various components for a vehicle) than of component-specific knowledge, which is supposed to be provided by the supplier. However, when the project involves new technology for the supplier, it is important for the automaker to have a higher level of component-specific knowledge to solve unexplored engineering problems together with the supplier. In innovative projects, effective knowledge partitioning seems to demand some overlap between an automaker and a supplier, rather than efficient and clear-cut boundaries, which are optimal for regular projects. Such ?fluid? nature of knowledge boundaries contingent on the project types poses a challenge for firms seeking both technological leadership as well as efficiency in established products. Developing and maintaining knowledge about an outsourced component is by no means easy. When the actual design tasks are outsourced, automakers miss substantial opportunities to gain relevant knowledge through learning by doing. Also, obtained knowledge may be diffused among competitors through shared suppliers. Another problem for automakers is that component-specific knowledge is important for only limited cases (innovative projects). Even worse, component-specific knowledge has a trade-off relationship with architectural knowledge. Such an inherent dilemma of managing knowledge, however, may provide some automakers with the opportunity to achieve sustainable competitive advantage. Additional analysis shows that one automaker managed both types of knowledge better than others in a manner that deals effectively with the dilemma. Its organizational mechanisms include career development policies, extensive documentation of technological information, internal training programs, and incentive schemes. The difficulty in implementing those mechanisms in a consistent and complementary manner seems to explain why there was a significant variance among automakers in knowledge level, even when the actual tasks were carried out by a shared supplier. (414 words)Massachusetts Institute of Technology?s International Motor Vehicle Program (IMVP),Grant-in-Aid for Scientific Research of Japan Society for the Promotion of Science