Limits to Modularity -- Reflections on Recent Developments in Chip Design

Debates on how business organization has moved beyond Chandler's vertically integrated multi-divisional firm have greatly benefited from the concept of “modularity”. There is however an important tension in the literature between the proponents of modularity and a small, but growing revisionist literature that contends that the enthusiasm for modularity has gone too far. Rather than exploring challenges and difficulties that management is facing in implementing modularity, there is a tendency in the “modularity” literature to generalize empirical observations that are context-specific and to confound them with prescription as well as prediction.This paper sides with the revisionist literature in cautioning against such strong claims of pervasive modularity. The objective is not to propose an alternative theory. More modestly, I am aiming to move the debate away from polemics to a scholarly discourse that explores why and when modularity may have limits, and what management can do to overcome these limits. I examine new evidence from a cutting-edge industry, semiconductors, that is often cited by modularity proponents as an indicator of broader industry trends. The paper shows that, even in this industry, there are powerful counter-forces causing organizational structures to become more integrated, not more arms' length. Evidence from chip design is used to analyze how competitive dynamics and cognitive complexity create modularity limits, and to examine management responses. I demonstrate that inter-firm collaboration requires more (not less) coordination through corporate management, if codification does not reduce complexity -- which it fails to do when technologies keep changing fast and unpredictably