Process Innovation and Learning by Doing in Semiconductor Manufacturing

TLDR

Knowledge gained through learning by doing is often environment‑specific, leading to loss when a new process is transferred to manufacturing. The study examines how process innovation and learning by doing interact to drive yield improvements and cost reductions in semiconductor manufacturing. The learning curve results from deliberate yield‑improving activities rather than mere volume; dedicated development facilities, proximity, and equipment duplication improve technology adoption, but new processes divert engineering resources and disrupt ongoing learning.

Abstract

This paper analyzes the relationship between process innovation and learning by doing in the semiconductor industry where improvements in manufacturing yield are a catalyst for dynamic cost reductions. In contrast to most previous studies of learning by doing, the learning curve is shown here to be the product of deliberate activities intended to improve yields and reduce costs, rather than the incidental byproduct of production volume. Since some of the knowledge acquired through learning by doing during new process development is specific to the production environment where the process is developed, some knowledge is effectively lost when a new process is transferred to manufacturing. We find that dedicated process development facilities, geographic proximity between development and manufacturing facilities, and the duplication of equipment between development and manufacturing facilities are all significant in improving performance in introducing new technologies. Once in manufacturing, new processes are shown to disrupt the ongoing learning activities of existing processes by drawing away scarce engineering resources to “debug” the new processes.