Research Article
Learning Effects and Cost Leadership Strategy in the Semiconductor Industry
1 Yonsei University, 2 Deloitte Anjin LLC
Published: January 2010 · Vol. 39, No. 2 · pp. 255-278
Full Text
Abstract
The primary object of this study is to examine the impact of learning effect on cost leadership in semiconductor industry. In semiconductor manufacturing, wafer fabrication (FAB) is the most complicated and important process, and is composed of several hundred process steps and involves several hundred machines. Therefore, the semiconductor fabrication usually aims to improve the wafer yield rate, which have been recognized as important factors to build core competences for semiconductor manufacture. The key contributor to a learning curve driven by manufacturing efficiencies is also the improved wafer yield rate. Rapid yield-learning rate determines profitability because high yield leads to low unit costs and high profit margins. This study, therefore, focuses on the output yield rate in the FAB process of Dynamic Random Access Memory (DRAM) to examine the association between learning effect and cost leadership. We differentiate our study by using output yield rate as a proxy for cost and analyzing an individual firm’s strategic decision by quarterly firm-level data. Data on FAB and DRAM was obtained from ‘DRAM Supply and Demand Quarterly Statistics provided by Gartner Dataquest from 1996 to 2006, which included the 44 quarters. The sample firm included 22firms operated worldwide. Empirical results show that first mover firms with a greater experience of new generation DRAM can attain cost leadership when it becomes a major DRAM. The learning effect observed is that the yield rate of current major DRAM increases with accumulated output and product time. We find that price premium exists in a new generation DRAM. However, we do not find the relation between the relative entry order and cost leadership, which indicate that first mover advantage do not arise simply from buyer switching cost. Our additional test results provide evidence of spillover learning effects. However, these spillovers are neither complete nor symmetric.
