As good as it gets: Optimal fab design and deployment

This paper presents and applies a methodology for determining the optimal scale, type, and timing of IC manufacturing capacity additions when demand is uncertain. Using the methodology, six different 200-mm fab designs are evaluated, including standard 10K- and 20K-wafer start per month (ws/mo) facilities, a 5K-ws/mo "minifab," and "sequentially deployable" or "expandable" 10K- and 20K-ws/mo facilities. The results suggest that in the presence of uncertainty, manufacturers should follow a conservative deployment policy (better late than early to avoid periods of costly underutilization). In addition, when optimally deployed, larger, more efficient facilities are economically more viable than smaller fabs, since they are capable of generating substantially larger profits with comparable risks of loss. Sequentially deployable large fabs are even more attractive since they offer the economies of scale of larger fabs but require a smaller initial capital outlay.