Kindle Notes & Highlights
Over time the number of transistors that leading-edge producers can fabricate in a given area of silicon has doubled roughly every three years, and from 1995 to 2003 the pace accelerated to a doubling every two years.2
determined investment of the Japanese combined with a dip in demand in 1986 led to a situation of severe overcapacity.
It reported losses of nearly $2 billion for 1985 and 1986, while twenty-five thousand people lost their jobs. The Japanese companies lost twice as much money … but … in a contest of deep pockets theirs were deeper” (1988, p. 55).
Around the same time, nine of the eleven US-based producers of high-volume memory chips also exited the memory market (Young 1992).
Intel carved out a quasi-monopoly that grew to account for nearly a fifth of the semiconductor industry’s sales at its peak in the late 1990s. In 1999, Intel captured 82 percent of the microprocessor market, even counting non-x86 architecture chips.5 This is well above the level of market share at which the US Department of Justice considers the industry to be “concentrated.”
Micron has periodically tried to diversify, first into system-level products like PCs in the 1990s, then into specialized chips like CMOS image sensors for cameras in the 2000s, but memory chips remain its primary line of business.
Regional shares of semiconductor sales, 1980 to 2006. Source: Authors’ calculation using top forty firm data from market research sources. The top forty accounted for 80 to 90 percent of total semiconductor sales during the period.
The end of 2007 saw the beginning of such a period. The relatively small Taiwanese producers are most at risk because of their reliance on foreign technology partners and a heavy debt burden.10 The Taiwanese memory firms appear unlikely to emerge in their present form from the deepening recession, with some kind of restructuring expected as part of a new government partnership with Elpida.11
Already burdened by high debt-to-equity ratios, Japanese firms reduced their capital equipment spending in 1992. Meanwhile Korean firms raised theirs.
Capital spending by semiconductor companies, share of world total, selected years Region 1980 1985 1990 1997 United States 60.4% 35.5% 29.8% 33.0% Japan 28.9% 46.7% 50.0% 25.0% Europe 10.7% 8.3% 9.9% 9.0% Rest of world 0.0% 9.4% 10.3% 33.0% Total 100.0% 100.0% 100.0% 100.0% Source: Macher et al. (1999, fig. 4); data supplied by author. Note: Columns may not total precisely, due to rounding.
The costly move that began in 2000 from 8-inch (200-mm) to 12-inch (300-mm) wafers (again, a 1.5 × increase in wafer diameter) provided a 2.5 × increase in wafer size. A further move to 18-inch (450-mm) wafers, which would again impose enormous development costs on industry leaders and equipment makers, entered an early stage of evaluation in 2008 and would provide a similar increase.3 As discussed in Crisis 1, semiconductor
One market researcher estimated that this would result in a 10 percent unit cost savings over megafabs because fixed costs, such as software for monitoring and automation, can be spread out over the larger output and because production bottlenecks from equipment failure are less likely to arise as a result of the increased redundancy of each type of equipment.4 However, even as the unit costs go down, the fixed costs of building fabs have skyrocketed and resulted in fabs that must operate constantly at extremely high volumes to be profitable.
but by 2007 the price tag for an advanced fab of minimum efficient scale had climbed to as much as $5 billion. A flash memory fab built in Japan by Toshiba and its joint-venture partner SanDisk reached a scale of more than 100,000 300-mm wafers per month at an estimated capital cost of $7 billion.5
