An end to the DRAM glut in '98? Who knows?
By Jack Robertson
More and more market analysts are forecasting that DRAM supply will
come into balance with demand in the second half of this year. Don't
bet on it. Sure, in the last six months more than a dozen fabs have been canceled or delayed. Sure, the South Koreans and the Japanese have cut their 1998 capital spending budgets by 20% to 35%, with much of that reduction slowing down planned DRAM expansion. Sure, Hyundai Electronics Industries Co. Ltd. and LG Semicon Co. Ltd., which together account for 12% to 15% of the world's DRAM supply, are facing potential U.S. dumping duties on DRAMs that could force the two companies to rethink massive exports to the United States.
All of these developments may be academic. What clouds the DRAM crystal ball is how fast the suppliers accelerate their die shrinks. Forget the cutback in brick-and-mortar fabs. Each generation of DRAM shrinks can almost double the number of chips produced on a wafer. In many cases, such increases offset the deferred output of a shelved fab.
We don't have a good handle on the timing of various companies'
shrinks. In most cases, this is confidential market information that
is not nearly as discernable as a new fab structure. Micron
Technology Inc. blew the DRAM market open last year with its
unexpected doubling of production from a 0.30-micron feature size
shrink. With almost all rivals now emulating Micron on shrinks, the
growth of DRAMs is hard to predict.
In fact, desperate DRAM makers are speeding up shrinks - even
forgoing generations. Micron is hoping to skip quarter-micron
processing and move from 0.30-micron to 0.22-micron feature size
wafers. The world's largest DRAM producer, Samsung Electronics Co.
Ltd., has moved from quarter-micron capability and is ramping up
0.23-micron processing for 64-Mbit.
A few DRAM players could finally be scratched from the Shrink
Derby's field. Such a shakeout would have far more market impact
than all the cutbacks in fab construction. That is assuming, of
course, that the shrink effort doesn't pick up the slack for the
departed DRAM makers.
Then, if you're trying to gauge whether the massive DRAM oversupply
will end this year, you have to project whether demand will keep up
its 90% to 100% bit-rate growth of the last few years. There are
signs that computer customers' appetite for memory may be
slackening. But this could be only a transitory dampening of the
always-volatile PC market. And though 77% of DRAM chips end up in
computers, there is surging memory demand from a wide range of other
sectors.
For now, it's anyone's guess as to when the DRAM glut will end.
Copyright (c) 1998 CMP Media Inc.
Aloha |
