Hi, I've been contemplating your questions this evening. A big order, and a very relevant question at this time. First, here is a fact filled post from the Blood in the Streets thread:
eet.com
Grim indeed!
This whole subject of 300mm capacity begs a bunch of speculation. Here's some assumptions that I'm making just to get it started.
300mm means each wafer has 2.25 times more surface area.
A new 8" or 200mm wafer fab now costs more than $1.0 Billion to build.
300mm will have a cost premium associated with it throughout the line.
Assume a typical 200mm DRAM fab does 30,000 wafer starts a month.
New fabs are built to handle larger volumes than older ones.
DRAM mfging capacity has over-produced from months and years.
Supply/demand law has driven price of DRAMs into the floor.
Cost of a 64 Megabyte DRAM chip is about $2.00/Meg retail.
A state-of-the-art PC has more than 64 Megs of memory.
A typical PC has 64 Meg or less, assume 64 Meg is average.
Chip lifetimes are measured in "bathtub" curves that extend 40+ years.
The PC market is reaching saturation in industrialized nations.
The typical PC runs Win95, with its DOS limit of 1024K memory map.
Japan still plans strategically, a few quarters out is not strategic.
Japan has been in a growth recession for 7 or 8 years.
China is the major market of the future, but doesn't have a high
percapita income yet.
Now for the wild speculation. The specter of 300mm processing has been lingering in the future for years. There are economic barriers that must be hurdled to bring it into the present. It takes a major force in the market to make the leap of faith that conversion to 300mm requires. I think NEC can pull this off financially, as you indicate their plans have been laid but they need a lot of help.
Guessing that they would like to build a 50,000 wafer-starts-per-month fab, that means their fab would have 3.75 times the productive capacity of a current 200mm fab starting 30,000 wafers, not speculating on any further multiplier factored in for line-width shrinkage. That means the world needs 1/4 as many DRAM fabs as it now has built with this capacity to get into the same glutted situation, and that's optimistic assuming no line-width multiplier. If a line-width reduction from .35 micron to .18 micron is factored in this number might reach 1/8 as many DRAM fabs in the future without some sort of demand driven stimulus increase.
Clearly there is too much fab capacity, today and in the foreseeable future for DRAM. This is causing 300mm to be delayed. To get rid of this extra capacity and cripple the potential competitors as a barrier to entry the world needs a major and permanent cutback in capacity along with a long term projection of demand and growth. A deep and long worldwide economic downturn might accomplish this. And thats what we are right in the middle of, right now. So the opportunity may exist for the select few to accomplish this fab line conversion.
I think Japan recognizes that China is going to be a tremendous market, and it will be a market that demands cost efficiency because it is not a rich market place now and won't be for many years to come. Therefore high volume/low priced/highly efficient factories that can tightly control their cost structure are the ones that will prosper. Given the potential size of that market, Japan recognizes that it will remain a reliable buyer for many years, and its this aspect that also suggests the possibility of migrating to 300mm.
Somewhere downstream I foresee a replacement operating system that will require the typical PC to upgrade memory to a higher level, perhaps the 1 Gigabyte level. Right now, PCs are not memory -constrained at 64 Meg. So there is no compelling economic rationale currently pushing existing PC users to upgrade maybe someday but certainly not today. And with the current situation with the Federal Government pursuing Microsoft, it seems to me that Microsoft is being distracted from pursuing the reinvention of the operating system as a goal. I may be wrong about this, but until they do come out with a compelling new operating system, and I think only they have the clout to unseat their own OS products,growth in the installed memory base will continue to flatten.
Fewer fabs handling a concentration of DRAM production is not the scenario that drives all the semi equip mfgers shareholder value higher. On the contrary, one must be a better stock picker to play this aspect.
Now RISC, ASICs, PLDs and the other types of newer technology are whole new risk arenas and beyond the scope of what I was trying to focus on here. |
