Gottfried (and threaders),
"Total world wide semiconductor shipments in 1999 were about, what, $140 billion?"
Close enough, although I've seen up to $165 billion.
That's expected to exceed $300 billion in 2004, according to estimates from SEMI, Dataquest and IDC. So $41 billion for large, fast memory in the US isn't unlikely.
I think that price of $500 a chip may be one of those "test shots across the bow." That'd be $4,000 for 512Mbytes, as a DIMM or some such. Rather pricey! It had better be faster than a speeding bullet.
If you look at how prices reduce during the lifetime of a memory type (due to Moore's law and technical advances), you soon find an interesting fact: A given generation of chip (such as the 16Mbit) drops to about $3.10-$3.20 per unit after 3-5 years, and then exits the market. Indeed, the average is $3.14, or pi. Something magical about this?
It remains to be seen if 64Mbit chips will reach this low number, they've been as low as $4.85 on the spot market, but not $3 in my recollection. Prices have firmed, and are about $7 a chip right now. This extends the product life cycle, of course; manufacturers (such as Samsung) can make money at about $5 a chip on that product right now, and would learn how to make it profitable at $4 I'm sure. (Track record speaking: They've done it for every other generation so far!)
But maybe Samsung and others are looking to break the "pi" barrier? The 128Mbit chips are in production several places, and 256Mbit technology is tested, although manufacturing is a problem. Maybe the memory makers plan to abandon the "venerable 64Mbit chip" early?
That's good news for semi equipments, as upgrades and outright replacements will have to increase. Some 64Mbit chips are being made on equipment that is fully extended technically to meet that challenge. Going smaller, adding Cu, etc., will obsolete a bunch of equipment used in current memory manufacture.
But you already knew, that, right? <grin>
Mitch |
