PA, you make some very good points. I'd like to believe that crony capitalism and ham-fisted strategic "management" of foreign economies has ended. Unfortunately, just as the blueprint for their economic rise was Japan, their ability to deal with structural inefficiencies in a timely manner, IMHO will prove to be built upon the "Japanese Bank Reform" model.
As a result we have seen a large drop in capex by the Koreans over the last year and only a very slight recovery this year - only 20-30% of levels that we saw only two-three years ago. Anyone else who expects a sharp rebound are deluding themselves and unaware of the underlying change that we are going though in Asia.
Despite that, the Koreans are expected to double their capex this year. From SBN:
Korean chip makers are expected to double their purchases of production equipment this year to $2.6 billion from $1.35 billion in 1998, Kim said. He said the bulk of the new gear would be to upgrade existing fabs, as chip makers aren't expected to launch any new plants this year.
semibiznews.com
Their capex is aimed at process technology, not expansion; however, one of the byproducts is "stealth expansion". Just ask Micron.
Now there was a certain analyst who postulated roughly a month ago that major process improvement initiatives could actually exacerbate the coming "shortage".
(02/04/99, 10:53 a.m. EDT)
SAN FRANCISCO — With unit shipments now increasing faster than new DRAM capacity, a memory shortage could occur by next year, warned a semiconductor analyst from NationsBanc Montgomery Securities during the company's technology conference here. The shortfall of supply and demand will be exacerbated by less-effective die shrinks and a dwindling pool of DRAM vendors worldwide, said analyst Jonathan Joseph.
Speaking at the NationsBanc Montgomery Securities Technology Week Conference, Joseph noted that several memory vendors have either cut their production levels or exited the market altogether in the past year. While this has cut down the worldwide DRAM capacity, overall bit demand is continuing to swell, and the overcapacity situation that plagued the sector throughout 1998 is likely to be offset within the next 12 months.
Another factor is the current generation of die shrinks. While previous shrinks have led to a significant gain in die-per-wafer, Joseph noted that some Montgomery studies have indicated that the move from 0.25-micron technology to 0.21-micron manufacturing is currently yielding only 40 percent more memory bits per wafer. This is not enough to offset growing bit demand which is running at more than 80 percent annually, and increasing.
"It will be pretty interesting by the end of 1999," he said. "And we could enter a DRAM shortage in 2000." ...
semibiznews.com
I believe this would be the same Jon Joseph from Nations Banc Montgomery Securities who today cut MU from a Buy to a Hold.
have seen Motorola, TXN, Siemens, LG, Fujitsu, Hitachi, Nippon Steel Semiconductor, Oki Electric either announce, about to announce, imply or deny rumors that they will be withdrawing from production of commodity DRAM.
I don't know where outsourcing fits in, but in the case of Fujitsu, I'm pretty sure I remember seeing that they expected to outsource something like 40% of their DRAM "production" to either TSMC or UMC.
You're right that we're undergoing a significant shake-out, but once it's over with, we'll still have 5 large players in a cut-throat commodity market where unilateral pricing power is non-existent. Conceivably, collusion have a greater chance of working, but MU has shown no interest in the past at collusion.
The point that Sanford Berstein's Zlotnikov made recently in CMP is worth repeating. If Dan Niles' "investment thesis" is that MU will continue to rise in part due to "increase in memory per PC", he should take the time to understand what made this increase possible. What made it possible was largely the cratering of DRAM prices. There was some elasticity at the margins. I have a hard time understanding why the rate of increase in memory per PC would continue to be the same during a period of price stability in the absence of the same drivers that made 1995 possible. Windows made increasing memory per PC a largely inelastic decision (i.e. it had to be done). Maybe on the high-end, that proves to still be the case for some systems, but given what's been going on in the low-end, I'd expect a deceleration in the rate of memory per PC growth.
Good trading,
Tom |
