WARNING - really long post!
Re: "won't help the dram mkt much in the next year or so though"
I disagree with that assumption, as outlined in the following section from my Industry Outlook writeup this past January (please forgive the length; if I were pithy, I'd be dangerous [speaking of pithy, that reminds me of the old joke about Thor, the God of Thunder - but that's for another time and place]):
"The secular case for semiconductors has been based on two trends. First and foremost is the industry's ability to shrink line widths in a fairly consistent fashion, which lowers cost per function. This feature has been evidenced since the invention of the planar integrated circuit, but repeatedly has been called into question on a forward-looking basis due to those pesky laws of physics. Chip fabricators haven't been able to overturn Isaac Newton, but they certainly have been able to fool him, and this should continue well beyond the next decade. To be sure, there have been and will continue to be tremendous costs associated with driving the technology, but as long as the increased production exceeds the costs, the investment will be made. The second trend has been the industry's ability to engender its own demand growth by opening new markets as a result of the first trend (increased functionality at a lower cost). This self-induced market expansion was what led to the misplaced sense of security in 1995.
We believe that both trends are still firmly in place and pose no incremental positive or negative to the semiconductor case. What should be emphasized in 1999 is the “fairly” in “fairly consistent fashion.” Over a sufficiently long period of time, the trends should appear to be quite linear. However, they do get distorted during periods of cyclical extremes, which is relevant to our 1999 outlook. During boom times in the industry, capital spending is focused on capacity additions in the form of increased wafer production. This tends to slow down the reduction in line widths for a period, which also creates a reduced rate of growth in demand. But as the supply increases and the demand growth slows down, the industry's pursuit of finer geometries adds to the price pressures already generated by the changed supply/demand balance and leads to new functionalities at lower price points. We have been and should continue to be in precisely that kind of an environment.
An interesting example of this phenomenon can be found in the memory business. Like many observers, we believed that the semiconductor industry would be able to take business at the margin from the vendors of rotating memory, since the disk drive business appeared to be driven by technology that was incapable of matching the decline in the price per bit of semiconductor memory. In fact, the shift to magneto-resistive (MR) and giant magneto-resistive (GMR) heads has allowed the disk drive industry to not only meet, but actually exceed the decline in semiconductor pricing. As a result, the price premium for semiconductor memory has remained sufficiently high to prevent the move from disk drives to RAM in all but the most specialized cases.
However, the decline in memory pricing has been in flash as well as DRAM and that has combined with lowered processor cost to enable the use of flash-based music players. With the advent of MP3 (MPEG 1 Layer 3), music can be coded in relatively small files without perceptible loss of fidelity. Originally used for Internet music, the format has spawned the Rio from Diamond Multimedia (DIMD, $6.75, Not Rated), a player the size of a cassette tape that can hold up to an hour of music in MP3 format with its standard 32MB of flash memory and uses a single AA battery. There have been some intellectual property issues (a strange term, indeed, given the nature of most of the music involved), but these appear to have been resolved. Since the product's introduction on November 20, 1998, it has been well-received and has enjoyed a tremendous amount of publicity. This doesn't replace the rotating memory of a compact disc, but it is an adjunct that creates a new pocket of demand for semiconductors.
The notion of new products created by increased functionality at lower price points has applicability to the PC market. The sub-$1000 PC is not a new product in the strict sense of the definition, but it does lead to a new class of buyer and generates additional consumers. This is an interesting twist to our discussion, because while it is an outgrowth of secular trends, the low-priced PC has led to an increased exposure to seasonality. We believe that much of the resurgence in semiconductor demand over the past several months can be attributed to three factors, with the seasonality of the consumer segment the most important. The other two factors would be the radical revision in the yen/dollar ratio, which has served to increase the dollar price of many segments of the semiconductor market, and the dearth of product in the channel that existed last summer and has served as a base for healthy sequential gains, particularly for products tied to the PC market.
There is one other aspect of the PC market that should be mentioned: the alteration of the channel. The shift to the build-to-order (BTO) model from the build-to-plan mode formerly used has clearly reduced inventories in the distribution chain. To a certain extent, that should mitigate some of the inventory swings that have plagued the chip industry over most of its existence. Unfortunately, that improvement is obtained at the cost of what would appear to be increased susceptibility to changes in purchase patterns. The use of 800 numbers and web sites in place of store fronts and wholesalers means that perturbations in buying will have little or no buffering before showing up in integrated circuit sales. Unfortunately, we may have gotten through the 1998 inventory drawdown at the OEMs just in time to develop an exposure in 1999 to purchase cutbacks by a user base that may have pulled some of 1999's buying into 1998.
We also believe that with roughly half of the semiconductors going into PCs and roughly half of the world's PCs manufactured in Taiwan, the semiconductor industry's exposure to the effects of Chinese New Year late in the first quarter may be quite pronounced. This was not a factor for the industry five years ago, but it has become an intrinsic aspect of the business.
Putting all of this together leads us to believe that the end of the first quarter may very well show some weakness that could negatively surprise those investors who have assumed that it is all onward and upward from last summer's bottom. The second quarter has an easy annual comparison (last May and June were atrocious) and should also have a decent sequential comparison. The third quarter will have to contend with more traditional sequential comparisons than in 1998, but the fourth quarter should be strong and set the stage for a very good 2000. However, it is imperative that the current rebound in equipment orders and shipments turn out to be an unsustainable blip for our post-1999 optimism to pan out. If equipment purchases show a dramatic gain over the next several quarters, it will call into question the pricing tightness that undergirds our industry viewpoint." ---from "The Winter of our Discombobulation"/January, 1999
To summarize, there are two things that should not be dismissed: seasonality and demand creation. I am convinced that the same seasonality that lulled bulls into thinking the worst of the memory pricing problem was behind us last fall is now misleading memory bears. And with psychology being so important in the pricing process, a relatively small shift in capacity from DRAM to flash to meet the needs of an emerging market can have a dramatic effect on DRAM pricing.
This leads to the $64 question: how much of the current pricing strength in the flash market that is attributed to demand from cell phone manufacturers is at risk if there is a seasonal pattern/slowdown in cell phones (I mention that because Sony's earnings release this week mentioned losses in cell phones; while I don't know what its analog/digital split is, if Sony is losing money at current cell phone pricing and other vendors are "enjoying" a similar dynamic, there could well be a cutback in production, either in response to the direct profit implications or because end-users have been artificially stimulated by below-cost pricing and a temporary surge in demand has been met)? The obvious danger is that any and all of the benefits of flash demand growth due to the MP3 players could be more than offset by sequential weakness in demand due to a shift in cell phones.
If, however, cell phone demand stays firm, then a buildup in flash demand from MP3 players could well tighten the flash market further and lead to a shift in production from DRAM to flash. While some of the largest and most important flash vendors do not make DRAMs as well (e.g., STM and Atmel), others have some exposure to other memory markets (e.g., Intel/Sharp and AMD/Fujitsu) and there are some marginal flash vendors that are DRAM guys (e.g., Samsung and MU). This sort of production shift could accelerate the attainment of DRAM supply and demand balance.
When all is said and done, a lot of what concerns investors is comprised of the shifts at the margin - Tad LaFountain |
