by Anna Filatova
03/26/2003 | 01:44 AM
In the end of February PC memory prices in the open market dropped already below $3 for a 256Mbit DDR chip. It means the price hit the psychologically important point, which is considered by many analysts to be the prime cost of these chips. That’s the outcome of the slump from the $5-6 where the chips were quite recently. As soon as memory makers started little by little making profits or hoping to make some, they again had to confront the depressing quarterly reports. How long is this going to last? Or we are up to another price rise bringing a respite for DRAM manufacturers and a trial for our purses? Let’s sort it out.
Big things come first. At the end of the last year the semiconductor industry pulled itself together and squeezed out a tiny sales growth: from $140 to $141 billion. This was largely due to increased demand on wireless equipment, including chips for mobile phones. Moreover, the only region to show growth was Asia Pacific (29% above 2001), while America suffered 13% sales drop, Japan and Europe lost 8% each.
And what will happen next? As for the current year, some quite respectable sources, like Semiconductor Industry Association or Merrill Lynch, forecast 12-20% growth, that is, up to $157.5-$169.3 billion. And once again the main locomotive power of the growth will be chips for wireless devices. These forecasts are more than optimistic; the analysts hope for a market rebirth, although there are still strong doubts about the growth scale. Moreover, the manufacturers themselves don’t seem to be agreeing with such a bright vision of their own future. A nice example: the CEO of TSMC expressed a warning at his last meeting with the company investors: “Many high-tech company CEOs are very conservative, even pessimistic, about 2003”.
We may recall the same analysts predicting semiconductor industry renewal a year ago. They all spoke of an imminent demand growth, but the results turned to be completely different. Actually, such optimistic forecasts may play a bad trick on those market participants who trusted these forecasts. It may turn out really bad, if they start increasing the production volumes and amount of goods in stock hoping for demand growth, and the growth will actually never happen in reality.
Just have a look around: do you see any evident prerequisites for improvement of the economical situation in America, Europe or Japan? While their politicians practice war/antiwar rhetoric and economists are more concerned with oil prices and nearly baffled economy, we are more likely to see companies cut short their expenses, including money spent on new PCs. We doubt anything is going to radically change here in 2003.
Well, as far as the DRAM section of the semiconductor industry is concerned, there will surely be some noticeable growth, dozens of percents, somewhere about 30-40%. It will be at the last year’s level when total memory sales volume grew by 36%, from $11.2 to $15.3 billion. The growth is even more (41%) if measured quantitatively, in chips. The number of chips produced is growing – their average price is lowering…
Why we give such a forecast for the DRAM market? Firstly, this market grows according to demand. Today the production facilities of this industry sector are used at 60-70% of their full capacity. Moreover, a number of manufacturers have their 300mm fabs currently under construction. So, the last-year 36% and currently predicted 30-40% growth this year is, roughly speaking, the growth of demand.
So, the only question is whether PC memory will enjoy the same demand as last year. As we have said, capital expenses are being reduced, companies are re-structuring themselves trying to reduce the costs and we just can’t hope for any big growth of spendings on IT. There is a curious opinion that those who bought computers in the end of 1999 to avoid the Millennium crash are now ready to upgrade their systems. But where are all those upgrades and is the upgrade so necessary? We can definitely say that there have not appeared any tasks (especially in the corporate market) putting much higher demands on the hardware part of an average PC.
By the way, the same is true for the purchases of a PCs today. It’s quite enough to have about 256MB RAM now for ordinary work. If the memory price dropped twice today, would it lead to users buying 512MB for the same price they pay for their today’s 256MB? No, what for? A mass user just doesn’t come across applications that require these 512MB.
So, of course there will be growth, but there won’t be anything exceptional. There are simply no incentives for that.
That’s why we have those 30-40%. It’s interesting that CPU manufacturers are also facing similar problems nowadays. If we use economics vocabulary to characterize the demand in this market sector, will sound as follows: the demand in this segment is inelastic. It means that the 50% price drop doesn’t result in doubling of the demand growth as there is simply no necessity to increase CPU efficiency and memory amount any further. So far, CPU manufacturers as well as DRAM makers couldn’t resolve this situation. Those super-overclocked “PC4000” modules or 2GB DIMMs, which have been available for servers only, are the signs of this dead point. The general tendencies for higher memory speed and volume are carried to an absurdity of a niche product here.
To sum it all up, we will have certain memory sales growth this year, but no big one, just like last year.
Let’s now turn to the second component of the end price: the supply. This is quite a different story. To begin with, the anticipated demand growth would push all the manufacturers to load their production facilities to the full. And that could be the only thing to be done this year. And in reality the things are much sadder.
Firstly, even if we disregard for a minute all new technologies and fabs being built now, the production volumes will increase due to the facilities, which haven’t ever been involved yet. Here we are definitely talking about the Chinese contract makers. The best example will be Elpida and Infineon having made friends with Chinese semiconductor giant - SMGC.
Secondly, we should keep in mind all those 300mm wafer fabs coming into service throughout 2003. Their manufacturing potential is immense. Here is an example: a new Micron fab in Virginia has already made the first samples of 0.11micron 300mm wafers, and when it reaches its full productivity, it will be able to produce 45% of the overall memory amount produced by all Micron fabs now! And there will be several fabs like that coming into operation this year.
Yeah, they are very unlikely to reach their peak productivity this year, but the shift towards 300mm wafers and 0.11micron manufacturing technology will be evident. One more example from Micron: 0.11micron chips have about 6% share in company’s total product output, but by the end of this year it will be about 50-60%.
By the way, do you think that the last year was a year of tremendous price drops and on the whole, just an unlucky year for DRAM makers? You are quite wrong if you do. Well, if Micron had paid proper attention to 0.13micron process and followed its habit of dumping, the prices would have been even lower! We mean that Micron did pay much attention to 0.11micron process, so beware: the company is going to have a good prime cost to fight with the competitors to the end. Plus Micron has carried out company restructuring as well as its competitors, which will allow them to sell chips at even lower price than usual.
Well, the memory market players will certainly need all their reserves this year. We may blame either Infineon or Mosel Vitelic, it is up to your taste. They quarreled apart in the end of 2002 over their joint venture aka ProMOS. So, what do we have now? And we have a company that hasn’t got any big clients onboard, but has started throwing its chips into the open market. The market reacted to this by a considerable price reduction in January-February. Is that all?
No, this is only the beginning. As the quarrel with Infineon was rather a unexpected thing, Mosel Vitelic had no production facilities at hand to package DRAM chips. And they needed enough facilities to digest about 60 thousand 200mm wafers from ProMOS at that time. Mosel Vitelic has got them now. And while the company’s output into the open market was only about 5-6 million chips in January, this number is going to reach 10 million in March! ProMOS is also expected to increase the production volumes of 300mm wafers from 7500 to 9000, which is about 3500 wafers in 200mm equivalent. There is one more thing: Taiwanese court confirmed that ProMOS has the right for 0.11micron technology from Infineon, although the latter is threatening with all possible and impossible measures. But Infineon has had no luck in lawsuits against ProMOS lately. The overall picture is impressive, isn’t it?
There is one more aspect in the situation. What was the main cause for the price growth last summer and fall? Firstly, the manufacturers were all going over to new production technologies, which required a rather long production lines halt. Secondly, there appeared first Intel chipsets with DDR SDRAM support and they provoked a jump in demand for this memory type. As the memory makers didn’t meet the demand and, moreover, hadn’t taken up DDR seriously yet, we experienced some shortages with all the evident consequences.
Does it resemble what we have now? Yes and no. On the one hand, new Intel chipsets supporting PC3200 are coming out. On the other hand, all the big sharks, like Samsung, have been producing it for a long time now. There are no causes and there hardly will be any to halt the production lines.
All these things mean we are up to a definite price reduction, at least until the coming summer.
What about the next summer? Well, there may be some changes then. As usual, most customers are now waiting for the price to reach the bottom and then start buying up. We haven’t reached the bottom yet, but it may be reached somewhere about this summer, when 256Mbit DDR will cost about $2-$2.5 and some manufacturers may be switching over to flash and other products and hence will leave the market. So, at that time mass purchases will start taking place with all the resulting consequences. Theoretically, it’s quite possible we see prices grow some time in summer.
There can be some smaller and really short-lasting price bursts in the next few months, though. We can recall the last November when the memory price soared by a third, but then quickly returned to the same point. We are still not quite certain what it was all about, because no demand growth was detected. But the market stood strong and we may only wish it to resist hysteria in the future. There are no evident causes for the price to grow; it will be going down until summer.