Memory Prices – A Rebuttal

Some Bad Information

Well, my hand has been forced. During the past week, I have been discussing the rising memory prices with people on the newsgroups. Just as I thought the rumor and misinformation had been dealt with – ACK! AnandTech posts an editorial on the subject with so much misinformation I was almost embarassed to read it. My intent is not to personally attack Anand (though I am quite sure I will now get hate mail from those who think he is a god, but that is a completely different issue), but to provide some factual information for those who just might care about such things. I think one can get enough sensationalist material from the tabloids. First, I will rebut each of the claims made in the editiorial in question, and then provide some boring facts about why prices have jumped so quickly

The first reason mentioned in the referenced editorial has to do with the Micron ‘dumping’ lawsuit against Taiwanese manufacturers. This one has been a favorite on the newsgroups, and is the only one with any grain of truth. Micron has used the threat of anti-dumping lawsuits against their rivals for almost a decade, with the most notable being the suit they successfully brought against Korean manufacturers in 1992. The most recent suit was filed in October 1998 against several Taiwanese manufacturers, whose total DRAM output is less than 4% of the global market. Interestingly, in March of 1998 the Commerce department levied a 12.64% dumping duty against Hyundai and LGS, who combined have a 21% global market share. The suit against these large companies had absolutely no effect on DRAM prices, yet we are now to believe that a possible tariff on 4% of the worlds DRAM has caused prices to double… nice try, but extremely doubtful.

The second possible reason given is the ‘shift’ to D-RDRAM by manufacturers. Unfortunately, this does not hold a drop of water because there has *been* no shift. Intel themselves indicated that they are supporting PC100 SDRAM on the Camino chipset because of the lack of availability of D-RDRAM parts – hardly an indication of any dramatic shift in production. D-RDRAM parts are extremely difficult to get ahold of, and they are very, very expensive (over 3 times what SDRAM costs even at todays prices). The reason for this is that manufacturers have *not* been producing D-RDRAM because they are not getting orders for the memory. In fact, I was told by a source within one manufacturer (who requested anonymity for obvious reasons) that Intel is the only real buyer for D-RDRAM currently, and they can hardly *give* the stuff away right now. No, I really don’t think this one even qualifies as a small possibility, and it is somewhat laughable that people have even brought this up

The final reason given in the editorial was a supposed problem with Micron (again) having huge numbers of .22 micron and smaller chips ‘unfit for sale’. This rumor cropped up several months ago regarding the .22 micron parts, and I asked some people within Micron to comment. These people laughed and indicated that they had been at .18 micron since before February. In fact, they indicated that their yields had increased by 75% in Q4 1998 as a result of their process improvements. In addition, Micron admitted to having a very large inventory of chips that they offloaded in early summer – resulting in an anti-dumping lawsuit against them by…. Taiwanese memory manufacturers! The actual fact is that 85% of all chips produced by Micron go to large OEMs, such as Dell, Compaq, Gateway, etc. (refer to Micron’s latest quarterly statement for this information) Certainly this would not be the case if their chips were defective.

Some Good Reasons

Now for the really boring part of this editorial – the bland facts about the rise in memory prices. It is important to note that memory manufacturers have been losing money for the past two years due to selling parts below cost as a result of a memory glut. It is a race between manufacturers to reduce the die size first, and therefore reduce costs first (more parts per wafer means lower cost per chip). Unfortunately, this has been an extremely competitive market and this race has been lost by virtually every manufacturer for the past few years. What this means, though it is not a popular thought, is that DRAM has actually been selling *below cost* for quite some time. There comes a point where a correction must occur.

Memory module prices are based primarily upon the price paid for chips. Most OEMs have long term contracts with the memory manufacturers – usually 6 to 12 months in duration – at guaranteed prices. Chips not sold in contractual deals are sold on the ‘spot’ market, at whatever the going price is. Most 3rd party module manufacturers purchase their chips at spot market prices, though some of the largest will have contracts. When prices are dropping, this is an advantage for these 3rd party manufacturers as they can offer lower prices than those with long term contracts. Unfortunately, when prices are going up their prices rise very quickly as well – which is what we are seeing today.

Back in May of this year, SDRAM chips were at an all time low of less than $4 for a 64Mb (megabit) chip. It takes 8 of these chips to build a 64MB (megabyte) module, plus a printed circuit board (PCB). The PCB is very inexpensive (a dollar or two, USD), so the overall cost of parts for a 64MB module was less than $35 at that time. With prices over three times that today, a 3rd party manufacturer will be paying over $100 for the parts to make a 64MB module, though large OEMs will not have been affected because of their contract pricing.

The cost of producing a 64Mb chip is approximately $5, according to those I have spoken to so this caused a great deal of concern amongst memory manufacturers. Many of them began to shut down production lines to save on costs (lay off people, save on some overhead) and to help reduce the memory glut. Though Micron sells most of their chips to OEMs, this is not necessarily true for all manufacturers. During this period, several DRAM manufacturers were offloading excess memory chips – including Micron and Infineon (formerly Siemens), which helped exacerbate the situation.

By late June, prices had begun to rise again (to about $7 per 64Mb chip), but production of chips was still keeping up with demand. For a few months prices remained stable – until the end of August. Suddenly, demand in Asia increased dramatically. It appears that the depressed economies in that region had caused businesses to push back their Y2K expenditures, and had kept many consumers out of the market. After several months of economic recovery, it seems that consumers and businesses decided that the time had come to open up the pocket books. Increased demand for the holiday season (by OEMs) has also contributed. For those who have been paying attention, there have been stories of chipset, LCD and even hard drive shortages over the past month – not just DRAMs.

During all of this, manufacturers have been not only shutting down some production lines, but had been quietly transitioning their fabs to produce 128Mb chips, rather than 64Mb chips. Unfortunately, the vast majority of motherboards currently in use cannot use 128Mb technology so most 3rd party manufacturers want only the 64Mb chips. One of the drawbacks of shutting down a production line is that it takes 12 weeks from wafer start to finished chips, so by the time a shortage occurs, it is 3 months too late to do anything about it. In fact, this is one of the reasons that there has been a glut for so long – restarting a production line is expensive, and it is usually cheaper to continue to produce than to stop/start the line.

The manufacturers have now found themselves on the horns of a dilemma. If they start up 64Mb production lines they will not have any parts available until after the holiday season is over (and after all the Y2K preparations are completed), and at the same time they will have produced obsolete parts because 128Mb chips will most likely be mainstream by then (all new chipsets, including those from VIA and Intel support 128Mb technology) making them basically worth less than the cost to produce them. It is also interesting to note that 128Mb parts are readily available, but few module manufacturers want to use them. In fact, as of early this week the spot market price for 64Mb and 128Mb chips were almost identical at somewhere above $15 apiece. 128Mb chips have been steady at about $15 or so for many months. It is even more interesting that when all configurations of DRAM are considered, there is still a memory glut.

Some Ugly Thoughts

During the past week of answering what I believe are ridiculous allegations on the newsgroups, I had briefly considered writing an article on the subject. I had decided not to do this because I figured the reasons were fairly boring (the old supply and demand issue), and prices would come back down a bit in a short period of time. As indicated in the start of this lunatic raving my hand was forced by the mentioned editorial. Normally I would not pay much attention to such an article, but AnandTech has gotten a reputation for being knowledgable and accurate amongst many ‘netizens, and I felt that this was too egregious an error to simply let it pass.

Unfortunately, this brings up the ugly issue of journalism. It seems that the name of the game for many ‘hardware’ sites is blatent advocacy or sensationalism. Articles which are supposed to be ‘reviews’ are in reality simply editorials railing against manufacturers and their ‘nefarious’ activities. Reports upcoming products are based upon rumor gleaned from other sites rather than directly from manufacturers. While it can be argued that there is a place for such things, it seems that a site that professes to be an ‘objective hardware review’ site should avoid such practices. It stinks of something called ‘yellow journalism’ that prevailed in the early days of print journalism.

Lest I be taken the wrong way, I am not suggesting that the editorial on AnandTech is one of these articles – to his credit, it is billed as an editorial. On the other hand, it would seem appropriate when writing about such a subject to have done at *least* a little bit of research and engaged in a smidgen of critical thinking about the supposed reasons for the price hike. On the other hand, it does make a much more exciting story than the one I have told…

Be the first to discuss this article!