In some of the first articles (1 2) I wrote on Micron (MU) I told the readers about the unique ability to watch the progress in DRAM price appreciation, and, presumably, the future fortunes of Micron, by watching and recording the DRAM spot price action on DRAMeXchange. Prices up , green light; prices flat, green/yellow; prices down, green/yellow; prices down a lot or many days, red light.
That has worked quite well since the Feb 3rd article the stock price has gone from $7.83 to $14.71 today, as the price of DRAM has increased over 100%. That was all the simple stuff.
We (and analysts) have used the 2Gb DRAM ship as the benchmark for pricing. Recently the price of these chips have been declining, then in an incredible turn around, the spot price of 2Gb DRAM chip increased nearly 8% in the yesterday session.
So, there is more to this pricing thing than watching the price of one part. DRAM generations are always increasing in density. The 4Gb chip sold today is 1,000,000 times the density of the 4Kb chips I sold as a rookie.
I need to bore you with progression of DRAM densities during the past 40 years:
4Kb, 16Kb, 64Kb, 128Kb, 256Kb, 512Kb, 1Mb, 2Mb, 4Mb, 8Mb, 16Mb, 32Mb, 64Mb, 128Mb, 256Mb, 512Mb, 1Gb, 2Gb, 4Gb, 8Gb - future.
So, there have been 19 transitions from one density to either 4 times that or 2 times that. You get a million times more bits than a 4KB parts for about 15% of the price of that 4Kb part (corrected for inflation.) Amazing business, isn't it? The point of this article is that there are some common characteristics involved in every one of those 19 density transitions.
Here we go:
When the new part first becomes available it is very dearly priced. This is to pick off those customers who really need the new higher density and are willing to pay for it. Then as output of the higher density part increases, the price declines rapidly (often willingly driven down by the manufacturers of the DRAM!), this encourages use of the higher density part. Along the way a tipping point occurs where the higher density part actually cost less than two of the lower density parts being replaced. This tipping point is somewhere around the point when 50% of the bits are produced in the higher density part. The end point is when two of the lower density part cost much more than one of the higher density part (check 1Gb DDR3 parts, for example.) The lower density part goes obsolete shortly after that.
Still with me? Good. The 50% point for 4Gb chips and Micron occurred in the last quarter (a mumbled detail in the conference call.) By now, the output of bits is probably 75% 4Gb chips. So, as this progresses, the 2Gb chip becomes less important to the business and the spot price for it conveys less information.
The reason for all this is that these conversions can't happen overnight; they are phased into the market. By the 50% point the DRAM manufacturer is making much more money on one 4Gb chip, for example, than they do on two 2Gb chips.
At this point in a density conversion, pricing for the "old" part becomes extremely volatile. Manufacturers might have temporary excess inventory that needs to be sold. They could have temporary shortages that cause price spikes. Even OEMs get into the market; they might have excess inventory than needs to be sold, maybe at any price. Since the volume of 2Gb chips is relatively small now, even manipulation of price becomes possible. Before long the 2Gb chip will become like the 1Gb chip (which is still available), no one will care about it.
From here on, if you are going to continue to play the Micron story based on DRAM (NAND is another story and will take care of itself), you need to focus on the 4Gb chip only (and watch for the introduction of an 8Gb chip, which will come.) We would like to see the price of the 4Gb chip stay at about $3.00 give or take. The current manufacturing cost of a 4Gb chip might be around $1.50, which makes for 50% gross margins (higher for Elpida since they don't have much depreciation due to the bankruptcy.) So, if 4Gb commodity DRAMs stabilize at about $3.00, Micron will do very well. Even as we worry about price Micron is shrinking the 4Gb chip which will increase gross margins. The shrink will push cost on the 4Gb chip down toward $1.00, and the shrink will bring on the 8Gb chip at an initial price of $8-10. Then rinse and repeat.
BTW, an 8Gb chip in a low power version will allow 1GB for iPhones in one chip (instead of two, stacked), 2GB for some of the new phones in two chips (instead of four), in standard power it will allow for the obligatory 4GB of PC RAM in a single, simple four chip module.
Stay in Micron.
It IS different this time.
Take analysts' recommendations with a grain of salt. After all, even they don't see things like Micron and Elpida happen every day. They pour over numbers all day to try to pick out a penny or two of more or less earnings in most of the companies they cover. That's their world. Micron "buying" $6-8 billion worth of now valuable assets for … umm … NOTHING is quite foreign to them.
In my model, Micron is worth $35-45 today and heaven only knows what it will be worth next year.