Thoughts on the DRAM market:
Unlike burgundy in bottle that increases value over time, or even wheat that can be stored and sold later in a (hoped-for) stronger market, semiconductors are basically landfill in the making (ouch! I know that sounds terrible, but it has certainly been the case). Given the half-life of ICs (with some being even shorter than DRAMs), I wouldn't be at all surprised if Econ 101 starts using DRAMs instead of wheat as the supply/demand example.
So when I hear that the DRAM market is strong, it leads me to wonder. When suppliers were all chorusing last year that demand was strong despite the low prices, I couldn't help but think that the wording should have been "due to" instead of "despite". The DRAM market is brutally effective at clearing itself through the use of price - supply and demand will always been in balance (it just depends at what price). The demand is virtually limitless without the pricing consideration, while supply is fixed at any point in time by the wafers that have been started in the past three months and the yields of parts on them. Over time, supply growth will accelerate or decelerate depending upon perceived rates of return.
So we have an ongoing struggle between the OEMs' memory budgets per box (and the number of boxes being produced/sold) and the suppliers' ability to fund and drive production. What happened last spring was bizarre (60% price decline in 64Mb in six months, as all the suppliers appeared to chase relatively high bit pricing compared to the 16Mb part), but was the mirror image of what happened in 1995.
This leads me to MU. Micron was an extraordinary beneficiary of the industry's inability to migrate to the 16Mb part in 1995 - as I understand it, the 1x16 part for the PC market suffered unacceptable amounts of noise and crosstalk on the bit lines when the parts were made on 0.40-processes and the delay in retooling to 0.35-micron was long enough that MU's production of 4Mb parts on what was essentially fully-depreciated equipment was a gold mine. It's hard to see an equivalent scenario developing - the industry is pushing new design/processes hard (and many are backing into the 128/256 realm from designs and processes developed for even higher density parts), Micron will be using tools with inherently higher depreciation costs and the buyers' shift to build-to-order from build-to-plan would appear to preclude a repetition of the hoarding that took place in 1995 right now (although it could very well happen next year).
Having acknowledged all that, it is true that Micron has one very special attibute - it is the only American-based supplier of DRAMs. For that reason, it should enjoy a premium among investors. The question remains: how much of one. - Tad LaFountain |
