Hi pompsander; Great post, I think you covered the reasons for owning RMBS right now perfectly.
Where we agree completely is that what matters most is marketing. Where we disagree is in how important marketing is, and who has how much of an advantage in it. These are subjects outside my specialty, there is a chance you are right, and that Intel will prevail.
But it is clear to me that the memory makers have succeeded in bringing DDR out, and that it will be not just cheaper than RDRAM, but considerably cheaper, over the next few years. Probably a significant question is how much Intel can convince the motherboard makers to support RDRAM, or, failing that, replace their production with their own.
That Intel is in a position to subsidize RDRAM machines I have no doubt. But I also have no doubt that the memory makers are in a position to drain quite a bit of cash out of Intel. I suspect that they are planning to do this. In no way are they going to step all over each other in order to turn the memory business into a profitless commodity business.
Hyundai said that they couldn't make money on RDRAM even with it selling for 3x SDRAM. Samsung says that they are making money selling it for 2.5x. In order to get RDRAM cheap, you have to get at least three vendors capable of making it cheap. Until then, monopoly pricing obtains and the price stays high. This is a very happy situation for the memory makers that can make good profits at that price, and the other memory makers don't care that much, as long as they can make money selling into the much larger SDRAM market. They are all (pretty much) capable of biting the bullet and getting their costs down enough to sell RDRAM at 50% over SDRAM, but there really is no reason to do this, there is plenty of demand for other SDRAM right now, and companies do not always make their decisions based on the most obvious economic profit motives.
Marketing is powerful, but don't exaggerate its importance. It hasn't, for instance, kept the Edsel in production. If you look at the responses to the recent Anand editorial you get an idea of what a good bit of the community thinks about Rambus out there. That community is what will provide a happy home for initial shipments of DDR, and they will ensure that DDR remains in production for years. I believe that it is now too late for RDRAM to capture enough of the marketplace to become the cheaper solution.
The marketing of DDR systems hasn't started yet. While it is true that Intel is bigger, they are only about equal in sales per year to the total mass of the memory makers (i.e. about $30 billion per year). The memory makers can pay for a lot of advertising, and their main task is to advertise to the motherboard and box maker engineers, not to the general public. Engineers are a tiny group, and relatively cheap to advertise to.
While it is partially true that Marketing likes to think that they tell Engineering what to design, it is even more true that Engineering gives products to Marketing to market. Like you said, Marketing can find nice things to say about it, no matter what piece of feces the engineers come out with. (Incidentally, that is why one should always get the sales people really drunk and friendly before asking them the truth about their product. Or simply ask one of the engineers.) Because of these ease that marketing has with supporting different products, the real underlying dynamic, in the absence of a corporate decision to tie itself to a technology like RDRAM, is to build the cheapest product that meets the customer's desires. Cost matters, and it is easier to sell performance than glitter. You can always add glitter to a good value product and get it to sell even better, no question, but the best result happens when you put out good product with good marketing.
As far as the box makers go, CPQ sells $38 billion per year, IBM sells $86, HWP sells $45, and Dell sells $28. Even Apple sells $7 billion per year. (Yahoo figures.) We have all seen Apple advertisements, but they are tiny compared to Intel. Also, by the way, note the industry rumors regarding Apple's plans:
Chip makers also expect Apple to launch a line of desktop systems this year that will use DDR SDRAM as main memory.
techweb.com (January 14, 2000)
Any of these companies can market any kind of memory they want to, and not all of them have their interests completely devoted to supporting an Intel/Rambus duopoly. Several of them are in development of their own DDR chipsets.
As far as AMD producing a good line of future microprocessors, who knows. While their continued success is undoubtedly good for DDR (as is the success of Apple), it is no longer necessary. The Micron chipset, which appears to be the first out the door, supports Intel CPUs, and will probably sign up a lot of board makers. Micron had that chipset first running back in November of last year, it is probably on third silicon by now and ready for production:
techweb.com
No, I think it is too late to put the DDR genie back in the bottle.
But thanks again for the beautifully reasoned post.
-- Carl |
