Good read from motley fool..explains rmbs-triple/amd-single digits.
Of the approximately 10 analysts following the stock, only 3 are now willing to even venture an opinion, down from 4 previously. Few analysts want the embarrassment of giving a poor rating to a stock poised to explode; yet most analysts know that Wall Street hates uncertainty. The latest height of controversy was reached Thursday and Friday, October 19-20, 2000, when twin panics swept over both those most pessimistic about the stock (10% of the float is still short) and those most optimistic. The stock plunged to $50, rose rapidly to $70, and nose-dived again Friday. Bizarrely, the very same report, harsh comments made by Intel's CEO Barrett on Wednesday, almost surely contributed to the panic of both sides. The professional shorts, reading Barrett accurately, reacted promptly, driving the price up Wednesday afternoon and more sharply Thursday. Longs took longer to digest the news, driving the price down Friday. Long time Rambus investors appeared on the boards Friday reporting that they had sold off 2/3 of their stake after Barrett's remarks. They also wondered why the stock was falling.
An extremely heavily shorted stock early in the year (>50% of the float was short in February), many of the hedge fund managers had moved on to fatter dot-com targets after Rambus began signing up Dram manufacturers for SDRAM royalties in June. A few hearty skeptics remained however, pointing out that the elements of these deals were secret (as such deal almost invariably are). They argued it was easy for Rambus to reach "agreements" so long as no serious money was involved. Ironically, the filing by Micron complaining of the "exorbitant" fees Rambus was seeking almost surely helped Rambus, keeping the stock relatively buoyant during the chip sectors hard fall. It is hard for anyone to read the Micron complaint and not gain an appreciation of strength of Rambus's position. Yet into July and August articles in the MorningStar, Individual Investor, and elsewhere suggested that there was no real money to be had from Rambus's legal strategy.
Two things changed that impression quickly. First, the Q4 '00 Rambus report (October 18), showed royalties already hitting $20 million/quarter, with none of the 3 major Dramurai yet signed. Moreover, chipset royalties, believed to be equal or greater in value than the DRAM maunfacturers signings, were not incorporated. Yet Rambus management promised that more signings were imminent. Simple calculations based on the reported royalty income strongly suggest that Rambus is indeed charging about 1% on SDRAM royalties as they have publicly implied.
In this context, Barrett's lament about Rambus becoming a "toll taker" hardly was reassuring. There have already been broad hints that Rambus and Intel have renegotiated their contract, presumably to allow Intel freedom of choice in chipsets for the P4. In exchange for releasing Intel from their contractual obligations, Rambus would presumably ask for a smooth settlement of the SDRAM/DDR royalty issue – money that Intel would inevitably have to pay anyway. Certainly the original contract forbade any non-RDRAM P4 chipset with more than 1.0 GB/sec transfer; yet the Brookdale (PC133) chipset violates this, and Intel has announced that it is evaluating the practicality of a DDR chipset for the P4.
Some wonder why Barrett developed a sudden distaste for a pure IP company collecting royalty. The hypocrisy reminded many of the inspector in Casablanca who was "shocked, shocked" to find out that gambling was conducted at the local casino. Surely Barrett knew all along that this was the Rambus business model. The answer is that the original Rambus/Intel contract provided the possibility of royalty free RDRAM use by Intel alone, provided certain extremely rigid criteria were met. These required Intel to virtually force a mass conversion by the industry to RDRAM (e.g., the criteria included RDRAM requirements for each DRAM manufacturer, not just for Intel). Barrett evidently thought Rambus royalties were a fine thing when applied to Intel's competitors. This would have given Intel a permanent cost advantage. However the contract requirements were never approached. Royalties for IP are no doubt a sound principle when Intel collects tolls from Via or Broadcom; or even when Rambus collects tolls from AMD or Via. But Barrett evidently feels Rambus collecting a toll from Intel is a sound principle carried a step too far.
To those knowledgeable about Intel and the chip industry, the idea that Intel would enter into a death match lawsuit against a partner with signed contracts, excellent funding, and all the legal cards is absurd. Like the world's number 2 and number 3 chip companies, NEC and Toshiba, Intel doubtless has closely examined Rambus's IP claims and knows how strong they are. Intel also faces the problem of the companies joint history: after signing a NDA to access Rambus's IP in 1992, Intel incorporated much of it in their 1994 SDRAM "open" standard. Indeed, Intel, more than anyone else, could be said to have induced manufacturers to infringe Rambus's patents. Intel has been doing so in its own chipsets for years. Finally, Intel has the special problem of the 1996 contract with Rambus, which Intel plainly wanted to alter.
It is one thing for a relatively small company that frequently teeters on the edge of bankruptcy, such as Micron, to play high stakes poker. They have done so before, losing to Texas Instruments, for example. It would be quite different for a rich powerful company such as Intel, NEC, or Toshiba to do so. With Intel's high margins, contractual obligations to Rambus, and embarassing history in the development of the SDRAM "open standard", a war with a litigator as ruthless as Rambus is extraordinarily improbable.
Thus, given Rambus's strong royalty results Wednesday, professional shorts were only further alarmed by Barrett's complaints about the "mistake of getting into bed with Rambus", and Rambus's status as a "toll taker." Royalties on controllers are reported to run in the 3-5% range, with the market size at tens of billions. Depending on the exact royalty figure, and Intel's percentage of DRAM controllers, a settlement with Intel alone could, at a single stroke, add anywhere from $1 to possibly several dollars per year per share to the Rambus bottom line. I believe such a deal has already been reached. In addition to the reasons aforecited (Intel, of all companies, could never risk a legal fight with Rambus; Intel wanted more flexibility in their contractural obligations; Intel has already hinted at newfound ability to produce non-RDRAM chipsets for the P4; and Rambus management has promised the announcement of new signings "soon"), Intel told FredHager.com that Barrett's comments referred back to an already settled situation that arose in June, when Intel first talked about a non-RDRAM chipset. Intel and Rambus might chose to delay the announcement until after the launch of the Pentium 4 in order not to confuse RDRAM manufacturers.
Clearly Barrett is right that Intel made a mistake getting into bed with Rambus. None of the advantages they hoped for (such as being the only chipset maker with royalty free use of RDRAM) materialized. Meanwhile the 1996 contract that required Intel to attempt a mass industry conversion has left Intel with an extraordinarily weak position in bargaining chipset royalties with Rambus. And of course, Intel could have simply chosen RDRAM for workstations and high-end PCs without making any special contractual concessions to Rambus. The dream of being the only company with royalty free chipsets evidently led Intel astray.
RDRAM does offer the highest performance. This position was reiterated by Intel Thursday and Friday in response to several media requests, and is scarcely doubted by anyone except AMD hobbyists. The elite of technology from Sun to Sony, from PMC-Sierra to Vitesse, from Brocade to Compaq/Digital, from HDTV makers to OC-192 designers, agree with Intel on the technical superiority of Rambus's approach. Even Via called RDRAM a "wonderful" technology at the 2000 Independent Platform Forum (before hastily denouncing the price). And AMD has licensed RDRAM, although according to Rambus officials, they have been told that AMD will not bring out a RDRAM chipset until prices rare within 10% of SDRAM prices. From Samsung's data, that is not expected until 2002.
Obviously, the price of RDRAM has declined at a significantly lower rate than Intel expected. Thus, while the preferred option for high-end systems, Intel realizes that a mass conversion of the industry to RDRAM in one or two years is not possible. Intel is entering into a price war with AMD, and cannot afford to have a significant price disadvantage for the mainstream and value markets. The "best performing" technology always has a place; the size of that role depends on price.
I believe Rambus critic McComas has probably summed up the new Intel view best. He believes that Intel will no longer try to "push" RDRAM on the market, but will let the market place develop "naturally." Since the natural development is for the price premium of RDRAM to fall at smaller etching sizes and with larger memory units (both the die size premium and the yield rise closer to SDRAM as these things happen), RDRAM prices will slowly but inevitably approach SDRAM over a period of several years. Meanwhile, the performance advantage of using RDRAM rises steadily as processors, front side busses, and chipsets speed up, and more users take to high-bandwidth applications. The virtuous cycle of declining costs and higher performance gain prompting greater volume manufacturing and further cost cuts is still in place. |
