Short RMBS? Read this (author has a RMBS enabled server farm @ Hopkins):
There is no DRAM but Rambus
by: ptnewell 6/22/00 11:12 am
Msg: 117388 of 117494
The second rise of Rambus began with an understated press release issued late in the evening of June 14. Indeed, so blandly worded was the announcement that Toshiba had agreed to pay Rambus royalties on SDRAM (including, or rather especially, on DDR-SDRAM) that mainstream media initially took no notice. Yet those familiar with the story, whatever their allegiance, recognized the thunderbolt for what it was. Put simply: There is no DRAM but Rambus.
The scenario of a monopoly more far reaching and more absolute than that of Microsoft or Intel loomed plausible. Rambus?s lawsuit against Hitachi, filed in January 2000, claiming dominion over SDRAM had initially been dismissed as a nuisance by the companies many outspoken critics. Independent observers soon reached the opposite conclusion, namely that Rambus had a strong case. Regardless, no opinion by any analyst could possibly carry the same weight as Toshiba?s decision. A rich multifaceted memory manufacturer, after careful study by patent attorneys and engineers in the business, decided that the case was too strong to contest ? even though Toshiba had not yet even been sued. Short of actual court decisions, no better opinion on the merits of the case can be imagined.
It was the perfect setup for a short seller?s massacre. Rambus stock surged 30% in premarket trading on news not even recognized as noteworthy by the mainstream media. By the time Morgan Stanley Dean Winter?s semiconductor analyst Mark Edelstone appeared on CNBC to announce an upgraded rating (to "strong buy") and price target ($200), Rambus was up 50%. Those who refused to believe the initial 30% rise could be real unless the news was widely trumpeted suffered badly.
A fascinating aspect of the situation which has received little attention is the apparent role of Intel in violating Rambus?s intellectual property. Both companies certainly have good reason to downplay this aspect, but the historical developments show the problem. Rambus was formed in 1990, when it applied for patents for techniques to create high speed memory. Rambus and Intel were involved in talks for years, ultimately leading to Intel?s adoption of RDRAM. In the interim, however, Intel ? in 1992, after two years of discussion with Rambus ? laid down the "open" memory standard for Synchronous Dynamic Random Access Memory, today?s SDRAM. If SDRAM violates Rambus patents, then clearly the promulgator of the SDRAM standard, Intel, must be the initial culprit. Given the companies disparate sizes ? Rambus was not even able to make an IPO until 1997 ? it casts a somewhat different light on the role of Intel as Rambus?s champion.
Rambus clearly now feels established enough to assert its IP rights. Despite Hitachi?s bitter protests, the Sherman anti-trust act does not make a monopoly illegal. However a monopoly is constrained in ways other businesses are not. In particular, a monopoly has far less power to reward its friends and punish its enemies than does a company with competitors. Intel, which agreed to such terms had its anti-trust case dropped, while Microsoft, which steadfastly insisted on such behavior faces breakup. It is imperative that a victorious Rambus not seek to punish its enemies by selective licensing. In this respect, the aspect of the Toshiba agreement which requires paying higher royalties for DDR-SDRAM is regrettable. Obviously Rambus believes DDR is an attempt by DRAM manufacturers to steal RDRAM without paying for it. Many of the techniques in DDR stem from RDRAM; in fact so blatant was the DRAM design committee that even the name "DDR" (Double Data Rate) was taken from previous Rambus announcements. Tempting as it may be to flex its new muscle, Rambus can best protect its fledgling monopoly by being as uniform and non-obtrusive as possible. Certainly any thoughts of "revenge" on Micron for its strenuous anti-Rambus and pro-DDR campaign should be discarded.
How does the decision alter Rambus as an investment? It is clear that most of the surge is due to the "insurance" aspect. If RDRAM becomes the industry standard, long term earnings will be unaffected. Nonetheless if, as now appears to be the case, Rambus collects royalties on virtually all types of high-speed DRAM currently contemplated, Rambus becomes a much safer stock. In essence, there are now two distinct paths to fortune, either of which is sufficient: (1) Win the legal battle, to which the Toshiba decision is the best available guide; or (2) win in the marketplace, which remains controversial. Even should the latter happen, the decision still makes Rambus more valuable, in that money received now is worth more (because it can be reinvested, or used to develop new products) than money received later. Obviously even if RDRAM ultimately wins, SDRAM will still dominate for at least the next 18 months, and perhaps longer.
In this regard, it should be noted that even before the Toshiba announcement, Rambus had been recovering steadily albeit more slowly from the Nasdaq collapse (up 60% from its low). The continued absence of a DDR chipset, almost a year after Micron promised their imminent arrival from multiple vendors, was being noticed even by advocates. Third party observers such as the Register wondered what the problem was. Clearly time is running out for DDR.
In February 2000 Intel?s Vice President Pat Gilsinger had announced that "DDR is too little, too late, and it doesn?t work in a desktop." By too little, he meant that although 266 MHz DDR was adequate up to a CPU speed of 1 GHz or a little beyond, it would already be a severe processing bottleneck by about 1.5-2 GHz, a level Intel hoped to reach within a year. By contrast, RDRAM performance actually improves with such higher CPU speeds, as the memory takes fewer power-conserving naps (the cause of the over-hyped "latency" issue). By contrast, conflicts between reads and writes greatly worsen SDRAM and DDR latency at higher CPU speeds. Already at 1 GHz standard benchmarks done with production model machines clearly favor RDRAM machines, whereas results had been more ambiguous in the early tests using 600 MHz processors. The same advantage for RDRAM (or actually a somewhat larger advantage) should hold at 1.5-2 GHz comparing DDR to RDRAM. If DDR was not "too late" in February of 2000, the clock is surely ticking. |
