Since Yahoo lifted one of my posts over the week-end, I took the liberty of reposting this from the RMBS board on Yahoo. It's a great summary.
Rambus Explosion
by: ptnewell 3/6/00 8:39 am
Msg: 53715 of 53816
This may interest those confused by the rapid
rise in RMBS share prices.
The Rambus Explosion
Over the three weeks ending March 1, the price of Rambus nearly quadrupled. After ten years of intense controversy within technological circles, and of steady innovation, Rambus was an overnight sensation on Wall Street. The story was the more dramatic because of the large amount of "smart" money bet the wrong way. Rambus had had an amazing 50% of the float shorted (half the shares had been borrowed from a broker and sold in the expectation the price would drop). Many of the shorters suffered severe losses or ruin. Many were technically knowledgeable individuals, and several hedge funds were reportedly short on Rambus. However equally passionate and smart (or at any rate luckier) individuals had gone deep into debt to bet all on the company. Many industry pundits who took strong anti-Rambus positions have had their reputations shredded. Others, like Morgan Stanley Dean Witter semiconductor analyst Mark Edelstone, find their already good reputations enhanced. And millions of small investors are significantly poorer or richer depending on which side they took. Some probably have been ruined.
The battle of Rambus is still being fought in the technical world, in the judicial system, within the highly charged politics of the PC builders, and in the media. And, of course, it is fought five times a week on Wall Street. The war has been won, but increasingly quixotic skirmishing continues.
Rambus has designed by far the fastest memory in the world, called RDRAM (Rambus Dynamic Random Access Memory). It doesn?t actually build RDRAM. Instead it wants the big DRAM manufacters to build RDRAM and pay a royalty fee of about 1.75% to Rambus. Nice work if you can get it. Manufacturing a commodity like wheat or DRAM chips is a low margin and cyclical business. Collecting royalties is pure profit.
Why would the DRAM manufactures play along? Actually, many have been reluctant, but they are signing up in a rush this year. They are doing so because Rambus has the patents for making the fastest (highest bandwidth) memory. Because the existing rival forms of DRAM apparently violate Rambus patents. Because the CPUs (brains) of computers have been getting ever faster for years, while the memory hasn?t kept up. And most controversially, because Intel, the 800 pound gorilla of the industry has decreed that it be so. While that has doubtless opened doors for Rambus, it has also fostered great animosity, both by Intel rivals (e.g., AMD), and by some PC box makers and DRAM manufacters who thought they were being dictated to by big brother. Intel?s position is that it knew what was best for future industry growth. The fact that it has proven right (at least in this case) doesn?t necessarily make swallowing the prescribed medicine more palatable
These rivals have been suggesting a less ambitious upgrade of SDRAM, called "DDR", could suffice, at least for a few more years. DDR would be closer to what the industry was used to working with, might be cheaper, wouldn?t require royalty payments, and ? best of all ? wouldn?t be RAMmed down their throats. DRAM manufacturer Hitachi in particular has been aggressively pursuing the DDR option, and AMD and Apple announced (still vague) plans to head that way. A series of delays in the development of RDRAM gave opponents ammunition. Moreover, the first versions of RDRAM were not obviously superior to even SDRAM in some ways. (Notably the component latency was actually slightly higher than SDRAM).
Rambus filed suit against Hitachi in 1999, pointing out that DDR violated several Rambus patents (these patents were granted in 1999, but applied for in 1990, making the latter the date of legal validity). In fact as industry watchers and lawyers analyzed the suit, it became obvious that the existing standard DRAM, called SDRAM (introduced in 1992), also violated Rambus patents. There is some question about whether some of these Rambus patents are too broad; a court may have to decide. But Hitachi is in a particularly bad situation since DDR uses techniques similar to some Rambus shared with Hitachi before Hitachi decided not to go with RDRAM.
In hindsight it is obvious that by the end of 1999, just as anti-RDRAM fervor was reaching its greatest heights, RDRAM had become unstoppable. Rambus had upgraded RDRAM to the point where those not caught up in the politics of the PC world were all choosing it. Nintendo chose RDRAM. More importantly, so did Sony, for their Playstation2, which looks to be a monster hit (it will also double as a DVD and Internet connector). Makers of Digital High-Definition Television sets were all choosing RDRAM, as were makers of Digital Signal Processors (DSPs), used for ultrafast Internet connections among other things. Three DRAM manufacters had committed to mass production, including the world?s largest, Samsung. And Rambus announced forthcoming RDRAM upgrades which would quadruple the bandwidth again, further leaving DDR (the evolutionary dead end of SDRAM) in the dust. Finally DDR remained vaporware, with growing doubts about it?s stability, and about the assumption that it really would be cheaper, if it ever got built at all.
But the anti-Rambus "smart" money actually upped their bets against Rambus in the month leading up to the explosion. The share price, which had hit 120 in the summer of 99, dropped to about 70 in January 2000, buffeted by strong short selling. At a DRAM industry meeting in January, most analysts predicted an explosion of RDRAM this year, with, e.g., DataQuest predicting RDRAM would capture 50% of a $30 billion market by 2002. However Sherry Garber of Semico predicted RDRAM would be a niche player, having only 0.8% of the market by 2002 (and dropping).
Prophets of doom always get the press. Who remembers the hundreds of pundits who said Y2K would be nothing? Who doesn?t remember Deutsche Bank?s Yardeni, who had a best seller predicting disaster? Predictably, the headlines coming out of the meeting played up Garber?s prediction, with the mainstream analysis buried deep inside the stories. The anti-Rambus people felt confirmed.
In the media battle, they continued to win the wars. But as RDRAM began shipping, and Dell enthusiastically adopted the use of RDRAM in their top line models, the price of RMBS (the Rambus stock symbol) began moving. The price shot up to about $120 on the day that Dell began shipping PCs with RDRAM. Before the Intel Developers Forum (IDF) meeting starting February 14, 2000, Mark Boucher proclaimed that "Industry support for RDRAM is evaporating. Intel may have a few words claiming continued support, but they will quietly back away from RDRAM toward DDR."
At the meeting, Intel Vice President Pat Gelsinger announced (February 15th, I believe) that "DDR is too little, too late, and it doesn?t work in a desktop." Intel did agree that in certain applications where it was more important to have a lot of memory than to have fast memory ? mainly servers ? they would continue to support SDRAM or even DDR. This marked the beginning of the end. The anti-RDRAM crowd claimed victory also, based on the comments about servers, but few bought it. The price of RMBS continued to rise. The media initially reported the action as panicked shorts covering (buying shares to repay the brokers they had borrowed from). In fact, Barrons has reported that short interest actually rose that week. The shorts shorted even more shares, but they lost out because new longs (including me) piled on even faster. As the price rose, Wall Street began to take notice. A SG Cowan analyst, who had downgraded the stock twice in a row just before it exploded upward maintained in an interview that the "stock was getting ahead of itself." He allowed as how RDRAM had the better technology, but said it was still "an all or nothing play." He re-iterated his "hold" rating (usually a polite way of saying "sell"). By then he was probably praying for failure. There is nothing worse for a Wall Street analyst than to downgrade a stock twice, just before it?s price explodes, especially when other analysts had had "strong buy" ratings.
Over the next two weeks, enormous good news went right for RMBS longs, much of it outside media scrutiny, but all reflected in the share price. Hyundai, a long time RDRAM skeptic, changed their tune, saying that they "wanted to make the chips our customers want." They announced an immediate start to RDRAM production, and intent to become a leader. They estimated that RDRAM would be 4.5% of production this year. Ten days later, they upped the estimate of RDRAM?s market share for this year to 5-8%, and 25% next year. Samsung claims RDRAM will be 20% this year and much higher next.
Intel bribed Infineon ? pardon, invested $250 million in Infineon ? to start RDRAM production. SDRAM prices plunged and RDRAM demand soared as Compaq, HP, Gateway, and Acer all joined Dell with RDRAM PC boxes, as did several smaller firms. Micron gave repeated clear signs that it will start RDRAM production next month (thereby choosing to produce high demand RDRAM chips over selling SDRAM for a loss). The price of RMBS kept soaring, briefly pushing past $300. CNBC stopped describing the action as purely a short squeeze, and interviewed semiconductor analyst Mark Edelstone of Morgan Stanley Dean Winter, who stated that he was quite comfortable with the price, and was in the process of re-evaluating earnings and price targets.
The Yahoo RMBS chat board is unusually active. The ideological shorts ? those who supported AMD or Apple and hated Intel ? began to get more desperate. One, whom the board has termed Sybill, was discovered posting under multiple aliases. Others, like this one, were more honest:
"Okay, I shorted RMBS and I'm ticked (and poor). Congrats to those of you who made money! There is so much mis-information concerning rambus I'll be posting my views in segments over the
next few days. As an industry person I see convincing evidence that RMBS will be only a niche memory technology?
?Rambus performance myth:
a. All independent benchmarks I've seen show 400MHz Rambus is anywhere from slow to really slow compared to EXISTING DDR solutions?"
I?d like to be able to feel sorry for this fellow and the other "smart" money that bet against RMBS. But he obviously has not kept up with developments (e.g., 800 MHz RDRAM is being shipped already, with further upgrades soon, while there are no "existing" DDR solutions ? it remains vaporware). I subscribe to the view that a "smart short" is an oxymoron; however skillfully one picks stocks, it is a benefit to have the odds favorably stacked by benefiting from the tendency for stock prices to rise.
This fellow, his compatriots, and various hedge funds, incredibly: shorted (risky) a tech stock (worse), which already had a large short float and therefore was explosively primed (insane) ? without keeping up with technical developments. One hopes his day job designing for AMD is safe.
As for poor Sherry Garber (with her 0.8% market share prediction) and the two Wall Street analysts who downgraded the stock before its explosive ascent, board members mostly hope that they had put their money where their mouth is, and were among the shorters.
In retrospect the fight was over by late 1999, when mass production of RDRAM which worked well began, Intel had committed itself to RDRAM with their i820 and i840 chipsets, and Dell had enthusiastically embraced RDRAM. But those who had lost their heads in the battle ran around for months afterward squawking. In hindsight an explosive price run up was inevitable when the victories Rambus had already won become recognized in the marketplace if not by opponents. |
