Alles hier nachzulesen:
By Phil Harvey
Redherring.com, February 29, 2000
If you had said Rambus (Nasdaq: RMBS) was a $300 stock a month ago, you would have been tossed in a padded cell. But that's just where the computer memory designer is headed -- with no end in sight.
Most investors bailed on Rambus in September, after Intel (Nasdaq: INTC) had problems implementing Rambus DRAM (RDRAM) in a chip set. They drove the stock as low as $51.50. But since then, Intel has fixed the glitches and reaffirmed its support of Rambus, driving the company's shares into the stratosphere. Tuesday Rambus announced that five DRAM suppliers are in volume production with its memory design, pushing its stock up as high as $303 in early trading, up 26 percent from its Monday close of $241. The stock is up more than 300 percent since January.
"What's happening is that Rambus has had so many disbelieve ... and now you have people looking on it as a real company," says Mark Edelstone, senior semiconductor analyst for Morgan Stanley Dean Witter (Nasdaq: MWD), who's been one of Rambus's most consistent supporters.
Redherring.com was one of the few stalwarts to declare in September that investors overreacted to news that Intel delayed its Rambus-enhanced 820 chip set and that Rambus was a strong long-term play.
If Rambus executives are surprised by the turn of events, they certainly aren't showing it. At an investor conference on Monday, Rambus chief financial officer Gary Harmon said that the company expects that in two years, more than half of the DRAMs being made will use its technology. That would make Rambus, which posted revenue of $43.4 million last year, a $300 million company by 2002, according to analysts' estimates.
At the Robertson Stephens conference in San Francisco, Mr. Harmon concurred with growth projections from market researchers Dataquest and Cahners In-Stat. Dataquest projects that Rambus sales will account for 16 percent of total DRAM revenues of $30 billion this year. It expects Rambus's market share to soar to 53.3 percent of the $63.1 billion market in 2002.
For those projections to come true, Intel must continue to support Rambus's technology over alternatives. Intel, which has a financial interest in Rambus, continues to persuade memory manufacturers to shift from making chip sets using SDRAMs, the memory used in most PCs today. Rather, it wants them to use Rambus's DRAM (RDRAM), which costs more to make but helps computer systems get the most out of today's fastest microprocessors.
The shift to RDRAMs is expensive for memory makers because Rambus DRAMS (RDRAMs) require a chip set that's not compatible with widely made SDRAM chip sets. This means memory makers need to spend hundreds of millions of dollars per company on assembly and testing equipment that's exclusive to Rambus memory. Additionally, memory makers have to pay Rambus an estimated royalty of at least 1 percent for each chip sold, according to a chip analyst who asked not to be named.
To help ease the pain, Intel has invested millions of dollars in memory manufacturers, including $250 million in Infineon, $400 million in Micron (NYSE: MU), and $100 million in Samsung, for RDRAM development.
At least five DRAM makers aren't blanching at the cost of making RDRAMs. Rambus announced Tuesday that two more manufacturers -- Hyundai Hyundai and Infineon Technologies -- had "passed component and system level validation tests," and that they would join Samsung, NEC (Nasdaq: NIPNY), and Toshiba in volume production of RDRAM.
Steve Cullen, principal memory analyst at Cahners In-Stat Group, predicts that by the end of the year Micron Technology also will be on board. That list is likely to keep growing. At its developers' forum earlier this month, Intel said it's choosing RDRAM over SDRAMs and Double Data Rate SDRAMs (or DDRs) for chip sets for mobile computers, workstations, and desktop PCs. Rambus executives have been reiterating Intel's position to anyone who'll listen since Intel made the statement.
"To put this in perspective, desktop PCs, mobile computers, and workstations account for about 60 percent or more of all DRAM sales," Rambus CEO Geoff Tate said during a conference call last Thursday. "The server market accounts for perhaps 10 percent of DRAM sales, and that's where Intel sees a niche," he said.
Even with all its chest-thumping, Rambus has yet to see its technology reproduced in volume. "Part of the reason [Intel and Rambus] make so much noise is to encourage vendors to get into the business," says Sherry Garber, a senior vice president with Semico Research in Phoenix. "This drives costs down as more units are produced, and Rambus gets royalties regardless."
Even with the renewed support from Intel and Rambus's soaring stock price, some investors are lining up to bet against Rambus. Earlier this month, short-sellers accounted for about half of the Rambus shares available. The spike in the stock price Monday drove some of the shorts running to cover their bets, says Mr. Edelstone of Morgan Stanley.
It's understandable why there's so much skepticism about Rambus. Remember the trashing Rambus took after Intel's famously inept public relations performance when releasing its 820 chip set? The 820, the first to use Rambus's memory-enhancing technology, was slated for delivery to PC makers at the end of September 1999. It didn't make it out the door until two months later due to "platform integration issues," Intel said.
The bigger problem at that time was, as Mr. Cullen puts it, that Intel "trotted it out at a time when Rambus was nice to have, but not necessary." When Intel ships its next generation, code-named Willamette, during the second half of this year, the company hopes things will go much more smoothly. Willamette's clock speed of more than 1 GHz and its move from a 133 MHz system bus to a 400 MHz system bus will make Rambus's high-speed memory all the more important for that chip set. At its developer's forum, Intel positioned that chip set as the top choice for PCs used in high-bandwidth homes.
Noteworthy, too, is the opportunity that Advanced Micro Devices (NYSE: AMD) (AMD) has to push PC systems running its high-speed chips between now and the time Willamette comes to market. AMD, also a licensee of Rambus's technology, isn't yet making RDRAM chip sets and may do well to wait for RDRAMs to become more ubiquitous before attempting such a costly change in memory technologies. "Why risk using Rambus when things are going so smoothly for [AMD] right now?" asks In-Stat's Mr. Cullen.
If Rambus continues to ascend, it doesn't mean that SDRAMs and DDRs will cease to exist. "It's a revolutionary technology, and there's a long learning curve involved," Ms. Garber says. "We believe Rambus has a place in the market, but that SDRAM will be the mainstay for a long time."
However, Rambus's support from so many large memory manufacturers could mean that SDRAMs become a boutique product, making them slightly more expensive than RDRAMs in a few years. Once RDRAM prices fall and volumes rise, manufacturers may start using them for things that don't require much power, such as telephone answering machines, Dataquest's Jim Handy says.
Analysts agree that it will be a couple of years, at the very least, before RDRAMs begin to penetrate the low-end PC market or get cheap enough to be used in commodity consumer electronics.
Long-term, Rambus plans to apply its technology to other markets, such as trying to solve chip-to-memory bandwidth issues in the networking equipment space. Considering that 10-year-old Rambus took this long to get its RDRAM technology to market, don't look for those new products anytime soon.