SMARTMONEY.COM: A Scary Time To Be Rambus
By Tiernan Ray
Smartmoney.com
NEW YORK (Dow Jones)--Rambus (RMBS) may well be the longest-running bubble stock of all time.
The company had only $37.8 million in sales last year and expects flat revenues for the rest of this year.
However, since it went public in May of 1997, the company, which designs advanced semiconductors
and leaves the manufacturing to others, has grown fat on sheer promise.
At the end of 1996, Rambus had the good fortune to receive Intel's (INTC) blessing as being the
provider of the next generation of memory technology for desktop PCs. The stock has soared 260%
since its IPO, and never looked back. Though it has fallen from its high of nearly 110 in January, at a
recent 63, it trades at an impressive 217 times this year's projected earnings of 29 cents a share.
That was until last week, when news started to dribble out about a poorly kept industry secret, namely
that Intel has prepared an alternative processor to the computer's memory circuits. This chipset works
not with Rambus' technology, called Rambus DRAM (RDRAM), but with a competing technology
dubbed synchronous DRAM (SDRAM). Rambus stock started to lose ground Thursday afternoon and
was down by as much as 13% at one point on Friday, thanks to speculation that after years of struggle
it is about to be given the boot by Intel.
Is it all over for Rambus? Not necessarily. What lifted the stock to its January high is the backing of
people like Morgan Stanley's Mark Edelstone, who values the company like an Internet stock.
Edelstone's looking at the company to make $8 in earnings per share in 2003, which would give Rambus
a premium of only 7.8 times forward earnings - a steal, in other words.
Insiders were not surprised at the latest drama about Rambus. Indeed, it is the second time in the last
two weeks that Intel appears to have hedged its position on Rambus; Intel VP Paul Otellini gave the
anointed technology only the mildest of boosts during the company's quarterly analyst briefing on April
13.
The real question is not whether Rambus will have to share the spotlight initially with competing
technologies. Rambus has never been on time with its promises, and it has always been perceived as a
long-term value story. Nor is it a surprise that Intel has had a backup plan ready to try and smooth the
transition to Rambus this year.
No, the issue is whether RDRAM will end up being a technology that's simply too expensive for the
market. Edelstone's financial model is built on certain assumptions about how fast RDRAM can take
over the market for PC memory. Specifically, he believes Rambus will have garnered 30% of the PC
main memory market by the end of next year, then 45% by the end of 2001 and 75% by the end of
2003, giving the company about 50% of the total DRAM market. It's only by hitting these targets that
Rambus can collect sufficient royalty revenue.
But here's the Catch-22: In order to spread as quickly as Edelstone hopes, RDRAM must rapidly come
down in price. When it rolls off the lines, RDRAM will have a price that's at a premium to PC 133 and
other SDRAM technology.
Estimates of that premium are simply all over the map, from 30% more than conventional SDRAM to
over 100%.
Over time, the price premium should decline, allowing the technology to spread from its initial rollout on
$2,000 desktops to even sub-$500 desktops. But in order to come down in price, RDRAM has to first
be purchased in enough volume so that Rambus" par tners, the memory-chip manufacturers, can bring
down the costs they incur to make the part.
In a Catch-22, something's got to give, and it may come down to how much money DRAM vendors are
willing to lose making RDRAM. NEC, which has a long relationship with Rambus extending back to the
Nintendo 64, is hardly negative on RDRAM. But even NEC is worried about having to cut its margins
to make RDRAM affordable. "We're trying to make it clear [to computer manufacturers] that we
cannot afford to lose money on this part," says Will Mulhern, NEC's DRAM marketing manager. Jim
Sogas, director of Hitachi Semiconductor's DRAM business unit, puts it thusly: "We, meaning all the
DRAM vendors, are disinterested parties. But if we're expected to produce a part that costs 30%, 40%,
50% more [to make], and then sell it for the same price [as cheaper products], then we won't do it."
But won't they? Part of the magic of Rambus' strategy is that the company has divided the DRAM
vendors against one another. After three years of declining DRAM prices the makers of memory chips
are finally hoping that 1999 will see a boom in demand and an end to profitless misery. Being the
volume leader in RDRAM - the proposed future of DRAM - plays a pivotal role. George Iwanyc,
DRAM guru with Dataquest, foresees a 50% growth in the DRAM market in 1999. "As demand
returns this year, there's going to be shortages in DRAM. Anyone with leading-edge capacity, and that
includes in RDRAM, will do be tter in the recovery." NEC, Toshiba and Samsung are believed to be the
leaders in RDRAM, while others such as Hyundai, Hitachi and Micron Technology (MU) are relatively
late to the project.
Intel is not taking any chances: It has rushed to prop up the DRAM folks with debt and equity
investments. But the need for something like Rambus is evident: Today's top-of-the-line Pentium III
processors running at speeds of 500 MHz consume tens of gigabytes per second internally. As the
Pentium speeds up to one gigahertz by late next year, the demand for a really fast memory chip to feed
that appetite for data will be even greater.
However, there may be other ways to solve that problem, and the Rambus bears maintain that the
premium on RDRAM puts the product out of step with a computer market fragmenting into different
kinds of devices.
According to Fisher Holstein's Danny Lam, a chip consultant, "Rambus is increasingly a product that is
late to market. In the time it has been delayed, conditions that wer e very favorable to it at first have
changed. It now looks decidedly passe, given the erosion in the PC business." Still, others believe
RDRAM simply cannot fail because it's the only long-term solution. "Intel's behind this; there's a ton of
capital invested; and the alternatives [to RDRAM] just can't cut it," says Michael Slater, head of
MicroDesign Resources.
"These are interim solutions," Morgan Stanley's Edelstone says of PC133, the next revision of the
existing SDRAM standard. "Intel's thinking four or five years down the road" about what kind of
memory chips are needed.
And he's right: The two main alternatives, PC133 and Double Data Rate (DDR), offer bandwidth
comparable to Rambus, but the movement to take them to higher speeds over time really only got
underway six months ago, when an industry-standards body called JEDEC decided to look at ways to
replace Rambus. A proposed follow-on to DDR, called DDR2, is still only a vague possibility.
These technologies have their fans, though.
Their hopes for an alternative to Rambus may hang on a small Taiwanese company called Via
Technologies, which recently went public on the Taiwan exchange. Via sells chipsets that compete with
Intel. When Intel moved to a different kind of housing for its chip last year, called Slot 1, Via increased
its business by selling a cheaper standard package, called Socket 7, to computer makers that didn't want
to go to the expense of converting. Via had $181 million in revenue in 1988, up from $129 million th e
previous year. The company is hoping Intel's stubbornness on Rambus up to this point will present a
similar opportunity this year.
"It's like [Intel's] doing it all over again with Rambus and Camino [the RDRAM chipset]," says Dean
Hays, head of marketing for Via.
"Some OEMs [original-equipment manufacturers] have been asking questions, and I think Intel knows
they have a massive problem on their hands," says Hays.
Frankly, Via and others had better pray for such a disaster, because it's clear that PC133 and DDR are
not going anywhere in the long term without Intel. The company was a major force in driving the
development of the last generation of SDRAM (PC100), so it's not enough for Intel to simply refrain
from opposing these efforts; it has to help.
Hans Mosesmann, who follows Intel for Prudential Securities, tells me that he thinks Intel may be
prepping a more extensive involvement with SDRAM based on its most recent rumblings. "I think later
this year you'll start to hear Intel talking about an additional [memory] roadmap, one that doesn't include
RDRAM," but is based on SDRAM, says Mosesmann.
"They've opened the door to alternative solutions." If Mosesmann is right, it's a scary time to be
Rambus. If indeed PC133 and DDR gain some foothold because of these latest slips, and if Intel gets
behind the technologies, it could add fuel to the anti-Rambus campaign. And if that happens to Rambus,
well, the question for the history books may be how one tiny company managed to generate a kind of
latent mistrust and resentment, among press, among analysts and among some DRAM manufacturers,
that is customarily reserved for big bullies like Microsoft.
For more information and analysis of companies and mutual funds, visit SmartMoney.com at
smartmoney.com. |
