As Delays Pile Up, Rambus Sweats Out the Memory Race

Intel stands pat on support of Rambus' revolutionary memory booster, but DRAM makers aren't holding their breath.
Author:
Publish date:

SAN FRANCISCO -- Three years ago, Rambus (RMBS) - Get Report promised to revolutionize the PC industry with the equivalent of ginkgo biloba for computers -- a chip design that would boost memory performance.

Rambus backer

Intel

(INTC) - Get Report

saw this as a cure to a vexing problem: Sluggish memory was slowing down the performance of its ever-faster chips, giving consumers a reason to pocket money they would otherwise spend on computer upgrades.

No wonder Rambus was the hottest IPO of 1997. Going public in May of that year, the stock shot up 150% to close at 30 1/4 on its first day of trading. Investors flocked to the new issue because they loved its business formula: Lots of royalty income without manufacturing costs meant high profits. Compare that to memory giant

Micron

(MU) - Get Report

, which spent $2.7 billion in fiscal 1998 to bring in revenue of $3 billion.

A Rambus-based dynamic random access memory chip, or DRAM, was supposed to be out by this summer. But now there is doubt whether we'll see the chip this year because of continuing delays. And memory makers, resentful of having to pay Rambus royalties, are racing to market with royalty-free alternative technologies, such as PC-133 and double data rate DRAM, or DDR.

If PC-133 makes it to market first, memory makers could put off buying Rambus-designed chips until they bring out with their own DDR chips that match Rambus-level speed, says Derek Nuhn, of technology-assessment firm

Semiconductor Insights

. The closer computer makers look at PC-133, the more pressure is placed on Intel to do the same, since a stampede to the alternative could cost the chipmaker some $400 million a quarter in sales, one Intel analyst says.

But it's not clear how much headway either Rambus or its alternatives are making. Rumors abound of Rambus bugs and secret plans at Intel for alternatives to the Rambus chip.

"I am being bombarded every day with wacky rumors having to do with making mountains out of molehills," says

SG Cowen

chip analyst Drew Peck, whose company has no underwriting relationship with Rambus. These rumors are difficult to confirm because nondisclosure agreements keep many industry consultants and others who work with Rambus from talking. And many stock analysts who track Rambus closely say they are leery of angering Intel.

"It's pretty clear now that Intel is willing to do an end run around Rambus and go to PC-133," says one oft-quoted Intel analyst, who asked this time that his name not be used.

Intel and Rambus, for their part, say the Rambus chip will be out by the third quarter, period. "There hasn't been any change in schedule since our announcement in February," says Intel spokesman Robert Manetta.

Inside the tame memory industry, this is steamy stuff. Behind all of it is big money: Some $13 billion in memory chips were sold last year, and that number is expected to grow to $45.5 billion in four years, according to

Cahners In-Stat Group

.

If Rambus meets In-Stat's predictions, memory makers will have to pay Rambus some $400 million a year in annual royalty fees by 2002. Meanwhile, Intel, which is Rambus' sole endorser, pays Rambus little if anything for its licenses.

Stanford

professor Mark Horowitz and

MIPS Technologies

(MIPS)

chief scientist P. Michael Farmwald founded Rambus in 1990, but it took five years before they got it into Nintendo and other game machines. It took another two years before they got Intel's attention for PC chips, in part by giving Intel options to buy 1 million Rambus shares at $10 each.

But the memory market was in a slump, and Rambus stagnated until June of last year, when

Compaq

(CPQ)

said it would license Rambus' designs for its servers. Around the same time, In-Stat predicted Rambus would be in more than half of all computers by 2002.

The company got another boost later on when Intel announced it would invest a total of $600 million in memory makers Micron and

Samsung Electronics

to help pay for costly upgrades needed for Rambus production. Rambus' stock also got a boost from a broader rally in the chip industry, peaking on Jan. 12 of this year at 109 1/2 -- nine times its offer price.

At that point, Rambus seemed like a sure bet. Memory makers were lining up to buy a Rambus license. The alternative -- synchronous link DRAM, a hot rage a year ago that was pushed by a consortium of some 24 chip and component makers -- had all but disappeared.

Then Rambus' own underwriter,

Morgan Stanley Dean Witter

, warned Jan. 12 that the stock was fully valued and downgraded it to neutral. A month later, Intel warned that its Rambus chips would be delayed by three months, pushing the rollout back to September.

"As far as we are concerned, there is no change to that," says Rambus CFO Gary Harmon. "We are far along. Samsung and a couple of others are ready to go now."

The stock, meanwhile, is 32% off its peak. Many technology fund managers say they won't touch Rambus because they see it as too risky, an all-or-nothing stock that will either be hugely successful or will crash and burn.

Many still say that Rambus proliferation is just a matter of time and that the rumors of problems with Rambus should be dismissed. "There are huge short positions on the Street," says Peck. "Are there still design issues? Absolutely. But you'd have to be nuts to be buying Rambus for what it can earn this year or next year."

Others aren't so sure. "What is dead is the 100% certainty that Rambus will win," says Danny Lam, an analyst with semiconductor consultant

Fisher-Holstein

. "If Rambus had come on board a year ago, it would have been a shoo-in."