MOUNTAIN VIEW, Calif. --
stock has zoomed up and down on perceptions of whether
will support the revolutionary Rambus-based chip-memory designs.
In the front seat of this roller coaster is Geoffrey Tate, Rambus' president and chief executive for the past nine years, a guy who both looks and sounds a bit like
and projects the same calm credibility. While a temperate attitude is expected in an industry as dull as computer memory, Rambus is at the heart of one of the great
controversies in the chip industry. The company's fortunes depend on a major industrywide technology change, one the memory industry doesn't want to make.
Animosity among tiny Rambus and its giant
partners is so strong that Rambus CFO Gary Harmon and
CEO Steve Appleton almost came to
blows in front of a crowd of investment managers at a party this past February.
Why such venom? To adopt Rambus' designs, memory makers have to invest huge amounts of money for new equipment, give up some control over their manufacturing, and, for the first time ever, pay royalties to the tune of 1.5% of the price of every chip sold. Instead, they are trying to push through an alternative standard called "double-data rate," faster than today's memory but not as fast as Rambus'.
Investors have placed a high
value on Rambus' stock on the promise that the standard will dominate the memory market within the next three years. On July 16, Rambus closed at an all-time high of 113, giving it a market cap of $2.7 billion, even though it produced just $2 million in income on $10 million in total revenue in the previous quarter. The stock has since dropped 25% because of Intel's July 19 announcement that it would evaluate PC133 SDRAMs.
Rambus' moment of truth is approaching. At its fall
Intel Developers' Forum
, or IDF, which begins Aug. 31 in Palm Springs, Intel is expected to announce both a long-awaited September rollout of its Rambus-based Camino chipset and support of PC133 SDRAMs, seen by some as a stopgap measure until royalty-free, double-data-rate DRAMS can be developed. But many people suspect that at the IDF, Intel will signal a move either toward, or away, from Rambus. For Rambus, the choice could mean fantastic success or a much-diminished future.
TSC: Why does Rambus inspire such emotion?
: What we are doing is pretty audacious and technically very challenging. In the early days, people thought that our technology just wouldn't work, that we had to be out of our minds or lying to them. It was just so outrageous that we could get so much more performance.
But it got a lot more emotional for some people because of Intel. Intel is one of those companies that generates emotion and we get a lot of emotion by association. If people are disposed to like or dislike Intel, it seems to rub off on us because we are coupled closely with Intel. But half of all DRAM bits couple up to Intel microprocessors or controllers, so if that is the price we pay for having Intel as a good partner we are willing to pay that price.
TSC: Could you have gotten where you are without Intel?
: It's a series of stepping stones. The
success is what got us Intel. Intel was looking for more memory bandwidth for the much higher processors and graphics platforms. But they wanted to see it proven. Nintendo gave us that.
TSC: Investors love your "chipless" model -- you don't make chips but license the design. This way you have high royalties and low cost, a formula for a cash cow. But it makes you dependent on DRAM companies to not just make the product but sell it as well. Does that make you vulnerable?
: Control is a two-edged sword. Selling product has advantages and disadvantages. You have to put in a lot of investment in infrastructure to create a worldwide sales force and logistics channels to carry inventory to sell that product. And inventory is a risky thing to carry. It loses its value quickly.
TSC: Did you anticipate the magnitude of the challenge?
: No. The technical challenge turned out to be bigger than we thought. Maybe we have 1% of the share of the DRAM market today after nine years. And the business challenge was really hard, getting these partnerships, getting people to work in the same direction at the same time and disseminating our know-how to all these companies.
TSC: One memory maker told us the industry resents that you dictate how to build their products.
: A phone or computer that is almost compatible is one that doesn't work. If people build parts 99% compatible, the systems companies won't buy them.
: There is a big difference. Rambus is a technology for communicating between chips. The
Playstation 2 uses MIPS, and Sony doesn't have to coordinate with anybody other than the MIPS semiconductor supplier. There is an extra degree of company-to-company coordination involved in our business model. Everyone wants multiple-sourced DRAMs, so to make
happy, you need multiple suppliers of DRAMs, modules, connectors and clock chips. That is at least a dozen companies that we have to orchestrate.
TSC: That's not always necessary, is it?
: Nintendo went into production with one semiconductor licensee,
. It gets complicated with companies that want multiple sourcing, which is the PC business.
TSC: At the Intel Developers' Forum people hope to hear whether Intel is moving closer toward or further away from Rambus. Are you nervous?
: We are used to it. We see two camps going forward. Camp A thinks the PC is as fast as it needs to be. And camp B says, "Five years ago I couldn't see what would be different but all these nifty, whizzy things came out like the Internet. In five years we will get all sorts of things we can't imagine today but we can't live without tomorrow." We are in camp B. Not just because it is self-serving, but because we think it is the best guess. Performance is going to continue to go up. That's a really critical assumption.
Rambus will come in first in performance desktops but will, over time, move to be the main memory of choice in all PCs. It can take longer or it can take less time. I'll be the first to say that if someone had a crystal ball to show that we would never need anything faster than today's software, then the transition to Rambus would be very slow. If you are a camp A investor, you might want to go out and short the stock.
TSC: Many have.
(In its Aug. 23 issue, Forbes advises investors to short Rambus.)
: Not just our stock. Look at Intel.
TSC: We've heard that double-data-rate DRAMs may be based in part on Rambus technology. If, under a worst case scenario, the industry adopts DDR and abandons Rambus, do you have patent rights on it that hedge your bets?
: We've made no comment on whether DDR infringes our patents. We frankly have had a hard time getting DDR samples. Our position is there is insufficient data. But every Rambus DRAM is basically a double-data-rate part. We transfer data on both edges of the clock and we were the first to do so.
TSC: Say Rambus becomes the standard. Do you guys sit back and count the royalties?
: Even we don't think that will happen. We have a lot of work to do to reduce costs. What does any company do that achieves goals. We set new goals. We have to find ways to expand our business. There are lots of ways we think we can leverage our business model and business relationships. We have a lot of good ideas but we haven't talked about them because we don't want people to think we are taking our eyes off the ball.