February 23, 1998

Craig Barrett Interview

BARRETT: We have interest. Our interest is uniform across the board. You want to build a box and put an operating system on it. We would love you to do it with an Intel processor.

INFORMATIONWEEK: I guess you can use all sorts of competing chips with that ...

BARRETT: It's just like NT servers -- you can use all sorts of competing chips.

INFORMATIONWEEK: Right, right. Although I guess with that there are a lot more viable competing chips than there are ...

BARRETT: Mr. [Robert] Palmer [of Digital, maker of the competing Alpha chip] wouldn't appreciate that comment at all.

INFORMATIONWEEK: Would you do anything with Strong Arm if that goes through?

BARRETT: Our current philosophy with Strong A rm is if the Digital [Equipment] deal passes the government's scrutiny, which I think it will, we will take that product line and push it forward into the spaces where it makes sense.

INFORMATIONWEEK: I wanted to ask you about Alpha vs. Merced. I have talked to some people at Digital who are, naturally, very bullish still on Alpha. The people at Digital claim that Alpha, around the time Merced arrives, should still have twice the performance of Merced. Analysts I talked to say anywhere from equal to twice or anywhere in between. So, with that in mind, and Compaq controlling, say, 50% of the server market [if the Digital acquisition is approved] ...

BARRETT: Both Compaq and Digital are very good customers of Intel and we expect them to be very good customers in the future. Both of them have announced their intention to go with Merced. You will have to go talk to Eckhard [Pfeiffer, CEO of Compaq] or somebody else and see what their philosophy is. But, Compaq has been o ur biggest customer and has pushed Intel architecture very hard in to the server space.

INFORMATIONWEEK: They mention the manufacturing process, once it gets downs to .18, they said that the Alpha would have about a one-eighth inch-sized dye, whereas, Merced would still be fairly large, and they claim that on a cost basis it [will be] much more competitive, as well.

BARRETT: [Others make similar claims. Look at] this thing called the PowerPC. There are certain things that you should take with a grain of salt. One is that our industry seems to have a penchant recently for technology announcements. It is all very interesting. But what really counts is being able to produce things at high volume and high performance at a reasonable value. We probably don't engage in much of those speculative technology announcements. Even though the whole world has been working on copper and Sematech has been working on copper for a bunch of years.

There is the whole notion of techno logy advancement and managing the expectations of what technology can do for you. We talk about the mechanics of these things and how difficult it is, and we all play the games about pre- announcement and how pre-announced are you. But, there is old Moore's Law. My point is that if you were to look at anybody's road map -- whether it be copper or gigahertz -- all those things will capture the imagination of the press and the reader. If you look at just about anybody's roadmap and the timeframe that all these products are out, everybody has a gigahertz processor. I find it a little bit interesting that it generates so much excitement.

INFORMATIONWEEK: We did want to talk about low-end vs. high-end desk tops. Our readers are grappling with measurements, return on technology investments. As you enter more areas, as you form more alliances, as technology power increases, what are you hearing from customers about what that they need to measure return on investments?

BARRETT: It is an interesting question which we could probably talk for hours on. The overview to this is the smart people have been publishing for years that with all this investment with computer technology, there has really been no effective increase in productivity, therefore no effective return on investment. The only problem is that all the companies who have not invested are out of business now. That is the part that never gets said.

The flip side is if you look at big corporations, they tend to fall into two categories. There are those who treat their IT investment as some form of competitive advantage. And there are those people who treat it as the cost of doing business. It kind of depends on which class you are in and who you are talking to. Why should I upgrade from my 486- based PCs or these servers or this enterprise structure that I have, because it will just cost me more money, as opposed to the key to the future is being able to handle data and make database decisions and comm unicate instantaneously? I am going to use this as a competitive advantage and therefore, like some of the Wall Street traders, if I can make this calculation a millionth of a second faster than the other guy, I have a technical advantage, I have a financial competitive advantage.

If I make the wrong strategic decision on product direction and I have the greatest IT infrastructure in the world, is that a good ROI or a bad ROI? If I happen to be in a product line which is being gobbled up by somebody else, is that an influence of my past market position or my future investment in IT? It is very difficult to answer that, and that is why people fall in these two categories. One class says I am totally interested in total cost of ownership and I am going to minimize my total cost of ownership and manage it in that fashion. The other one says by leap of faith I am assuming that I will use this information technology as a strategic weapon. I will have better access to data -- be able to commun icate better and make faster decisions than my competition.

INFORMATIONWEEK: Do you lean one way or the other?

BARRETT: Well, I think that they all ought to use it to their strategic advantage. But, if you look at the Intel strategic objectives, basically it says to use IT as a strategic competitive weapon. Strategic advantage doesn't mean to ring every last nickel out of it, but to give people the tools on their desk to be as productive as they possibly can. That is first and foremost. And, then once having done that they do it at the lowest cost possible.

INFORMATIONWEEK: Can you talk a little about the longevity of cheap desktops vs. high end, adopting somewhat of an ROI prospective?

BARRETT: We roll over our desktops every three years in this company.

INFORMATIONWEEK: I wanted to look at that in the context of the sub-$1,000 PC market. The argument often is that you are going to have to retire them much more frequently if they ar e sub-$1,000 PCs vs. $2,000 machines.

BARRETT: Probably a more interesting discussion is how to look at it in terms of whether you spend $1,100 for this piece of hardware, or $1,500 or $1,900. How big a contribution to the total cost of ownership and productivity is that extra $800 in capital up front? I don't know what your total cost of ownership of a desktop is. At Intel it is currently $6,000 to $7,000 a year. That is what you pay for the hardware, the software, the network infrastructure per desktop. Over a three-year period that is $20,000. The question is then, of that $20,000, $1,000 of it is the delta between a $1,000 client and a $2,000 client. So, it is a relatively small portion.

INFORMATIONWEEK: What do you see as the biggest MIPS-munching corporate applications in the future?

BARRETT: I would guess that it would be visual stuff and visual communication -- a combination of videoconferencing and training and the ability to broadcast one-to-many.

INFORMATIONWEEK: Multicast IP stuff?

BARRETT: Yes, that kind of stuff. Rich Internet and intranet applications. Some people argue that you should apply it to the lowest common denominator, and other people argue that you ought [to go] with the richest stuff you can. We are obviously on the rich side of that thinking. We think that the people who are on the other side of that argument are [similar to the one who might have said], Why would you upgrade from a black-and-white TV to a color TV if most of people own black-and-white TVs? I never quite understood that argument, but, that is what it seems to sound like to me.

INFORMATIONWEEK: Many of the PC manufacturers that you work with are moving to just-in-time manufacturing models and more direct-selling models. They have less stock around. How does that affect your manufacturing model? Do you have electronic links with many OEMs? How does that work?

BARRETT: Well, we increasingly go into basic electronic commerce with our major customers. The really wonderful aspect of it is we basically have a fill-to-order system at Intel which we like to match up with our customers. The wonderful thing is that as soon as you start to decrease the size and width of the channels, be it either a work in process or in the channel inventory, you allow new technology to flow into the system faster. Taking it to the extreme, Intel introduces a new processor today and, if there is six months of flow into the channel, it takes six months to flush that out before that new technology gets to the user. If there is two weeks of float, we introduce something new, then it is two weeks to flush everything out. Being able to get the new technology into the marketplace rapidly increases the efficiency of the channel and the efficiency of the whole system and that has to benefit everybody.

INFORMATIONWEEK: It is definitely good for you and for PC manufacturers. It makes the rapid introduction of tec hnology more of a reality for the IT manager. It means that the push to innovate is stronger and these people have to learn to change more quickly and to assume and manage more technology. As great as that is in terms of margins and all that kind of stuff, all that technology still has to be consumed.

BARRETT: We understand that. You know, we buy a lot of this stuff ourselves and our VP of IT has the same issues as every other Fortune 500, Fortune 5000 company. You basically buy to standards and update your standards. You go to a single architecture because that makes your life simple. But, you still get the new technology into the marketplace and satisfy the end user. The end user happens to be the person who has the keyboard and the mouse and wants a rapid response time. So, we just push internally too.

It used to be the stock answer was, "This is our standard for the next two years, I am not changing." And now, because models in this industry only last for six or nine month s, you can't say this is a standard for two years. You have to say, "OK every six months I am going to reset, but, I have to reset it in a standard way with hardware, software manageability, low cost of ownership, and backward compatibility. So that what I have got today, will continue to play and I will just upgrade it."

INFORMATIONWEEK: We have a couple of questions on Merced performance.

BARRETT: We obviously believe that Merced is going to be an industry standard. If you just look at the number of OEMs who are cognizant of the performance characteristics, of people who have signed up for Merced already, including Digital, Compaq, and a number of other people. I mean, you look across the board at all of the current RISC processors, suppliers, including Hewlett-Packard. I would just rest on their interpretation in the fact they are all signed up with us to go. They are not signing up to go with a second-rate processor family. Now, will it get better as it goes on in time? Absolutely.

INFORMATIONWEEK: What do you think will be Merced's biggest rival?

BARRETT: There will be the standard RISC guys that we have gone against in the 32-bit world.

INFORMATIONWEEK: If you could just single out one, who might that be?

BARRETT: That would be other folks who are in the workstation server business: Sun, Sparc, Alpha.


Return to "Intel's Enterprise Push"

Send Us Your Feedback

Top of the Page