Undoubtedly one of the world's most powerful businessmen, Craig Barrett, president and CEO of the Intel corporation, gave a presentation of his company's vision for the future at a conference in London yesterday. Having inherited the hot seat from the legend that is Andy Grove, Barrett is to the computing industry what George Soros is to international finance. This man is hot. You know he's hot by the sheer size of the engine around him; photographers, PR types, camera crews, assistants.
What Barrett says affects world markets. Decisions he makes affect jobs and increasingly, his beliefs and opinions have influence on our governments. Craig Barrett's products have a brand recognition that challenges the likes of Pepsi and Levi's and a jingle that's as familiar as "Always Coca-Cola".
But unlike the consumer giants that reap the rewards of continuous demand, Intel has to create its own markets--a process which has landed it in court under charges that almost duplicate those faced by its close partner and ally--Microsoft.
Barrett took half an hour of out of his hectic London schedule to talk to the editor of ZDNet News, Richard Barry. There were no rules, the only restriction was time.
Richard Barry: We took the opportunity to explore our readers interests on the News Channel earlier this week by asking them to email some of their own questions. And there were a lot of questions.
Craig Barrett: Ha. (Laughter) That's a novel approach.
Richard Barry: I think so. Hopefully so do our readers.
Craig Barrett: Well why don't you send me the questions and I'll answer via email.
Richard Barry: (More laughter) That would be cheating and I don't think my boss would approve. By far the most popular topic was Celeron. One of the questions we received was whether Celeron (Mendocino) is as capable as Pentium II and if so, why do we need Pentium II?
Craig Barrett: If you look at the performance characteristics for the Celeron family it is aimed at the entry level. Pentium II family is aimed at what I call the mainstream performance desktop. There is a slight overlap in performance at the top-end of the Celeron family. You could probably look at analogies in the automobile industry. We have different brands in automobiles and there is overlap in price performance but as we move forward into the future that overlap will be minimum between the two as they both move up in performance characteristics. And you'll also have at the top of the line the Pentium II Xeon processor family with higher performance for workstations.
Richard Barry: So where we are now? Today, there is clearly an overlap.
Craig Barrett: There's a very small overlap in performance. I mean if you look at today in the mainstream, Pentium is going to 350, 400MHz, 450MHz. Celeron is down in the 300MHz range. It's minor overlap, very minor overlap. There'll be some grey space forever as these brands move forward.
Richard Barry: The message from our readers is that they are confused by Celeron. Confused at the initial launch of Celeron. A question we are constantly asked is why did you launch the Covington Celeron followed by the hasty launch of Mendocino when the message from the experts was why didn't you wait for Mendocino to be ready instead of releasing what is widely regarded as a failure?
Craig Barrett: Well the experts always have opinions! (both laugh) Ha, they always do. The experts had opinions on the 386SX, the 486SX, Pentiums. Basically what we do is sell products to people and not to press experts.
Richard Barry: (Interrupts) But Craig we're not just talking about press experts here. Organisations like the Microprocessor Report for example, organisations, people with unquestioned expertise. They believe that Intel reacted, there was a gap in the market that was being filled by AMD and Cyrix at the low end. Intel was seen as lacking, not only by the experts but by Wall Street, the company had to be seen as doing something in that space. It had to address that market and therefor released a product before it was ready. Isn't that what happened?
Craig Barrett: We released a product, in fact if you look at the multimedia performance of the Celeron without cache (Covington) and compared it to the K6, which one has better performance?
Richard Barry: On multimedia only, the Intel product.
Craig Barrett: Then why are we critiquing a product that has better performance in the same price space as the K6?
Richard Barry: Maybe we're missing the point. When the Celeron was released it was universally slated and harassed by the press and the chip experts. Very poor performance, certainly not the performance we were expecting from the Intel stable.
Craig Barrett: No, and I'm not playing words with you but let me take you back through history. I wasn't kidding when I said the 386SX when it was introduced was harassed by the press as was the 486SX but they both did exceptionally well in the market place. We introduced the first version of the Celeron in a rapid fashion. It didn't have integrated cache...
Richard Barry: (Interrupts) Why did you introduce it rapidly?
Craig Barrett: Because the $1,000 market -- popularity came on very fast in the marketplace.
Richard Barry: Did it take you by surprise?
Craig Barrett: Of course, we've acknowledged, admitted that, so we responded rapidly with a product we thought had good characteristics and committed to a brand, sold our brand and follow on products behind that and that's exactly what you are seeing. The current version with integrated cache is a very good product. Compared to anybody else in that price range at that performance, it does super.
Richard Barry: Celeron undoubtedly had a poor start, [for] people who spent $1000 on a box...
Craig Barrett: (Interrupts) They got a great product.
Richard Barry: Despite the benchmarks, the negative press, you still believe they got a great product?
Craig Barrett: I believe. I rhetorically asked you the question, what's better: a K6 or a Celeron -- the original version of Celeron. Mendocino Celeron? That's a code name; forget about it. What's better a K6 or a Celeron in terms of multimedia performance, which is probably the most taxing performance for processors that consumers are going to use?
Richard Barry: I'm glad you've brought up multimedia, it leads me nicely into Katmai. The industry is certainly abuzz with rumour and speculation about it. But AMD is already there -- "It's here now" to quote a common Intel retort -- with 3DNOW! its been here for what, six months, is gaining support and we're not expecting Katmai for what, another six months. Right?
Craig Barrett: Uh huh.
Richard Barry: That gives AMD a six month lead on you already and by the time Katmai arrives, and correct me if I'm wrong, it will probably launch for around, say $800, that's the usual. If I'm correct that's going to place the Katmai at the high end to begin with, giving AMD more advantage in that arena.
(This question was prompted by a conversation with AMD's North European marketing manager, Richard Baker, who said he was "delighted" that Intel would have to launch Katmai at the top end of the market, while AMD settles back to enjoy maximum output from its fabs increased and unchallenged brand recognition and, importantly, much reduced prices.)
Craig Barrett: Y'know what? I wouldn't hire you as a marketing man. (Laughs)
Richard Barry: Why thank you sir, I'll take that as a compliment!
Craig Barrett: Ha! No, let me give you an alternative marketing scenario. If you want to introduce a new instruction set into the market place you don't simply throw it into the market and say 'Great, I've got 3DNOW!' Or whatever. Katmai has 70 new instructions, the only way the end user sees some benefit from Katmai, or 3DNOW!, or MMX technology is by working with the software industry two years in advance to create software titles that take advantage of those instructions.
Launching a processor and saying 'Isn't this wonderful?' brings the end user no benefit at all. So what Intel does is it says fine, we've got 70 new instructions, it's gonna mean better animation and richer video etc. Let's work with the software industry such that we have a coupled introduction of applications and hardware.
Your concept that it's gonna cost $800...
Richard Barry: (Interrupts) Is that incorrect?
Craig Barrett: Ha! I'm not gonna tell you what the price is. Katmai instruction set is going to go right through the entire processor line very quickly. So trying to segment, target it out at the high end most expensive processor, is a ridiculous marketing position.
Richard Barry: Agreed, so will we see this targeted, say more closely to the gamer?
Craig Barrett: You can expect us to push Katmai very aggressively across the entire processor line very quickly. This is significant because Intel has rarely launched a processor below $900. A break in what has become something of an Intel tradition illustrates the strength AMD is currently languishing in relatively unchallenged.
(Could it be that Intel will launch Katmai at similar price points to the -- by then -- established 3DNOW! Family? A follow up on this point with analyst and OEM comment will follow.)
Richard Barry: In the interests of time I'll move on... there's a lot to cover. Intel's capacity, its fab capacity. There have been reports that you delayed the planned fab in Texas, [and] yesterday you announced over 600 job cuts.
Craig Barrett: That was nothing to do with wafer fab.
Richard Barry: And obviously there was the recent announcement that 3,000 jobs were being lost, and reportedly you have shut down fabs for weeks due to over-production. Can you confirm or deny that you have had to shut down fabs for weeks? According to Michael Slater, senior editor at the Microprocessor Report, Intel over-shot its estimate for demand and had no alternative but to close down the fabs -- casting doubt on the company's ability to predict demand as well as inflicting unnecessary cost.
Craig Barrett: We shut down one or two of our fabs in Oregon for a short period of time. But, well. OK, can I paint the whole picture for you?
Richard Barry: For sure. Please do.
Craig Barrett: The truth is we are spending $4bn on manufacturing capacity OK. The whole picture is that we delayed a fab in Texas for a couple of reasons. One is that PC demand is slow this year, slower growth than historically, so demand is down. We decided to delay that factory in Texas to bring it up as a 300mm factory on .13 micron technology because PC demand was slow. Yeah we shut down a factory or two for a period of time to keep inventories in line. I think it was a prudent thing to do. We four or five months ago announced that we were going to shed 3000 jobs this year, that was after we had created 35,000 jobs in the previous three years. So put it in perspective -- that's all I want you to do. You know we haven't just gone in and shut down our factories like some of our competitors have done in and around the local geography and we're not bleeding red ink like just about everyone else in the semiconductor industry. We're still investing, I said $4.5bn in manufacturing capacity this year.
Richard Barry: Craig, what you say has a direct affect on the whole industry and I promise the next question isn't loaded! (laughs)
Craig Barrett: (Laughs) Hey I'm just getting conditioned here.
Richard Barry: The industry is in a terrible state at the moment.
Craig Barrett: It is.
Richard Barry: Do you think things are going to pick up and if so, when?
Craig Barrett: You know it's tough to tell. Two things happened this year that really hurt the component industry, the suppliers in this industry. What it is, there was a major inventory glut with a lot of the major OEMs first half of the year. It took six to nine months to work that out. Second, the growth rate this year was lower than it is historically. People had put a lot of capacity in place assuming that everything would be happy and healthy so the over-capacity situation compounded the problem which is why people have been bleeding red ink. That's the bad news.
Click here for part II of ZDNet UK's exclusive interview with Intel CEO Craig Barrett.