Part 2, Barrett unplugged: 'You can't keep a proprietary vertical stack'

"110"="" hspace="3" vspace="1">In Part 2 of his interview with Tech Update, Intel CEO Craig Barrett talks about the battle for supremacy in data center architectures, why the road to innovation lies in modular building blocks with open interfaces, and the "interesting strategic choices" facing Sun CEO Scott McNealy.

Tech Update: Speaking of the other guys, a lot of what's been happening over the last couple of years has come at the expense of other architectures, so MIPS is no longer showing up in systems.

Barrett: Well, MIPS is still around. It's targeted at embedded applications.

Tech Update: PA-RISC is being phased out; Alpha, too. So, it's a three-horse race at this point, at least when we talk about the data center.

Barrett: There is SPARC. There is Intel. And there's Power, off of IBM.

Tech Update: A lot of people who depend upon tried-and-true, battle-tested hardware in their data centers are not convinced of IA-64's virtues in the same scenario.

Barrett: If you look at a typical data center today, what is it? If we say that the world is going to be Web-centric, it's a three-tier data center--right? So the typical data center has Web servers, app servers, and database servers. What percentage of those three classes of servers are on Intel's architecture today? The answer is 85 percent of them are IA. The big database server is the part that is still to some degree the domain of IBM's Power stuff.

Tech Update: What's the road map for penetrating that space--the mini/mainframe or the telecom equipment areas?

Barrett: You asked two pretty divergent questions there in the same sentence. Well, the first is the one that we are pursuing here, which is you go with people who supply big iron. The main customers for this capability are companies like Hewlett-Packard, Compaq, and IBM. IBM is a little bit agnostic in this space. They give their guys free rein. You can go with Intel. You can go with [IBM's Power] architecture. We go with those people who are the competitors of the other guy in this space for this business. [Editor's note: Barrett is referring to Sun.]   We give them a price/performance advantage and we just say, go at it. That has been the game model. The game model has been, "OK, here is the solution. Now go sell it." To do this we need the OEMs who are going to take the product and then the operating systems to go with it--and we talked about the OSes already. Then, you have to worry about getting the applications supported and tuned on [IA]. That's what is ongoing right now. Meanwhile, we are just about to introduce our second generation of I-64: Itanium II (formerly McKinley). Itanium II just ratchets up the whole price/performance capability another notch for us in this space.

The big iron database machines is a kind of a topic on its own. The database machines in the computer world are really the last vestiges of the old vertical integration systems. They are the last vestiges of the mainframe computer. Now it is going toward this more horizontal approach that we've been talking about. Here's the hardware--standard hardware. Put the OS on it that you want. Put the application on it that you want. Put the database on it that you that you want.

As for the telecom industry, you have seen this same sort of massive dislocation in the last year. It is 20 years of computer learning crammed down the mouth of the telecom folks in 12 months. It's very difficult to digest that. Just look at the change and the impact on companies like Lucent, Nortel, Ericsson, and Motorola. Whether you pick the wireless side or the fixed wire-line side of this business, look at the dislocation in those companies as they have had to say to themselves, "Oh my goodness, this old vertical stack model doesn't work anymore."

That's the effect of price/performance. With price/performance and standard modular building blocks, you can get the capacity you want. Our message in that space is precisely that. You don't have the capability. You don't have the R&D. You don't have the engineering resources and you don't have the ability to hide your costs any more. It's a competitive marketplace. From the time-to-market, competitive, and TCO standpoints, you're going to end up having to go to with modular building blocks that have open interfaces and let the whole industry innovate on these platforms. You can't keep a proprietary vertical stack.

Tech Update: What happens if Scott McNealy doesn't hear that message? Especially the part about hiding costs?

Barrett: Scott has some interesting strategic choices to make.

Tech Update: Is decommissioning SPARC one of them?

Barrett: Well, that's Scott's choice. That's not our choice. But you're probably getting a feeling for what's happening. Our server market is going pretty good. It is the desktop part of our business that has been hurt the most in this recession. But, if you look at Scott's business over the last year, it has not been sterling. That's the word, I guess, I would use. He made the first layoffs in [the] history [of his company] and he's struggling for profitability. If you read the tea leaves going forward and you look at the design-win activity that is going down, which is a forecaster of the next five years of business--I'm not sure that I would be comfortable if I were in Scott's shoes.

Tech Update: Would he be better off hitching the Solaris horse to the Intel engine?

Barrett: Only if he was serious about it. We've been there before.

Tech Update: I mean strategically, to the same extent that HP or Compaq have committed.

Barrett: Well, if you start to argue that this is going to be a price/performance sell and you assume that he has been hiding behind his ability to sell hardware at margins that are atypical of this industry, you can expect that whole issue will be exposed as we go forward. So, the question is, if he can't hide behind that margin model anymore, then what does he do to compensate? But, that's an issue that Scott has to figure out. If he wants--and he's serious about a design win with us--we are willing to talk to him. But I sometimes sense it is more of a religious issue with Sun than it is a pragmatic business issue.

Tech Update: As you said earlier, IBM is saying, "You can take this Intel thing or you can go in-house with Power." When I interviewed IBM Senior Vice President Steve Mills, I asked him about IBM's duplicitous relationship with Microsoft. On one hand, he relies on Microsoft for a large part of his business. On the other, they're huge competitors, especially on the software front. With hardware, IBM has almost the same sort of relationship with you. They say, "Intel's what we want to push out at the low end, but consolidate using our high-end hardware." What's it like doing business with such a competitor?

Barrett: As long as we have a fair shot, and the best man wins and the best woman wins--or the best person wins--we'll go at it. IBM has franchises that they are trying to protect, just like everybody else. They are a very good customer of ours. Their Intel business is good in the server side, and we intend to make it better and continue working with them.

Tech Update: You've collaborated with them on the server stuff, with Linux in particular. But, because of the way Linux runs on everything else they sell, that also creates a path that leads from Intel to Power and allows IBM to siphon customers away from you.

Barrett: Well, our server business continues to grow.