Q You?re only 36, you?ve been at Intel for 17 years, during which time you?ve headed several important projects and product groups ? surely now you can realistically aim for the top job?
A Last time I checked Andy [Grove ? Intel?s president] and Craig [Barrett ? its chief operating officer] were still occupying their positions well, and I have a long career in front of me yet. Obviously I have a pretty significant role at Intel, they?ve given me incredible opportunities. When I was 25 I was running the 486 design team, a fairly critical role for a comparatively young individual. I now work very closely with Craig and Andy trying to learn and following in the footsteps of some of the greatest leaders the industry has ever known. So, even if I don?t progress any further, I feel pretty privileged.
Q What do you see as Intel?s role in the industry today?
AWe like to contribute to what we call managing the ecosystem of the industry ? we need to make the whole industry move forward. We need standards, products and technologies in place so that we can move forward. One of those for the business user is manageability. The PC is now part of the fabric of the enterprise, and it needs to be managed as such.
The Wired for Management [WfM] initiative encompasses many places in the industry where we don?t directly sell products. At the client level our objective is very simple: we want every single PC built to be a manageable, managed PC, so every one can be remotely configured, upgraded, installed and booted. This common level of manageability is set by industry standards like DMI [desktop management interface] that define the key interfaces which are incorporated into everything we build.
QThe whole industry seems to have taken up that mantra. But many people think over-sophisticated software is to blame for lack of manageability.
AYeah, if the PC is manageable, but if it doesn?t have management tools that allow that to be used by corporate IT and MIS, then it doesn?t matter. It really has to be connected up end-to-end throughout the enterprise.
We can?t do everything, and we need the co-operation of Microsoft, CA, Tivoli, all the software people, so we work closely with them. Our WfM initiative is entirely complementary to Microsoft?s ZAW [Zero Administration Windows] initiative. Their software layer builds on and complements our hardware level.
QOne model of manageability getting a lot of press is thin client and the network computer.
AThere?s a lot of myth, hyperbole and rhetoric about NCs, so let?s decompose it a bit. It began with TCO [total cost of ownership], but when you understand the numbers, a fully-managed PC has substantially less cost of ownership than an unmanaged PC, and if we compare that against a model of where an NC might be, they are about the same. Because the NC requires bigger servers, bigger networks, bigger storage on the server side, and when you add that in, they end up being about the same.
So our first response to the TCO thing was that we were going to aggressively manage the PC. So we got WfM and the NetPC. The second issue was the cheaper client, the idea of a $500 NC. Well the $500 NC never quite made it, and PCs can be purchased for a little more than the actual cost of an NC.
The third tenet was Java. And it turns out that by far the best client for running Java is the PC. We have a client who says his Java runs 10 times faster on a NetPC than on an NC. Java requires more performance, not less, from the client; it benefits from having local disks for cacheing and memory for storage.
QIsn?t the NetPC just Intel?s and Microsoft?s answer to the thin-client movement?
AOur perspective on this is somewhat different to Microsoft?s. We also have companies like Oracle which are creating Intel-based NCs. Essentially, they are taking a NetPC design and taking out the hard drive and creating an even smaller client. We?re happy with people doing that.
But a NetPC fits into the same network infrastructure as a PC. It?s a very straightforward solution for people to deploy: they have the networks, they have the applications, the development environment ? it?s a no-brainer by comparison.
QYour deals with Oracle look like securing you a toe-hold in the NC market in case it takes off ? a kind of insurance.
AWe will participate at all levels of the market. We see a lot of experimentation happening where people are looking for terminal replacements, they currently have 3,270 terminals that are 9 years old and they?re wondering what to do next. That?s the kind of place where people are doing more thoughtful evaluations, and that?s where NCs might get a toe-hold. And we?ll support it.
QMicrosoft has entered a new market for portable computing with Windows CE. What is Intel doing for that handheld market?
AIf it?s reprogrammable, then it?s an Intel-Windows product. We?ll draw the line between reprogrammable and narrow, fixed-function devices, and that?s where you?ll see Intel draw the line strategically. On the reprogrammable side we want to take the Pentium family and scale it down, drive the power down, drive towards long battery life, flexible user interfaces, integrate functionality ? just as the laptops have gone from luggable to light, we?re going to push the envelope as far as possible.
QPower consumption has always been a problem for portable devices. How are you tackling that?
AWe continue to push to lower voltage devices with their more flexible power management techniques, to push that envelope. I think there?s a breakthrough at about 6 hours of battery life; when you get close to 10 hours of battery life and it?s less than 2lb, I think you will have broken through some sort of user paradigm. But I?m not a marketing expert.
On the embedded side, we have an entire family of products that are embedded processors. For example, smart phones where we have deep relationships with Nokia, Ericsson and people like that, putting PCs in cars?
QThat?s a way of finding uses for older chips. Doesn?t bringing out a new generation of chip simply kill the revenue from older ones?
AWe made a fundamental decision about this many years ago. If we don?t eat our own children, someone else will ? if we don?t advance a stage of technology as fast as we know how, then someone else will do it.
Economically, any processor generation that we introduced was a bad decision. When we introduced the Pentium, it was a rotten decision ? there are enormous costs associated with marketing, ramping up production. The 486 was at the peak of its lifecycle, we had driven the cost right out of it, the market understood it. So why did we introduce the Pentium? Because we knew if we didn?t advance the market, someone else would.
And every time we move to another processor, we open up the next set of market opportunities. With the Pentium II market, we?re seeing this tremendous momentum in the new market segments where we never were before. In the workstation market, for instance, all of a sudden the Intel-based workstations have half the marketplace, and IDC predicts we?ll have 90 per cent of it by 2001. That?s a $5bn market where we had no participation but, because we pushed the envelope, it came into our purview.
QLiving up to Moore?s law is one thing, but is there a maximum useful speed for a processor?
AOne of my jobs at Intel was helping to introduce the 386, and we really wondered whether anyone was going to be able to use a chip that fast. But we found uses. With Pentium II we see digital content creation. I?ve asked that same question with every generation of chip that I?ve been involved with at Intel, and over the past 15 years it has all been used up and a lot more. Our business customers are upset with us that we cannot deliver higher performance servers any more rapidly; they just want to scale things up.
QOne could argue that Intel and the software companies egg each other on to keep people buying newer technology, finding ways to a sell a technology whether there?s a demand or not.
AThe industry is definitely in danger of doing that. It?s just that we?re excited about technology. The cost of ownership example is one where we became exuberant about the technology and forgot about some of the business problems ? that was a pretty good wake-up call. To be honest there wasn?t any reason why we couldn?t have solved some of these manageability problems three or four or five years ago. We just didn?t recognise some of those customer needs.
We look at the major trends that businesses are deploying and looking forward to, and we see tremendous opportunity for increased capability on the desktop. Internet connectivity, for example, opens up more and more doors with intelligent capabilities. Electronic commerce changes the way businesses interact with their customers, and they?re going to want modern PCs, they?re going to want enhanced security, operations security ? these algorithms take lots of mips.
QIntel is often seen as the biggest gorilla in the jungle, and with MMX you cut out a lot of companies making graphics chips.
AThere was a time when people asked ?should we go to integrated flip-flops or stick with discreet flip-flops on the motherboard? Boy, we?re taking value away from the flip-flop manufacturers.? It?s an age-old argument, but you get to a point where the technology stabilises, it?s standardised from a cost perspective. If I can build more transistors onto a chip, I can look at integrating the right degree of functionality. That can make a much larger marketplace as the specialists look to integrate at the next level up.
QYour main rivals use pretty much the same architecture as Intel ? some even call their products ?x86? ? so you?re setting the pace. Can Intel ever be displaced?
ACompetition is wonderful; it keeps the market moving forward, customers get better value and if anyone stops technologically, or stops giving customer value, there?s always someone to fill the gap. So how do you move the market forward? Better features, new capabilities, lower prices, better availability ? it?s the same whether you?re selling T-shirts or computer chips.
Of course, if we miss a market transition, if we fail in an execution, if we ship inferior quality products?
Q...but you?ve done that before with the Pentium bug and ridden it through.
AWe?re vulnerable, and we had to respond aggressively. We succeeded in that case. Can we be displaced? People said that about IBM. They missed a whole technology transition.
QYet IBM is still the world?s biggest technology firm.
AThey missed a market transition, they lost a huge opportunity, and they?re not in the position of strength they used to be in.
QDo you get fed up with the Wintel thing?
AYeah, because we are two very independent companies. It?s an inappropriate way to look at the industry. We both have strategies that are independent of each other. We support Unix on our platform ? they sort of don?t like that. They have people who do Windows CE chips, we sort of don?t like that.
It?s natural people should come to that conclusion, but it?s also inappropriate to assume it?s a conscious or unusual alliance. Our relationship with Microsoft is very beneficial, but it?s also very tense in other areas.
QHow far does Intel plan ahead?
AWe have people working on chips today that we won?t introduce until about 2003. That?s our development horizon, and we have people dreaming about things even further beyond that. In fact, we have people developing the process technologies that will take us through to about 2005.
As big as the PC market has become, it still only represents about half the overall computer market today. There?s still enormous markets at the high end that we?ve only just started to touch. There?s about $250m of mainframe and mini-computer industry that we haven?t entered yet in any significant degree. So moving forward the 32-bit architecture and the 64-bit architecture (in particular the Merced architecture), that?s a really big deal for us. We believe that?s going to move us into those enterprise markets.
We have to up the performance level, we have to make much larger servers available, go to much higher degrees of processing, much more robust transaction level systems that never go down, I/O systems that are larger than the largest mainframes today, RAID ? some pretty dramatic new areas for us to go into.
QHow do you feel about standards? Intel has the de facto standard for the PC.
AThere are standards that the industry builds on, and there?s de facto standards, associated with product implementations. The former are the things that create the horizontal industry, the ecosystem of the PC. People build on them because they see product opportunities or ? as in the case of Windows ? because they have to to be in the Windows application market.
In the Java case, McNealy is trying to have his cake and eat it. ?I want it to be an open standard for the industry, even though nobody has any products built on it yet, but I want it to be proprietary and only controlled by me.? We?ve taken a negative and aggressive stance on this. I have people going to every ISO meeting arguing against the Java standard. We say, free up the technology so it can really evolve in the industry or get outta here. If you?re going to propose this as a standard, but control the definition, the evolution, the technology, the copyrights, the patents and everything else associated with it, don?t call it an ISO standard, because it isn?t.
We would love Sun to open up Java to the industry. I?ve met with McNealy, Barratz, everyone associated with this, and I?ve literally begged them to make it a real standard, because I think the technology really has legs if they?d open it up to the influences and direction of the industry. But if they don?t, it?s going to be yet another failed proprietary attempt.
The industry over time will not build on something controlled by someone else. You do that after you?ve created the market position, like Microsoft has with Windows. It supports its developers very carefully, it evolves those APIs very carefully, and people do it because it?s good business.
QWhy should Microsoft and Sun be treated differently?
ABecause Microsoft has an installed base of Windows. We do numerous standards for the industry. The PCI bus ? we defined it by taking good engineers and creating the bus. Then we created the industry standard to have help from the industry in finishing and evolving the specification; we also freely licensed all the technology associated with it and took the PCI name we had trademarked and gave it to an independent group. Now that?s a standard. And it?s used on every PC, Apple machine and Sun machine on the planet.
QThose standards help you sell more motherboards ? but if Sun surrenders Java, its revenue stream just disappears.
AMicrosoft gets its power in the industry because ISVs build on their interface. Sun could look at that and say, ?Boy, if we could get people doing that in Java?, then they could potentially create a lot of industry ISV influence and they don?t need to control the language to do that, they could truly license it, they could have a lot of influence, a rapid evolution for it, and we think it would be a much superior strategy for them than the one they?re on at the moment.
QBut they might not make any money that way.
AI think they?re going to make no money with the language itself. In fact, they?re going to drive people away from their implementation of the language by trying to control it so tightly.
Microsoft and Intel both make money from having a huge installed base, so you can afford to take the moral high ground on standards. Are you the king and queen of the IT industry?
Standardisation is hardly a magnanimous act; we look at it as a vital act to continue the expansion of the industry ecosystem in which we live.
QSo if the king doesn?t look after his realm and his barons, they?ll turn against him?
ANo that?s the wrong perspective. Every time we expand the industry, we make it better for everyone in it. Getting a little bit more market share from AMD ? so what? We win when we make the market bigger.
QTo push the analogy, the king invades a new country and all the barons benefit ? but the nearer the throne you are, the more you benefit?
AIt?s critical for us to expand the uses and the users of the PC. That?s the only way we benefit our customers, our shareholders and the employees of the company. If we succeed, and we continue to deliver the best products, we probably benefit financially as well.
Boy, have we spent a lot of money on DMI, and we sure don?t have much revenue associated with it.
QHmmm... but I?m sure it?s not just pure altruism, is it?
AOh, absolutely! The pittance of PCs that are managed in the world? The amount of revenue we get off managed PCs? What a rotten investment! If you were any businessman in the world you?d have fired the guy who came up with that one.
QWhat?s next from Intel?
AYou?ll see a succession of products from us. You?ll see increased clock rates when we introduce the mobile and server versions of Pentium II next year. You?ll also see much higher end versions too, with much bigger caches, much higher degrees of multi-processing scalability. And more clock frequencies and lower prices as well. By the end of next year almost everything we sell will be a Pentium II.
QHow do you cope with the expense of having to ditch old fabrication plants?
AThere?s major economics associated with fabrication ? this year, I think it?s $4.6bn capital investment in new fabs, next year it will be substantially bigger. But we?ve succeeded in refurbishing our fabs for multiple generations. Back in the 386 days, when you stopped building 386s, you just closed it down and built a taco stand.
Obviously, we?re going to ramp up the 0.25 micron technologies, we introduced that for the mobile products, the Pentium II products, which is where we?ll get the higher clock rates next year. And that?s a huge deal for us.
Pat Gelsinger factfile
R Age:
36
R Job:
Intel vice-president; general manager, Desktop Products Group
R Education:
Degrees in electrical engineering from Lincoln Tech, Santa Clara University and a masters from Stanford University
R Career:
Lifelong Intel man. Having joined the company in 1980, Gelsinger has all the credentials to go right to the top of Intel, with experience in both technical and general management. A prodigy, he led the i486 processor development team while still in his mid-20s and was a key designer or a manager on the development of all the Intel processors from the i286 to the Pentium Pro.
From 1992 to 1996, Gelsinger was responsible for much of Intel?s communications products and was instrumental in the Proshare video- conferencing range. He now heads the Desktop Products Group, which encompasses every Intel concept destined for your desk: processors, chipsets, motherboards, systems and graphics components, as well as strategic projects like DMI and the Wired for Management initiative.
Gelsinger became a vice president of Intel when he was only 32. He?s written 20 books on processors, holds six patents and doesn?t look like leaving Intel in the near future. Only a fool would bet against him taking one of the highest posts at Intel ? probably before he?s 40.
Gelsinger himself jokes about his physical similarity to Microsoft?s chairman (?I was introduced on Monday as just a bad haircut away from Bill Gates,? he quips); should he ever take over from Andy Grove, seeing the two most powerful men in IT side by side looking like clones from the same pod, will send shudders down the spine.