What’s Changing Inside Intel?
By Ed Sperling -- 4/28/2006
Pat Gelsinger, senior VP and general manager of Intel’s Digital Enterprise Group, sat down with Electronic News to talk about the company’s future directions, how that will build on its existing core competencies, as well as new processors and who will benefit from them. What follows are excerpts of that interview.
Electronic News: Is innovation leveling off or is it increasing?
Gelsinger: In some ways, you’re seeing clear trends toward less innovation at the chip level because of the cost. What does a 65-nanometer chip cost? That’s why we’ve seen the number of ASICs declining. At the same time, you’re seeing things like FPGAs growing. You take all that together, and there are less silicon designs than there were in the past. From my perspective, as we make these architectural transitions to a quad core or oct core and multicores, we’re seeing an acceleration of innovation. I think we were in a period that was boring for a number of years. I see them getting pretty exciting in the future.
Electronic News: But it also looks as if other pieces have come a long way, too, so that the integration of all of these technologies can make a much more innovative device.
Gelsinger: I’d agree with that. Innovation is a non-linear activity. It’s not like Moore’s Law is just plunking along and innovation is moving with it. You get to points where dislocations happen. When Windows happened, it wasn’t just like one day you couldn’t do it and then the next day you could. All of a sudden you had enough excess capacity where you could change the paradigm of the user interface. We have gotten to a period where there’s now enough excess capacity where you’re seeing that next step of innovation. There is general connectivity in place across broad wireless system. There are thin and light form factors. You can start putting a lot of functionality into small form factors, and there’s enough computing capacity left over where your application types become interesting. I do see a cross with high-density memory changing the memory hierarchy. On the horizon there are major changes in what we consider a computing device.
Electronic News: Will those dislocations come from software or some application of the technology that already exists?
Gelsinger: You’ve always seen these dislocations happen with hardware and software. If there is enough hardware capability, the next step happens. We’ve also seen with mobility and wireless that they’re occurring at this form factor, as well. As you tie that all together, there are new domains of usage that customers are interested in.
Gelsinger: About 15 years ago, I said there would be five global communications companies and five local communications companies. There are monopoly effects and regulatory effects and all these other things that get in the way of the progression of technology and business. There is an incredible period of innovation happening in communications infrastructure. The world is going wireless. If you think about our business over the past 15 to 20 years, the most important customer for us was the CIO. Over the next 15 years, it may be the CTO of the carriers—the communications companies. They may become the pools of the biggest infrastructure services that get built and delivered.
Electronic News: The flow of information is centralized for them. Intel has always worked most effectively in a decentralized environment. Will your future be in a new direction?
Gelsinger: It could be. You’re seeing new battlegrounds emerging. If you look at the current debate on Net neutrality, you see new companies like Google lining up on one side and the carriers lining up on the other. Intel has taken a middle ground to allow innovation to occur.
Electronic News: This is a radically different approach for Intel, isn’t it? The focus is now on the way the customer interacts with the technology rather than the technology itself.
Gelsinger: It is a new perspective. It’s not that Dell and HP and IBM aren’t important to us, because they are. But now we have other constituents, such as the carriers, ISVs and the service providers, that become very significant to us, as well. Ultimately they become different channels for us to understand what the real end user wants in the underlying products. Increasingly we are stretching ourselves out to understand the end user.
Electronic News: So ultimately do the carriers become the glue between you and the customer?
Gelsinger: Glue is too strong a word. When you think about your cell phone, do you think about the components inside that may come from TI and Intel, or do you think of it as the phone by Motorola, Nokia, LG or Samsung, who may be building it, or do you think about the relationship with Verizon or AT&T or whoever is providing the service?
Electronic News: Most people think about whether they’ve got a good connection, don’t they?
Gelsinger: Yes. That’s the service portion. But how do you continue moving the technology along so you can add new services? Market shifts occur largely based on a particular service that a carrier drives forward. You can imagine that in the future a lot of PCs will be delivered like that, because I’m buying a service relationship for a combination of my WiFi and 3G service connections. That becomes the billing relationship. And since more and more of the applications get delivered that way—and we see more and more software as services and more and more of the big data centers like Google and Yahoo! and Microsoft Live delivered as services—that relationship will become very different.
Electronic News: Something else seems to have change here, as well. Intel used to be far more independent. It now seems to be more reliant on a ring of partnerships.
Gelsinger: In many ways, I see that as a skill we’ve developed. We couldn’t do USB. We had to use third-party silicon, we had to facilitate plugfests, we had to enable the usage model branding in the industry. We did it with PCI [peripheral component interconnect], we did it with AGPs [accelerated graphics ports]. It’s not a new skill, but the scope and the breadth have changed, because now we’re trying to facilitate an entire service relationship through broader ecosystem players across global environments. WiFi was a baby step for some of the things we’re trying to do now.
Electronic News: Has Intel’s core competency changed?
Gelsinger: There are elements of our core competency that are unending—underlying process technology, being able to manufacture in high volume, and doing sophisticated high-volume design. That’s what we do. Some of these other competencies—we now have thousands of software engineers engaging with service providers, being able to facilitate end user models—these are things we layer on top of those core competencies. It’s pretty far from where we were a decade ago.
Electronic News: Intel is experimenting in a lot of other areas such as life sciences, too. Where is Intel going with that?
Gelsinger: Our healthcare activities fall into three areas. One is IT in institutional healthcare. How do you accelerate adoption of that? My PC is a pretty lousy form factor when you think of a doctor or nurse trying to handle their activities. That’s traditional IT in institutional healthcare. IT penetration in healthcare around the world is the lowest of any of the major verticals. That is a pretty frightening statement. The second area is personal healthcare. We have mainframe healthcare, called hospitals. Now we need the PC version, or personal healthcare, with connection into the home environment and into self care.
Electronic News: Is this like one specialist serving several thousand people over the Internet instead of a few hundred in an office?
Gelsinger: Possibly. Some of it might just be extending the relationship into the home of self-diagnostic capabilities. We’ve had some experiments with Parkinson’s [Disease]. The patient now shows up once a month to do a Parkinson’s test. We can have them do it three times a day at home. It may require sensors in the home to be able to assist the dispensing of drugs. It may be things like being able to facilitate and monitor health and activity, where you have the personal device that measures your movement during the day. You create the wireless connection so it uploads your heart rate into your PC.
Electronic News: What’s the third area?
Gelsinger: The things we build at the transistor level are now smaller than all of the components that are the basis of life—proteins, genes, DNA. We can start analyzing at the single DNA level. This is the deep biological science. We have world‘s greatest chemists and physicists and the finest analytical tools on the planet. We added a few biologists in with them and we’re getting great results. We can do a single DNA analysis or a single protein analysis. We can detect protein deviations that have never been analyzed before.
Electronic News: What you’re talking about is approaching the quantum level. Is that correct?
Gelsinger: Yes. You’re literally at the single protein level, looking for deviations that have never been analyzed before. You can combine that with spectrometers and weight measurement. But now you can actually look at the shape of a protein. Instead of watching the onset of cancer or some other disease in two years, when your body is filled with it, you can actually detect the first change. Today, when you’re doing DNA tests, you’re looking for statistical deviations at a minimum of 10,000 stands of DNA. We’re talking about 10,000 versus one—being able to detect at the single level versus the broad statistical level. The key, we believe, to treatment of many life-threatening diseases is early diagnosis.
Electronic News: Will Intel play a significant role in that market or provide the tools and technology and maybe even investments for other companies?
Gelsinger: It’s too early to tell. When we were first coming out with single-board computers, we realized that people needed tools to design them. We created development stations to go with that. What we’re doing today in life sciences is essentially that—the single protein analysis station to enable people to begin analyzing these things. These are the first very, very crude instrumentation tools for people to run these experiments. Will that lead to in-body technology or tools that change our business model or create a new business model for us? It’s too early to tell.
Electronic News: Would a market slowdown in one area, such as PCs, provide sufficient impetus for Intel to grow its business in another area?
Gelsinger: Based on the core capabilities that the company has, we’re going to expand into as many areas as we can leverage those core capabilities into. Will we stop investing in PCs and servers because our handset business grows faster? No. We’ll starting building more fabs and expanding that way. I would hope the Intel of the future becomes more diversified in terms of the markets and the products we’re delivering even though I don’t see the underlying competencies will deviate that far from where we are today.
Electronic News: What has stopped Intel from doing that in the past? There have been a lot of ventures into new markets that have gone nowhere.
Gelsinger: For every industry, you can point to a few cases where companies have successfully leveraged their competencies in other areas. But for the most part, they don’t. It’s very hard to do, and I don’t know that we’re going to be any more successful in the future than in the past. But we’ve taken a much more structured approach, with much more application focus as to how we’re trying to partner and develop those relationships. We’ve complemented that with our ICat [Electronic Commerce Suite] quite successfully. We’ve taken a very disciplined, bottoms-up view of our competencies, the markets and the partnerships that we need to form, that lead me to be optimistic there is a good chance for success.
Electronic News: Let’s shift subjects. How many cores will you have in your chips, how will they be managed and what will they be used for?
Gelsinger: I don’t think we know how many there will be. That’s a topic of both research and product and market understanding. We’re on track. We have lots of duals out already, and we have quads and octs under development. Each of those cores can support multiple threads of execution. You can have 16 or 32 threads each.
Electronic News: But each core is still a processor that deals with ad hoc queries. Haven’t those always been a nightmare for multiprocessing systems?
Gelsinger: Yes, and as a result we’re spending a lot of time looking at workloads so we can take advantage of those. If everything’s a single stream of execution leading to another single stream leading to another, there’s not a lot of parallelism to be gained. Those types of applications are not going to see enormous leaps in capability. At the same time, our observation is that most of those applications don’t need a whole lot of performance. New types of applications, such as recognition and human interface, mining, synthesis—those have enormous potential for parallelism. If I was going to model this room, how many rays of light are in it? There are a lot. That’s almost perfectly parallel. When you do a text search on Google, that’s embarrassingly parallel. Imagine running Google Desktop for everything. And that will be done on not just text. It could be an image, it could be voice, it could be a particular set of data or financial information. Those are the kinds of domains that will create entirely new steps in computation.
Electronic News: One of the things you mentioned at the Intel Developer’s Forum is virtualization. Virtualization historically has meant sharing of resources across an enterprise. How does that extend into the consumer world?
Gelsinger: That’s too narrow of a definition. If you’ve ever received a virus or some cookie over your e-mail, you just say, ‘Let’s run the browser in some virtual machine.’ When the virtual machine ends, it’s gone. You can get to the point where every application runs in a virtual machine. An operating system is providing a platform to share scarce resources. In the future you will have plentiful resources. You simply virtualize a pool of resources and you change the relationship and the way you think of applications and the operating system. That’s may be a little abstract. Let’s take a simpler example. We’re working on capabilities such as manageability, where we create a virtual partition, we create a secure environment and now my IT department can reach into the PC to manage it remotely. Can you do that at home? Can your service provider reach into your home network and have a virtualized set of resources that are protected and secure and manage those resources for you? Yes. They might be backup resources. They might be network management, security and virus protection or home security. There might be a whole set of things they can do. You can partition your environment between your kids’ games, your media center, your tax preparation and your secure documents.
Electronic News: Then it’s a matter of being the gatekeeper of your data?
Gelsinger: It certainly creates that ability. You could then delegate that responsibility. You may want Verizon to be the gatekeeper. I want to be able to make sure the entertainment things I want to do tonight are made available, my digital memories are secure forever, and my personal information from my heart-rate monitor to caloric monitoring are stored in my home machine.
Electronic News: Looking at all we’ve discussed so far, does this mean your close partnerships begin to change? It used to be Dell and Gateway and HP.
Gelsinger: Yes. I see it as a broadening set. I don’t see it as replacing Dell. To a great degree, if we come out with a new platform, they’re going to ask what are the applications, who is the audience. Many of the things I need to do to make Dell happy require us to go further into the ecosystem, as well. If it’s a data center, they’ll ask if SAP is ready to go to market. ‘Have you done that work? Are the service providers lined up?’ It’s what they need to move through the next chain of the ecosystem.
Electronic News: On the flip side of that, what will be some of the most important technologies in the future that potentially could be disruptive?
Gelsinger: You’ll see a few going on that are really interesting. The whole transition to software as a service is a big deal. Watch what’s going on between Google and Yahoo! and Microsoft Live. That’s big and it’s important. The whole idea of these very parallel data searches—Google everything—that’s a big deal. Google has taken the parallelization of data and they’re applying that across a broad domain. People like Microsoft and Yahoo! are taking this very seriously. We’re selling lots of chips to these guys.
Electronic News: And we’re just dealing with the free services. There no doubt will be paid services, right?
Gelsinger: Yes. That, to me, is when things get really interesting—when you start overlaying contextual information and fee-based services into different vertical domains. Obviously Microsoft brings their software as a service, as well. Maybe they never catch up to Google in the consumer space, but they’re going to own small and medium-sized business and play a much bigger role. We’re very intrigued by some of the Web 2.0 trends, with RSS and mash-ups. The browser has moved from an HTML interface to being a platform for a pretty broad set of services. Just as hardware development is starting to heat up, the software development is starting to heat up. It’s like the Internet craze starting all over again with a different set of technology and the infrastructure built out. That’s another one we see as being a big deal.
© 2009, Reed Business Information, a division of Reed Elsevier Inc. All Rights Reserved.