IDF Spring 2000
Dr. Andrew S. Grove
Palm Springs, CA, USA
February 15, 2000
DR. ANDREW GROVE:
What I would like to talk to you about today is the magical "powers of ten" that have taken over our lives.
This is the second time that I have had the opportunity to address the Intel Developer Forum. The first one was the first Intel Developer Forum two and a half years ago at San Jose. And the main theme that I was espousing then was that our world was changing and we were rapidly heading toward a world of a billion connected computers.
And consequently, because of the forces of connectivity and connected computing taking over our universe, market growth was going to be accompanied by market segmentation. Different types of devices were going to evolve for different places on the network.
And the hope I was expressing, and kind of the cold reaction to those of you that were there, was that we should work and use the same skills and characteristics of personal computing that has stood us in good stead for the decades before and help adapt the personal computing platform to the different segments of the connected universe.
That's basically what has happened in the last two and a half years.
And to start with, I would like to comment on the impact of this universe on ourselves, ourself being Intel. We have operated with a very clear, well-focused mission statement for a 15-year period before that time, which was that we were going to be the predominant building block supplier to the PC industry. And for the first time in this decade and a half, we have found it necessary to adapt our mission statement. And as you probably have heard many times, our current mission statement is to be the building block supplier to the worldwide Internet economy.
And these are nice words, but they are really words. And I find it useful to remind ourselves of our environment, what it was like when the previous mission statement operated, and compare it with our environment today.
And the best way to do that is to look at one of our very first television commercials where we started advertising the significance of the key Intel building block, microprocessors, and the role they played inside the personal computing.
Let's run that commercial -- for those of you who were here, to refresh your memories. That commercial embodied what we were for 15 years.
VIDEO: "Want to run your windowing software fast? Then you need a real power source inside, a source that can generate the power your software needs, the affordable Intel 486. Power it up and run your software at light speed.
Intel, the computer inside."
DR. ANDREW GROVE: In retrospect, this was a very simple world. Our world, and I submit to you all of our world, has changed to something that is a little more difficult to depict.
But I try to depict it in an animation that I'm going to offer you in a moment that traces the bits that are involved whenever a person sits down in front of a computer and makes a request to the Internet infrastructure to download a piece of software, to download a piece of music.
Let's follow the bits as they traverse through the Internet and do that.
DR. ANDREW GROVE: That's what happens every time you hit the return key and send a request to the Internet, in an oversimplified fashion. But the contrast between that commercial and the image that we get when we trace the flow of the bits in the Internet reflects the changing environment in which Intel and most of you in this audience operate. And, correspondingly, all of us have to make major adaptation to our product strategies and to our business strategies.
I would like to give you kind of a quick status report of what we have done in the last few years in readying ourselves to follow those green and blue bits and provide equipment and silicon chips that keep those bits going -- storing, controlling them -- and provide as complete a solution set to that bit flow as we can.
And to do that, I'm going to ask Dan Nainan, from the Intel demo group, to give us a run-through of a current Intel product line. Dan, it's yours.
DAN NAINAN: Good morning.
Well, let's start out with some of Intel's familiar core products. First we have the Intel® Celeron® processor, now available in speeds of up to 533 MHz.
Then we have the Intel® Pentium® III processor, running at speeds of up to 800 MHz.
Now we have the Intel® mobile Pentium® III processor with SpeedStep technology, which runs at 650 MHz while plugged in and 500 while on the road. And notice we also have a Bluetooth cell phone on top of that.
And then we have the Intel® Pentium® III Xeon processor-based workstation running at up to 733 MHz, while the Intel® Itanium processor will be one of the fastest workstations possible in the near future.
Now, let's look at some other applications of Intel silicon. For example, here we have the Intel® Play QX3 Microscope, the PC Camera, the Intel® Celeron® processor-based Intel Web appliance, combining the best of phone and the Internet. We have the Internet set top box, powered by Intel technology, the HP Jornada, powered by Intel processors. And we have the Nokia 6162, powered by Intel flash memory. And we have a cell phone with a cellular chipset from DSP Communications, a recent Intel acquisition. And, finally, over here we have the Blackberry E-mail pager, which has an embedded Intel 386.
Back to you, Andy.
DR. ANDREW GROVE: That's the client. And how about networking?
DAN NAINAN: Okay. For networking, over here we have the Intel® AnyPoint Home Network. We have the series of Intel DSL external and internal modems. Speaking of high-speed access, we have the Stanford Telecom products, another recent Intel acquisition.
Over here, we have some networking products, including the PRO 100 management adapter, the Intel LAN-on-motherboard, and then the PRO 100 LAN plus modem CardBus 56. And we have products from another recent Intel acquisition, Level One, including the 10/100/1000 Ethernet gigabit switch. We have the Intel Express 10/100 Stackable Hub, assorted routers and switches, the award-winning Intel InBusiness series for small business, some products from Shiva, another recent Intel acquisition, the VPN Gateway and Dial Access Concentrator. A product from Dialogic, another acquisition, the IPLink card. The Intel Cache Appliance. And, finally, two products from iPivot called the -- Should we give out the name yet?
DR. ANDREW GROVE: Well, no, no, no, no, no. Let's hold something in suspense for the next couple of days.
DAN NAINAN: Okay. Great. And we have the e-Commerce Accelerator and the e-Commerce Director.
DR. ANDREW GROVE: Thank you, Dan.
DAN NAINAN: Thank you.
DR. ANDREW GROVE: So that follows the bits all the way to the servers. But the picture is not complete until we get a good look at our server architecture. Video.
BURT: I've got a delivery here for Andy Grove.
DR. ANDREW GROVE: Before I sign, want to walk me through what you've got for me?
BURT: We've got a whole rack here of leading Intel architecture servers.
What we have on the top is our small form factor 1u. This is a Pentium III processor server. And this -- let me take this opportunity to show you this one. This is a 2u dual Pentium III processor-based server with a 133 front-side bus and a very small form factor for the power. This next one we have is a 5u two-way Pentium III Xeon processor. This is a 7u four-way Xeon. This is a four-way Xeon. That will be coming out later this year. And then down here we have our eight-way server.
And up at the top, we have an added surprise. We have a four-way Itanium processor server. Even though we're not shipping these right now, this will power up and actually run.
DR. ANDREW GROVE: Actually, we are shipping a few of them.
BURT: Are we? Okay.
DR. ANDREW GROVE: I'm told too few of them.
DR. ANDREW GROVE: So you want to take that back and send it to a customer who does development with it.
BURT: Would you like to see what's inside this thing?
DR. ANDREW GROVE: Absolutely.
All that fuss about that little chip?
Burt, you have delivered.
BURT: Thank you, sir.
DR. ANDREW GROVE: Thank you very much.
DR. ANDREW GROVE: I think the spread of this product line, which is really a work in progress, indicates what we are trying to do. We are trying to follow the bits and put devices everywhere where they can benefit the flow of technology and take the same skill set that we have brought to personal computing, mass production, cost effectiveness, and apply it to the string of the bits as they travel on the Internet.
What is driving all of this is a major change in what drives computing today.
The phase corresponding to that -- the phase that was epitomized in that video -- was personal computing times, which were characterized, which were driven, by productivity software and basically games.
Today, and in the future, it is better to look at the Internet applications as being driven by three different classes of applications involving information, transactions, and markets.
But before we take a look at what we mean by information, transactions, and markets, I think it is important to realize that the growth scale, the rate of growth that we are dealing with, outdoes everything that we have experienced in the personal computing age.
The Internet is driven by powers of ten in just about all aspects: customers, transactions, and power.
And the best way to illustrate the workings of this power -- powers of ten -- is by asking some people -- who live it in their daily life and struggle with it and build the Internet in these categories that I mentioned -- to tell us about it.
And the first among those is going to be a representative of Google, in fact, the CEO of Google, Larry Page.
Google is a search engine company. It takes some brand-new techniques to bring intelligence to searches. And it is a company that lives the powers of ten perhaps as well as anyone else.
The number of searches that Google has implemented have grown through orders of magnitude in a little over a year's time, from 10,000 to 11 million in 14 months' time.
Please welcome Larry Page, the CEO of Google, who is going to tell us about this.
DR. ANDREW GROVE: Have a seat, Larry, and tell us about Google and tell us about the powers of ten as they work for you.
LARRY PAGE: So thanks, Andy.
Google is a search engine, as Andy mentioned. And let me just tell you a little bit about it. When I tell people Google is a search engine, they always say, well, why should I use it?
And there's a bunch of reasons. One is that we worked really hard to develop technology. So we look a lot at the link structure of the Web. We basically data-mine the whole link structure and try to determine what's important on the Web. We're able to do that in an unbiased and objective way.
And we have a bunch of other technologies as well. We look at linked text and all this other stuff.
And -- but that's not enough. You say, okay, I believe you guys have great technology. But why should I use it?
And we've had great success with user satisfaction. In a recent study, we were ranked highest among all major Web sites, with 97 percent of our users rating us as much or somewhat better than other sites, which we were pretty happy with.
DR. ANDREW GROVE: What has been important in putting all this together?
LARRY PAGE: We basically started at Stanford University. It was part of my Ph.D. work and also the cofounder, Sergey's, work.
DR. ANDREW GROVE: By the way, this is a different sign of the new age. It used to be that people dropped out of college to start companies. Now they drop out of Ph.D. programs to start companies.
LARRY PAGE: And even professors these days.
DR. ANDREW GROVE: That's right.
LARRY PAGE: So, yeah, we started at Stanford. We had all sorts of equipment. Basically, we grabbed whatever equipment that came in the door that we could get our hands on. And we had equipment from Sun and IBM and also from Intel.
And, basically, when I came to start the company, we used the Intel equipment running Linux, because that was by far the most cost-effective for us.
DR. ANDREW GROVE: Why is cost-effectiveness so important to you?
LARRY PAGE: Well, we have about 2400 computers all running Linux, all Intel machines. And we see providing good search is basically providing computer --
DR. ANDREW GROVE: Wait a minute. You didn't start with 2400 computers.
LARRY PAGE: No. We started with two.
DR. ANDREW GROVE: And you grew to 2400 in what period of time?
LARRY PAGE: I guess about a year and a half.
DR. ANDREW GROVE: So how fast are you growing now?
LARRY PAGE: Actually, today, we have to add 30 computers per day just to keep up with our traffic.
DR. ANDREW GROVE: Thirty per day. Okay. So how do you do that?
LARRY PAGE: It's a lot of work.
DR. ANDREW GROVE: We all have our problems.
LARRY PAGE: Yeah.
Actually, manageability is really important for us. That's probably the number-one biggest problem we have -- it just takes a lot of time to keep all these machines running.
DR. ANDREW GROVE: But, basically, your method of scaling Google is adding small computers, 30 at a time?
LARRY PAGE: We like to say, actually, we have a sort of redundant array of computers instead of disks.
DR. ANDREW GROVE: Redundant array --
LARRY PAGE: Inexpensive computers.
DR. ANDREW GROVE: Inexpensive computers. Powerful, but inexpensive.
LARRY PAGE: That's right.
DR. ANDREW GROVE: I'm just helping along.
DR. ANDREW GROVE: Where are you heading with all this?
LARRY PAGE: I guess the ideal search engine, we think, would basically understand everything on the Web, it would understand what you wanted, and it would do the right thing.
And that probably requires a lot of computation.
DR. ANDREW GROVE: When you say "understand everything on the Web," what do you mean by that?
LARRY PAGE: Well, I guess people used to call it, say -- well, we say artificial intelligence, basically. If you had intelligence, you have the search engine that understood everything. And it would be able to answer any question.
And in between where we are now and that sort of ultimate search engine, there's a lot of room to improve things.
DR. ANDREW GROVE: What do you need to do to your infrastructure to move along that room and close to that ideal?
LARRY PAGE: I guess we just need a lot of computation. I mean, I guess Sergey and I both -- who started the company -- were big fans of "more computation gives you better stuff." So Google, we actually believe, does significantly more computation when you do a query than other search engines. And that's part of why it works better.
DR. ANDREW GROVE: Is it simply adding more computers like you have? Or is there a role for more powerful computers for each unit?
LARRY PAGE: No. Actually, we use what's available now. We pretty much use maxed-out Intel machines in every way.
If there are bigger machines available, I'm sure we'd use large numbers of those as well.
More memory especially is very, very useful for us, because, you know, you can start to think about having, like, the whole Web in RAM, things like that.
DR. ANDREW GROVE: Say that again.
LARRY PAGE: We'd like to have the whole Web in memory, in random access memory.
DR. ANDREW GROVE: That requires a fair amount of memory.
LARRY PAGE: The Web, a good part of the Web, is a few terabits. So it's not unreasonable.
DR. ANDREW GROVE: And how do you make sure that you are redundant in schemes like this?
LARRY PAGE: Right now, we actually have our own software to do that. And it actually works pretty well. We have fiber to a whole rack of 80 machines. So the steel beam fell and cut the fiber. And our system stayed up, with slightly higher latency. So we have that now. It's a lot of work, though, to make all these things happen.
DR. ANDREW GROVE: You've got basically the creme de la crème of the computing industry in this room. If you had a wish to make for them, what would it be?
LARRY PAGE: I guess my two biggest wishes would be, you know, low-cost machines that have a lot of memory, and that looks very promising with the new stuff that's coming out of Intel.
And the other thing would be, you know, for people to really think about, a lot of these machines are going into data centers now. And for people to really think, you know, "can I maintain this machine without having to drive to the data center," or, you know, have somebody on the other side of the country to maintain the machines.
DR. ANDREW GROVE: Remote management?
LARRY PAGE: Remote management, right.
DR. ANDREW GROVE: Have you got that, folks? I hope so.
Larry, thank you very much for coming.
LARRY PAGE: Thank you.
DR. ANDREW GROVE: Our next representative of the powers of ten revolution is eToys. eToys was the number two retailer last Christmas -- or, I'm sorry, number-two E-tailer last Christmas. By the way, just to call attention, everything I'll show here is on a semilog chart corresponding to the powers of ten. The customers have grown from 20,000 to 1.7 million unique customers in two years. And to tell the eToys story, please welcome John Hnanicek, the CEO of eToys.
DR. ANDREW GROVE: Again, let's start by telling us about eToys.
JOHN HNANICEK: Well, eToys is the number one retailer of children's products on the Internet. We're more about toys. We're more than just toys. We're all about products for kids: Books, videos. We strive to be a safe place for parents to come to be able to get stuff for kids.
DR. ANDREW GROVE: Safe place for the pocketbook?
JOHN HNANICEK: No, no, no. Safe place for the content.
DR. ANDREW GROVE: Oh, okay.
How fast are you growing?
JOHN HNANICEK: Well, obviously, the chart you showed before is -- is indicative of our growth. We've had -- members have grown from 20,000 to 1.7 million over the last two years. Our sales in the holiday quarter, this past quarter, were 106 million, and that compares to 23 million in the previous quarter. And Wall Street has us -- I think Wall Street estimates currently are about 300 million this coming year.
DR. ANDREW GROVE: What implication do that have on your infrastructure?
JOHN HNANICEK: Well, it's obviously a huge challenge to keep up with the growth.
My background has been traditional retail. And so I always think of our Web site -- if our Web site goes down -- as being the same thing as Wal-Mart closing all their stores, the doors to all their stores, at the same time across the country.
So, you know, availability is key. And when you're growing so rapidly, planning for the traffic that you're going to have is probably the biggest challenge we have.
DR. ANDREW GROVE: What infrastructure did you design for this?
JOHN HNANICEK: Well, our infrastructure is basically two-tier. Our Web application servers are all VA Linux machines, dual Pentium processor-based systems, back-ended by an IBM or Sequent NUMA-Q system.
DR. ANDREW GROVE: Which is based on?
JOHN HNANICEK: Intel.
JOHN HNANICEK: It's still early.
DR. ANDREW GROVE: I didn't want to have a misunderstanding or misapprehension to be propagated here.
How do you grow each of those to keep up with the growing customer base?
JOHN HNANICEK: Well, what's important to our architecture, and when you look at our server farm -- for the Web site applications, you know -- is that utilizing the Intel Architecture, we're able to grow horizontally very easily, adding more servers. It gives a lot of flexibility. Also, it supports tremendous reliability. Because if any one machine goes down, it's a minimum impact to the overall customer experience.
We have over 100 servers right now on our primary site.
DR. ANDREW GROVE: Web servers?
JOHN HNANICEK: Web servers, right. And one large NUMA-Q cluster serving our Oracle database.
DR. ANDREW GROVE: How do you grow the NUMA cluster.
JOHN HNANICEK: The NUMA cluster grows kind of separate from the Web site tier. It really grows up. As we continue to scale, we're just going to be adding more quads to the cluster to be able to get more through put.
DR. ANDREW GROVE: So in a way, one can look at your growth strategy as scaling outward on the Web server tier and scaling both out and up --
JOHN HNANICEK: On the database.
DR. ANDREW GROVE: -- on the database server?
JOHN HNANICEK: Yes.
DR. ANDREW GROVE: Does the Itanium processor have any significance in all of this?
JOHN HNANICEK: Yeah, it has a lot of significance from the perspective of -- and one of the reasons why we're so excited about Itanium processor is that -- given our current growth projections for this year, our Web site would have to go to 4-, 5-, 600 servers possibly. And that begins to just have a lot of management challenges, space challenges. So we're looking for also, you know, higher speeds and higher throughput on the Web servers so that we don't have to grow out as much. Really give us more manageability.
DR. ANDREW GROVE: And just so we can kind of picture, back to the database server, how powerful are these machines?
JOHN HNANICEK: I've had very good experience in utilizing NUMA-Q now for several years, even in a prior life. I think the NUMA-Qs are, you know, comparable, and I would call them the UE10000 killer. And with --
DR. ANDREW GROVE: What is the UE10000?
JOHN HNANICEK: Sun.
DR. ANDREW GROVE: Oh.
JOHN HNANICEK: This wasn't prerehearsed. I would definitely call them an option to the UE10000 in an open environment, in an environment where eToys can feel comfortable that people are writing software and there's lots of tools and there's lots of applications to be able to use for it. And then with IA-64, you know, I think it becomes a mainframe class machine that -- specifically for eToys -- is the right infrastructure for us to continue to grow. Because we expect the kind of growth that we've been seeing for the next, you know, three, four, five years.
DR. ANDREW GROVE: Super. So do we. John, thank you very much.
JOHN HNANICEK: Thank you.
DR. ANDREW GROVE: The third category of the powers of ten has to do with e-Commerce. And we are going to have a representative of Commerce One, who is a leading e-Commerce software and infrastructure company, talk about it.
The power of ten chart that I'm showing here has to do with the growth of e-Commerce, not by Commerce One itself, but e-Commerce that is expected to take place in the next several years.
The numbers, just so I can read them off for you, are going to go from $400 billion of electronic, largely business-to-business commerce, taking place in the current year to $7 trillion in four years' time.
The market potential for this, just to put it in context, the total commerce, electronic and otherwise, in that same period of time is 100 -- I've got to learn new words -- $100 trillion. So the $7 trillion is still representing less than ten percent penetration into that.
In the implementation of change of this magnitude, not just in terms of growth rate, in terms of the power of ten growth rate, but in the sheer physical size, the challenges are really very, very exciting and daunting.
And I will ask you to welcome Sam Prather, vice president of engineering for Commerce One to tell us about their effort to make this happen.
SAM PRATHER: Thank you.
DR. ANDREW GROVE: Commerce One. What do you guys do?
SAM PRATHER: Well, we're building business-to-business electronic commerce portals around the world where we can hook up thousands of buyers, that is, companies that are buyers, to tens of thousands of suppliers, and then we're interconnecting all of those portals together into a global trading web so that any buyer in the world can connect to any supplier in the world and conduct electronic commerce around the world.
DR. ANDREW GROVE: Let me have the picture of this. You and I are members of the trading community. We each are connected to an intermediary portal. And those portals are horizontally connected to other portals.
SAM PRATHER: Right.
DR. ANDREW GROVE: So we can replicate the commerce equivalent on the Internet?
SAM PRATHER: Right. Exactly. Exactly.
We use XML documents to transport purchase orders, acknowledgments, advance ship notices, all of the other kinds of documents that are classically done by paper and are associated with electronic commerce, but we can make them happen almost instantly around the world instead of them having to go by mail or Federal Express or fax or other mechanisms.
DR. ANDREW GROVE: How does all of this compare with EDI? EDI was supposed to be doing something similar to this for the last several decades.
SAM PRATHER: EDI starts out as a point-to-point solution. So the buyer has to connect up every supplier and they have to negotiate with every supplier what the format of the data interchange is going to be. And it has to be a very specific format. So you have all these point-to-point connections, which if you think about it was the way things were before the Internet. So it's kind of like you have a virtual circuit to each one of your suppliers. So only the big guys could do that.
We're using XML documents, so we can have a standard format for the buyer to transmit their documents into our portals. We then transform them using transformations from XML into what format the supplier needs to have. And consequently, we can hook up not only the big guys and replace EDI, but we can also hook up the little guys, small suppliers, like, for example, a guy who mows lawns, he can actually go to the library, he doesn't have a computer, he can go to the library, use a standard Web browser, and get his purchase orders off the Web using our software.
DR. ANDREW GROVE: This is hypothetical or real?
SAM PRATHER: It's a real example. The county of Los Angeles lawn mowers use this.
DR. ANDREW GROVE: Through a library computer?
SAM PRATHER: Right.
DR. ANDREW GROVE: And the beauty of all of this is that he doesn't have to do anything but learn how to use a Web browser?
SAM PRATHER: That's right.
DR. ANDREW GROVE: And from there on, the connectivity through XML document takes place through your portal?
SAM PRATHER: Right.
DR. ANDREW GROVE: Good. You know, our industry measures performance in terms of frequency, operating frequency, and the like. In a world like this that deals with orders, what is your predominant performance metric?
SAM PRATHER: We look at purchase orders per hour going through our portals as a really significant metric.
DR. ANDREW GROVE: Where are you on that metric?
SAM PRATHER: We're able to do 10,000 POs per hour through a four-way Xeon processor-based system.
DR. ANDREW GROVE: 10,000 POs --
SAM PRATHER: Per hour.
DR. ANDREW GROVE: Per hour.
SAM PRATHER: If you think about what that means, the average American corporation, say the average Fortune 500 company, does less than a couple hundred thousand POs per year. So if you multiply that out, you can imagine how many POs we can put through an IA-32 box.
DR. ANDREW GROVE: I'm multiplying it out and it doesn't look like a large enough market for Xeon processors. We need to slow them down.
DR. ANDREW GROVE: What kind of infrastructure have you chosen?
SAM PRATHER: We're using primarily four-way Xeon systems in our portals. We also have some two-ways. We're -- everything we're doing today is on IA-32. So we're building all of these portals around the world all based on NT 4 and the IA-32 architecture.
DR. ANDREW GROVE: What has led you to those choices?
SAM PRATHER: Well, first of all, of course, we're building this huge infrastructure around the world. And we want to do it cost-effectively.
Also, when we go out to a small supplier who wants to have a computer system, we want to be able to provide them something which they can afford and which at the same time has the reliability and the capacity to do the job.
DR. ANDREW GROVE: What's in the future as we move up in the powers of ten chart toward the $7 trillion and you have your appetite to take a good chunk of that on your infrastructure? What does that mean for your infrastructure?
SAM PRATHER: Well, we're going to process vastly more transactions per day than we do today. And so as you can see, you know, from that chart, the amount of business-to-business e-Commerce is growing exponentially.
And so the volume of work that we're going to put through these portals is going to flow dramatically, which is why we need IA-64.
DR. ANDREW GROVE: Specifically what does IA-64 do for you?
SAM PRATHER: Well, I suppose first and foremost, it allows us to process a lot more transactions in a smaller number of physical systems, which means that we improve the manageability of the environment.
Secondly, I think we will ultimately be able to process transactions more cost-effectively, in other words, at a smaller number of micropennies per transaction.
DR. ANDREW GROVE: Micropennies per transaction. That's a real metric?
SAM PRATHER: Sure. And that's an important metric, because in the portal business, you know, a significant portion of your revenue comes from transaction fees. And so what you're interested in is how cost-effectively can I process transactions, and can I do it with complete and total reliability.
DR. ANDREW GROVE: What does -- Kind of as an end, as a benefit to a customer of yours, what does all this mean in terms of transaction cost reduction?
SAM PRATHER: Well, ultimately, a big part of what we're doing is reducing the cost of doing commerce.
So if you look at it back at the buyer in the buying company's organization, if they do a manual purchase order, it costs anywhere from $70 to $120 to process that purchase order through all of the hands it has to pass, through all of the signatures it has to have, and so on. We can reduce that cost by an order of magnitude. And so by increasing our efficiencies in these portals, we can actually continue to bring that cost down in the future for both the buyers and for the suppliers.
DR. ANDREW GROVE: So Moore's Law is supplanted by Prather's law of logarithmic decrease in transaction cost?
SAM PRATHER: Thank you.
DR. ANDREW GROVE: You're welcome.
Thank you for using our product. And we'll try to help.
SAM PRATHER: You bet.
DR. ANDREW GROVE: Thanks.
DR. ANDREW GROVE: These three gentlemen told us about different applications, all of them characterized by the powers of ten-driven growth rates: Information, both aggregating, finding, and viewing; transactions where e-Commerce transactions have to be increasingly aggregated and consolidated per customer; and markets that Sam was talking about, where not only transactions, but decisions are made through the intersections of computers to computers.
That's the environment that is driving all of us. And that's the environment that's driving Intel both on the product side and as an e-Commerce player.
And I want to just spend a minute to share with you some experiences what our own use of e-Commerce technology has meant to computer utilization.
A few years ago, when IT was basically for internal use, the customers were the information technology department and its customers, and the utilization of computing infrastructure mirrored the working day.
As we started to increase the sophistication of the use of our databases internally, data-mining came into it as an additional factor, an additional user of our internal infrastructure, internal customers grew, and with that, as you can see, the utilization of the computers, particularly since we run a worldwide operation, has smoothed out and increased through the entire 24-hour workday.
External customers got added more recently. And that's where we are today, external customers hooking into our internal databases and internal transaction operations, adding yet another dimension of data processing usage. And we are working on adding the universe of suppliers to the community of customers, again, adding more requirements for capacity, competition and capacity. And as Sam talked about, we look forward to the future where we will become members of automated exchanges, an interesting step that ourselves, Intel, and one of our larger distributors, Arrow, has implemented just a week ago. We use RosettaNet to process an information exchange between our information technology infrastructure and theirs, somewhat analogous to the kind of thing that Sam was talking about, but still on a one-to-one basis.
But coming back to the point: when you add all of these things up and aggregate them, you see the total demand way, way exceeding what we started with when information technology inside Intel represented only enterprise-related applications.
The addition of all of these different communities as part of the information technology infrastructure brings about inside of one company the power of ten type of growth in infrastructure ad competition that we have been talking about all through this talk.
The question really is how do we meet this demand. How do we grow the infrastructure without dealing with software migration time and again. And the simplest way is, just add more servers to the existing infrastructure.
We can do that in the development that you saw Burt demonstrate, extending, scaling outward with the addition of similar servers that we have used in the first place. And in addition to that, we can see the growth in the other dimensions where we can add computing infrastructure based on higher-performance microprocessors, higher-performance computing, culminating in the future with the addition of 64-bit processors in the Itanium processor family.
The point of all of this is that the Intel Architecture meets the key requirements of our customer base in the Internet universe and ourselves as an Internet customer. It brings about cost-effectiveness both in terms of redundancy and availability. It allows scalability because of the uniformity of the instruction set architecture, and allows interoperability for the same reason.
It has satisfied a range of demands ranging from information processing to transaction processing and to implementing electronic markets.
And all of this is predicated on us collectively, ourselves and you all, members of the industry, to do the same thing that we have done to create this industry in the first place: Relentless drive toward higher performance at a lower price; developing and industry-wide embracing of standards such as we have done in the PC platform -- we have a paradigm in this industry of collaborating and competing with each other at the same time -- and the use of varied and rapidly developing distribution channels.
We have done it for the PC industry. We are doing it for Internet infrastructure collectively. And the three gentlemen that told their stories all are counting on us to do this job. The PC formula has worked and will work for Internet infrastructure.
And from Intel's standpoint, we stand ready to help. We stand ready to, as you have seen in the breadth of the product line, deploy silicon power wherever the bits travel, recognizing that, basically, the Internet runs on silicon. And the powers of ten demand best performance and volume economics, which is what we have been about and what we are about and what we are committing to be about. And just kind of as a transition to what Albert Yu is going to tell you about next, in terms of what we are going to do about developing higher-performance microprocessors, let's take a look at where we stand today.
Albert? Albert Yu, senior vice president in charge of microprocessor development. Albert, show us what we can do today.
* Other names and brands may be claimed as the property of others.