SIA Technology Management Conference
New York, NY, USA
June 19, 2002
Download presentation (3.8 MB Zip file)
ANNOUNCER: Ladies and gentlemen, please welcome president and chief operating officer, Intel Corporation, Paul Otellini.
PAUL OTELLINI: Thank you, and good afternoon. Before I start, I couldn't help but think, as John McKinley opened his presentation, that perhaps our branding campaigns have gone a little too far. The day that the CTO of one of the world's largest financial services organization confuses the world's fastest four-way server with an Intel employee, we may have crossed the line.
PAUL OTELLINI: In any event, I'm happy to be here, and I thank the Securities Industry Association for the invitation to speak today to a room full of customers. And we don't see a lot of customers nowadays. Too few. Fewer than we would like, certainly.
You may be surprised by my use of the word "customers" in the context, because I doubt that any of you have ever written a check as an IT manager to Intel Corporation. You've certainly bought our products through computer manufacturers.
But what I want to talk to you about today isn't just the products you buy from us, the "Intel Inside" chips, but a lot more about the ecosystem that we built around those chips to try to bring you compelling solutions to run your businesses, particularly in the financial services industry.
So today I'd like to cover three things: Our view of the Wall Street IT environment. And it's not inconsistent with the views that you heard from the prior few speakers, but I think our slant may be a little bit different than theirs.
Secondly, talk a little bit about the worldwide trends in computing. And thirdly, lastly, end up scaling with Intel in terms of how you can use Intel-based hardware and solutions to drive productivity into your businesses.
I'm going to begin with the overarching problem that I think those of us who manage IT spending have to deal with, however you cut it in any industry today, is that IT spending is being squeezed.
After a good run-up in the late '90s, principally the Y2K environment and dotcom environment, IT spending has declined the last two years. And Mr. Wiser, two speakers ago, showed you a couple of points, but he left some points off the same curve.
The last point he showed was '01, then he skipped to '04 and '06. And if you don't see the stuff in between, you never notice the decline.
In fact, the spending has come down. It peaked two years ago. It's come down the last couple of years. And I think it's very generously optimistic to consider that IT spending will jump up again in the next couple of years. I think the pressure is going to remain on all IT budgets, not just in the financial service industries but everywhere, to hold flat and drive productivity into their organizations and around the corporations.
This is spending for U.S.-based securities industry firms.
Another way to look at this is on the headcount side. This is IT staffing for security industries on a worldwide basis. And you can see that it grew very rapidly through the late '90s and peaked at around 71,000 by the year 2000, and then took a pretty dramatic drop the last couple of years, dropping almost 7 percent in total employment. So you're being asked to do a lot more with a lot fewer heads.
Inside of that headcount today, Intel surveyed the top 10 firms on the Street, and we tried to find out what the 24,000 developers in those 10 firms were doing.
Five years ago, over 90 percent of those developers were developing code on RISC-based machines, principally Sun machines, to be deployed in a principally RISC-based environment.
Last year, over 50 percent of those developers were writing code, developing code on Intel-based machines to be deployed in a mixed environment. I think going forward, that trend will only continue in terms of using much more cost effective hardware to do the development and increasingly cost effective hardware to do the deployment of that code into a multitude of operating environments.
The net result of the combination of the squeeze on spending and the decline in resources and the complexity of the tasks that IT in your industry is being asked to do is really a set of diverging curves.
And the diverging curves are driven by a number of things. It's driven by the transactions that the prior two speakers described in terms of the number of sheer transactions and the complexity of those transactions not declining as the markets go through perturbations, but also the regulatory environments are going to cause more and more work for the IT shops that are represented in this conference.
So you have a divergence coming up, and clearly a problem.
The complexity of the work is also changing. You folks know much better than we do in Silicon Valley, but the regulatory impacts, the removals of laws going back to the 1929 time frame, the requirements to do one day clearing of transactions are changing your business.
Globalization, having to be able to settle things on a worldwide basis, consolidation, the mergers in the industry, the changing structure of the industry, the changing structure inside of firms in terms of how certain divisions of firms relate to other divisions is forcing you to change the way you do work. And of course, all the other aspects of this that are part of the impact in terms of perhaps not just the wealth creation side but nowadays, the wealth preservation side of your client base, is requiring you to deploy new services to satisfy them.
Batch mode in terms of communications doesn't work anymore. Everyone is requiring real time data to all of their service desks worldwide. And of course, everyone has legacies and they have to migrate out of them.
So quite frankly, when I think about our job of designing chips, which are probably the most complicated machines that man has ever built, that task sometimes seems daunting compared to the task that you have to do in terms of the sheer complexity of the tasks that are being asked by your firms for you to do every day.
So in that context, let me talk a little bit about the server market and some of the worldwide trends that we see coming up. There are really two mega trends from my perspective.
One is the migration from proprietary, RISC-based, UNIX environments to platforms built on hardware and software standards.
The slide here is from IDC, and it looks at the cost per thousand users of a typical RISC UNIX system, and they happen to compare it here to an Intel-based system running Linux, but you'd get a comparable result if you were looking at an Intel system running a Windows environment.
Almost a drop of 50 percent in the total cost between the hardware support, the software costs, and the overall cost of operations of these machines.
And we know at Intel, as we migrate to this kind of hardware from proprietary systems, our engineering systems have been RISC-based systems for quite some time. In the last couple of years we've moved them to Intel-based systems. We know that not only do we save on the hardware and software, but as we can scale that out over time, the support costs also drop, which is not yet reflected in the study.
So I think, number one is the compelling cost economics associated with the move from a proprietary environment.
This is not necessarily architectural. It's economic. The example I want to draw here is Apple. And it's not a perfect analogy to what's happening in the workstation back office world, but I think it's close enough to make the point.
These are really two sets of curves stacked on top of each other. The top one, if you can't see from the back, represents the share of market that Intel-based desktops have from 1979 through today, and the blue line is the Apple market share. You can see before the IBM PC was introduced, Intel-based machines had about 20 percent of the market and Apple had about 20 percent of the market, and a bunch of other architectures had the rest.
As the PC came out, it created a dynamic that standards beget volume, and volume begets standards, and it reinforced itself over successive generations some 20 odd years now to be able to drive to the point where the sheer scale economics of the PC make it increasingly difficult for a company like Apple to compete. The software isn't there. The software is now developed for the PC first. They have the same kinds of costs to be able to develop platforms and new architectures that the PC does, but they amortize them over far fewer units.
Well, that same impact of scale economics is at work with respect to some of the proprietary architectures that are out here. The one I'm going to pick on today happens to be Sun, just because they're near and dear to our hearts.
This is the same curve 10 years later, and I've plotted servers. In yellow is Intel and in pink is Sun. The server market segment share from 1989 to today. And you see that same dynamic, where Sun has sort of a low single digit share of the server market in terms of units.
These scale economics cannot be denied. The scale and the volumes that a standard drives attract software. The software attracts more people, it attracts diversity, it lowers cost, it increases the investment in the architecture, and the cycle continues over and over again.
Many of you and your peers are making increasingly strategic bets based upon Intel server architecture. This chart looks at all of the servers sold from 1996 through 2001. It's a projection through 2006, so a 10 year period, and we're six years through that already today.
It represents all server systems two way and above. And the little bar on the bottom is RISC-based systems. The purple curve is Intel-based systems.
So not only have we moved out of the simple file and print sharing. We are now a major force in the 2-way servers, representing some 70+ percent of the unit share of the market. Increasingly with our architectures, the Xeon-based architectures, and, coming soon, the Itanium 2-based machines, you're going to see Intel machines move farther and farther up the food chain into the eight and above marketplace.
And this is IDC's data, it's not Intel's, but it shows that lead continuing to grow over the next few years.
So trend one is this move to standard platforms, this move to standard operating environments, and away from proprietary systems.
Trend two that really makes this all work is what I call the horizontalization of the industry. And this is, again, not a new trend. It's a trend that came from the PC industry.
If you think back to when people before the PC, the industry, in terms of computing, was represented by companies who had complete vertically integrated stacks. They produce their own chips, they produce their own software, they produce their own platforms, they produce their own applications, et cetera, et cetera. And they competed with each other by taking their stack against the next company's stack. The last two speakers talked about the bunch. The bunch competed in that fashion.
The same phenomena that disrupted that industry in terms of the PC on the desktop is about to be at work in the back end. And that really has to do with the fact that companies can be much more effective when they compete. Not vertical stack against vertical stack, but when you as a user can start looking at the horizontal elements of computing and picking the best of breed of each of those elements and combining those best of breed, best of choice, best of cost, best of performance, best of flexibility, to create an environment that is best for you as a user.
And we're seeing that today. There's a new horizontal industry being created for financial services. You have a number of OEMs participating in that, some of whom are very familiar to you. They sold you some of the proprietary systems that you purchased for many, many years. Companies like IBM and Hewlett Packard.
But you also have new players like Egenera, which is a startup company here in the East Coast, Dell, and many, many others, all competing to win your hardware business with relatively the same architectural-based products based upon Intel processors, but their value-add is at a different layer. It's at their chipsets or the way they do their I/O subsystems and so forth to get differential performance and capabilities for you as a user.
Operating system competition is another layer. And while, on the desktop, there may not be a lot of differentiation today, you certainly see a lot of activity going on in the operating system choice in this area.
As people move from principally a UNIX environment, they have many choices. They can choose to go to Linux, they can choose to go to Windows, they can choose to go to Intel-based flavors of their previously proprietary operating systems on UNIX.
Again, at the application layer, many new players coming in here. I'm showing some of the logos, all targeting their software to run on Intel, to run on this multitude of standard platforms, to be able to give you, as IT buyers, the ultimate in choice.
So you add all this together, and what we see, from our perspective, is a very strong degree of momentum. The movement really is happening here.
We've picked a couple of quotes. The top one is from Credit Suisse First Boston where they replaced a four-way RISC-based machine with an Egenera machine, and subsequently received 20 times the performance at 25 percent lower cost. Those kinds of economics are very compelling.
There's a number of other examples here. Chris Conde, who is here at the conference, is president of SunGard, is very excited about the new Itanium 2 chip coming out soon. He talks about this as being an inflection point, an inflection point that, in his mind, is analogous to the 386; one that really brought high performance platforms out to the desktop.
So you have to ask yourself why is all this happening. At least we asked ourselves why is all this happening. And I think this is where there's a different Intel at work here.
I think most of you look at Intel or view Intel as a chip company. And in fact, that's what we are from a revenue standpoint. Most of our revenue is from chips. Most of our R&D is from chips. All of our manufacturing is principally focused on building chips.
But in order to make those chips compelling to you, we have to do much more work than we used to do, particularly as we enter into new markets and move farther and farther up the computing food chain.
We do, and have done for a number of years, a lot of work in platform enabling. Many of the bus architectures that you buy inside your computers, things like PCI and InfiniBand, were principally architected by Intel; involved in the specifications, involved in the deployment of those going back almost 20 years now in terms of a variety of architectures.
But in the last couple of years, we added two more layers onto our stack. The first was a couple of years ago, a very significant move where we put roughly 2,000 engineers on the job of solutions enabling.
As we moved farther up into the enterprise, we recognized that the key limiter wasn't the hardware availability, but, rather, the software availability. And it wasn't just operating systems. It was making sure that each of the applications that all of you need for your business transactions not only were ported to Intel, but were optimized so they ran fastest on Intel.
These engineers, these 2,000 plus engineers in over 100 solution centers or porting centers around the world that we operate independently and with our customers, are out there doing the porting job day in and day out, moving the stacks. We work on ISV stacks in terms of standard software. We also work with our OEM partners to move proprietary stacks over in the case of a lot of people, particularly here on the Street, where some of that software is home grown. Sometimes we need to help move it over and optimize it and help you understand the innards of our architecture to be able to make that happen.
Then the last layer we added is the channel activity where we've done quite a bit of work not just marketing but engaging directly with some of the key partners in your industry that we hadn't engaged with prior to a couple of years ago.
On one extreme, this is the Oracles of the world, the people who provide the databases, the SQLs, the DB2s, but also the CRM folks and, increasingly, the solutions providers unique to your industry, people like SunGard and a number of the people that are out here today.
So Intel is spending roughing 10 percent of our effort in terms of computing in this area of creating the solutions, enabling the solutions, and working with our channel partners to make sure that we have a free flow from the chip to the end user in terms of being able to deploy product.
The other thing that is coming out as these solutions become deployed is their very compelling price/performance characteristics.
This is a chart from Accenture. It's actually up on their Web site. It's from January of this year. In this analysis that they did, they compared three environments, machine environments, all running the same version of SAP. I think it was 4.0b. And the first one they compared at the time was one of the high-end Sun machines, a UE 10,000, 34 processor machine, to a Unisys machine running Windows and SQL. It was a 24 processor machine, and a Compaq eight-way machine that was also running Windows and SQL.
What they found is the response time or the performance was substantially better on the Intel-based machines. In fact, the Unisys and the Compaq machines, even though they were much different in terms of their overall configuration, had about the same response time. And the overall cost, as a percent of the UNIX machine, was substantially lower.
So think of it, in this environment, what Accenture found was a 30 percent higher performance at 20 percent of the cost. Not a 20 percent discount; an 80 percent discount. 20 percent of the cost. A greater than 10x factor of improvement in price/performance.
It's not an isolated case. If you pop open the TPC database today, you'll see that 10 out of 10 leaders in terms of price/performance are Intel based, and six out of 10 in terms of just raw performance are Intel based. And these are all still with our older generation hardware. As newer versions of Xeon and newer versions of Itanium come out this year, I fully expect this to improve and do us well on the right hand side of this page as we've done on the left hand side of the page.
Now, the proof is kind of in the pudding here, and maybe to give you a more tangible view of what I mean by the high performance computing environment, we could give you a demo. Let me bring Diana up to tell you a little bit about Itanium 2. Hi, Diana.
DIANA: So Reuters has developed an adaptive platform to build fast, reliable trading solutions quickly and correctly. They call it Structured Negotiation Web Services Technology. And the really exciting part for us is that today we're showing this running on Intel architecture, specifically the Itanium processor family, for the very first time. Reuters is planning to put the Itanium solution into production by the end of this year.
Now, on the screen over here, you can see the Reuters dealing for swap service. It's designed for OIS or Overnight Index Swaps, and it's an example of the kinds of applications that you can develop with the Reuters Web services framework and the Itanium processor family.
Reuters was looking for a platform that could deliver an order of magnitude increase in performance in order to allow them to scale this core technology across multiple products.
They're using Itanium 2 processor-based servers to give them the power and the performance they need at an affordable price.
Reuters is running a pre-release version of windows.net enterprise server, and they're also currently running Itanium 2 processor-based server from HP, like the one we have here. It features HP's ZX-1 chipset and four Itanium 2 processors. These processors are currently running at a production speed of 1 GHz apiece.
So rather than just keep talking about it, let me show you a little bit more about the application.
This is the Dealing for Swaps home page. You can see there are three major sections: negotiations, these are ongoing negotiations among the traders; review, these are negotiations that have been agreed to between two traders but not yet confirmed; and this is the market interest board, essentially a messaging board for the traders.
So I'm going to pretend I'm a trader for a minute, and I'm going to try to get 25 million Euros. So I'll show you how easy that is to do it. I just fill out a quick form here, Euros. Okay. We'll pick the currency. Yeah, six-month term, that sounds good. 25, so that's in millions. And I'll just send a quick message here to say what I'm looking for. Looking for 25 million Euros, and I'm going to pretend to be Paul Otellini. Paul is going to be sorry he let me have access to his financial system.
And I'm just going to send this off. This is going to go to all the traders out in the system. They'll see it real time, and they'll get a chance to respond.
Now, this is just one example of a transaction, but Reuters can actually handle thousands of similar exchanges simultaneously, and that's due to the computing power of the Itanium 2, letting Reuters manage these complex SPML-based transactions safely, securely, and in real time, which is obviously a must for this kind of application.
Oh, I received a response already. Check that out. Let's see. Hope I'm going to get my 25 million Euros. Requesting data. All right. "Paul, I will loan 25 million to you for the specified rate. Andy." Hey, it's some guy named Andy. That sounds good. It's a deal.
Then all I need to do here is agree. So go back out to Andy, and as soon as he comes back and agrees, it should show up here as dealer agree. Just hit "confirm," and it will go off, and it will shortly be appearing in this review section as a confirmed deal.
So you can see how easy it is to do swaps using this Reuters system. And it's also a great example of how the Itanium 2 processor can handle commercial business critical applications. Plus with the Itanium processor family, you have a choice of operating systems, of solution providers, the kind of choice that you just don't get with other architectures.
You put it all together, the price, the performance, the choice, and you can only conclude that the Itanium processor family gives you the best solution in the market today. Thank you.
PAUL OTELLINI: Thanks, Diana.
PAUL OTELLINI: What Diana showed there was a real live application. It wasn't running on this machine. It was actually running on the Reuters server, and we were actually accessing the Reuters server real time there.
To summarize the kind of performance that Reuters found as they created this environment, we've had benchmarks run, on the same application over three different hardware configurations.
The first one was the top of the line Sun UltraSPARC III machine with eight processors. The second one was a four processor Itanium system, the first version of Itanium. And the second one was the Itanium 2 system, the four-way machine here, that happens to be by HP in this case.
What we found on the Reuters application was that the performance between the Sun eight-way machine and the original Itanium machine was about 2x. That's 2x with half the processors and $45,000 versus the $230,000 machine.
As it moves to Itanium 2, the performance goes up again by almost a factor of 2. So a factor of four from the top of the line Sun; four times performance. Price isn't going to change. The price for the Itanium 2 systems from HP is expected to be around $50,000, sort of the same as it was for the Itanium system.
So you have this very compelling price economics paradigm continuing to work, and the software compatibility, the same software that ran on Itanium runs on Itanium 2. It doesn't require recompilation or re-optimization.
And in fact, this is the second generation of the product line. It will be going to production very, very quickly. But we are already working on the generation beyond that. In fact, we're working on multiple generations, but we have silicon on the generation beyond Itanium 2, which we've code named Madison.
Madison is probably the most complicated semiconductor device or chip ever built. It's a half a billion transistors on one die. We have working silicon today in-house. This is simply taking the Itanium 2 chip that's in these kinds of processors and moving it to the next-generation silicon technology, the next highest performing silicon technology that's out there that we are putting in mass production today.
Beyond this Madison chip, we have four more generations in development. So this is an area where Intel is using its R&D muscle to be able to continue to out-invest the industry, to take the elements of the architecture and continue to refine them generation after generation after generation, and bring them to mass production the way that only Intel can do.
In fact, in terms of being able to take advantage of the R&D budgets, this really is one of the more compelling stories for you. Between last year and this year, 2001 and 2002, Intel will invest $13 billion. So capital equipment to build new fab plants, new wafer fabrication plants, $13 billion. And spending in excess of $4 billion each year on R&D.
So think of this as a $10 billion a year investment in your business, not so much in ours but in your business. We are doing the basic research, the basic factory networks to be able to build the products that you will end up being able to take advantage of over the next few years.
So my summary to you today is really just three messages. One, we certainly share your pain. Having managed a number of IT shops in my life, we are continuing to ask you and people like you to do more with less. You will not get the luxury of more and more resources, I think, for a number of years. You will have to find ways to be much more productive.
In the history of computing, there's only one way to really become more productive, and that's to embrace standards. If, in fact, IT is your competitive differentiator, and I believe it is in your industry, standards give you that base level capability on which you can add value.
Using your value-add to replicate things that are done in standards is a waste of time and money. And so you move up the food chain, move your value-add above where it has been in the past.
I think, quite simply, you would be missing a lot to not take advantage of the fact that we are out-investing the industry, and in fact, investing about $10 billion a year on your behalf.
With that, thanks very much, and enjoy the rest of the show.
* Other names and brands may be claimed as the property of others.