Windows Financial Analysts Summit 1997
Remarks by Bill Gates
Thursday, July 24, 1997
Seattle, WA

Good afternoon. Today's been very on schedule and certainly compared to last year in some ways kind of low key. We didn't bring in any military hardware to emphasize our competitive message, and we're keeping things rather straightforward. [Last year, a bomb scare caused the analyst meeting to be moved from a building on Elliott Bay to a nearby aircraft carrier, which was in town for Seattle’s annual summer festival.]

What I wanted to wrap the day up by talking about is just a little bit more of the product direction and look a little bit further out in the next fiscal year. But I wanted to start out with taking some of the product achievements of this year and putting them into perspective.

A key thing that is always exciting to me is to look at how we're doing in bringing in world-class experts, particularly in the new areas that we're doing work. And this means things like networking, securities, graphics, linguistics. And we had a fantastic year. We're certainly in the situation where the advanced R&D efforts we have, because of the caliber of people there, is just drawing in more high-caliber people. And we've become one of the few places that you can come knowing that not only will you be able to do the research, but it will get into products and then get out there and have a big impact.

I was very pleased with our product shipments over the last year. In any of the areas we were able to move even faster than we expected; so a lot of great deliveries with high quality.

In terms of the message we're sending to the world we've got two new themes that you heard a little bit about today from Steve and Pete and in some of the other speeches. For corporations we're talking about a digital nervous system. This is to get them to not think of PCs and networks as something that are just there and you have to fund them; rather to think very creatively by stepping back and saying, "How should the information in this company move; do we have the ability to call up everything about a customer and what's happened with that customer and get everybody who needs to be involved in making the decision quickly to have that information and let them collaborate together; can we make all the processes of running a company dramatically more efficient than they are today."

A lot of companies are making investments to buy the hardware and put those networks together, but they don't go the last ten percent to put the intranets and standards in place that really let them get this immense competitive value.

And so what we're going to be doing is really taking industry by industry examples of where people are doing that and therefore increasing the value that people see from the PC, getting people to understand the new information age. Decisions need to be made efficiently enough that using digital tools is really the only way to do that.

Now the analog for digital nervous systems brought down to the individual level is what we're calling "web lifestyle." Somebody who is comfortable enough with the PC and the Internet that whenever they're doing something that requires information they turn to the Internet, whether it's planning their weekend or buying a car or turning to get financial services or simply following the news or sports. It's going to take a long time before the majority of people are living that web lifestyle. But we've made a bet that that will happen and that people who get in early with technology and brands will have great positions for what will be a very, very significant market.

The key emphasis for us is on technology, although there are elements of that strategy that have content as well.

Our focus is really unchanged. It was about a year ago when we went from saying that our only big initiative was getting the Internet standards built into all of our products that we really brought in manageability as a top priority and then scalability as an additional item.

I think it's important for people to remember that we're really in four big businesses: The Windows business, which is the desktop operating system part; our Office productivity software; server software, which is Windows NT and Microsoft BackOffice; and the new activities, which we've now grouped all in Pete Higgin's Interactive Media Group, are our fourth area.

When people write in the press, "Is Microsoft becoming a media company?", it's always a little surprising to us because the first three businesses are growing very well, very profitable, proven businesses where our position in terms of the customer relationship and the excellence of the people we have doing that work make us a very strong incumbent. Our products will become obsolete, but we're in the strongest position to hear what customers are saying and deliver the improvements that they expect there.

So those first three businesses will continue to be three of the best businesses in the world. The third one is the smallest of the three, but it is the fastest growing, in a lot of new investments this year that relate to really getting that out there as rapidly as possible.

Interactive media is something we believe in very much, but the possibility here is over the next five years to grow that to be a business of similar size to those first three. That would really be the best case. And even when we look at what's inside there, technology is playing a very central role. Certainly the success of Expedia is based on that. And it's more and more of an emphasis of the things we're doing there.

So it's fair to say that Microsoft’s center of gravity is as a software company, although we're willing to mix content in, whether it's doing the games or doing the templates that go with Microsoft Word. Or perhaps in the extreme cases doing things like Encarta or Slate Magazine, which have both been very successful because we brought great people in to work on those properties.

As Microsoft takes on initiatives, these span all of our businesses. And we've been very pleased at how quickly the individual product groups respond to those initiatives. I'm going to talk a little bit about each one of them and where we are.

Just to complete the progress for '97 -- Office, a lot of big breakthroughs: Office '97 was a huge release. It was the one we are putting a huge amount of R&D effort into, even before we released Office 95. In terms of features and code changes it was substantially more than Windows 95 -- or Office 95. This is where we brought in the first of the things coming out of our research group. You know, year after year you've heard us talk about the increase in that pure research and how important it is. Now with Office 97 you're really starting to see a big payback.

The natural languages systems and grammar checking have an engine in there that comes from our linguistic group. Outlook was a new product. And we also changed the product line somewhat not only with the new release but having the Developer Edition and the Small Business Edition. The Small Business Edition has been very helpful for us in terms of differentiating the channels that we sell through. The normal Office Professional product more and more is a product that sells through the Select channels, through our licensing channel primarily, whereas the Small Business Edition used a broad set of channels, including sometimes letting hardware manufacturers include that as part of an offering.

The Internet area I'd say people have given us a lot of credit here, which is very gratifying, but I don't think they've really appreciated the wide range of products that are coming out. Some of the products we give away, and because of that they don't get the amount of visibility they deserve. Two of those really at the top of the list would be NetShow and NetMeeting. Now there are lots of companies that have adopted these products and are doing very exciting things with them. NetMeeting, of course, lets you share work on the PC, not only talk to each other, but also to run applications together. You can edit a document, manage a spreadsheet. And the companies that have caught on to this are using it in some very powerful scenarios. You don't have to use the Internet transport for the voice piece; you can set up a normal audio connection, or you continue to use NetMeeting for the document collaboration. This is a product we've been bringing out new versions of every six months. And it's achieving pretty incredible popularity.

NetShow is the product we use to broadcast video and audio information across the Internet, so you can watch things as they happen or go and get them later. For example, yesterday's events and today's events are being sent out on the Internet and will be available for people to get at those.

Now the only tricky thing with NetShow is you have to have pretty good networking architecture. And that's why a new partnership for us this year is working with Cisco. We both share a vision that network bandwidth will be going up a lot, but that there are missing pieces; directory being very important and so our cooperation with just one directory was a big milestone for us this year. We set up a multimedia laboratory that not only Cisco, but also Intel has helped to create, where we can show people what the scenarios are for NetShow and how you make sure that when you set that up you're not overloading your network.

On the web server side the popularity of IIS -- Internet Information Server -- has risen dramatically. We're actually not far away from releasing IIS 4.0 and it wasn't that long ago we put out 3.0. So it's a good illustration of how quickly we move. And each of these releases has very significant features. Now that we've got the popularity we're getting even more feedback from a variety of customers: corporate customers and ISP customers, the idea of making it easy for people to host using Windows NT where before it was mostly done on UNIX. It’s been a key focus and a lot of pretty amazing progress.

Index server, our string indexing capability, is simply built-in. Proxy server, the product that lets you access the Internet but protect your local area network from outside access. A lot of enhancements there.

That's always where you get caching capabilities. It means that if you look at how people are accessing the Internet inside your company, most of those hits will actually come from the cache; you won't have to go out to the Internet because it will be there because somebody else has gotten that information, and you saved that page on the proxy server. And you'll see more and more of that going on in the Internet at many levels because it's a key technique to have caching at many different levels. There's a great deal of rich caching built into the proxy server.

Site Server is a product that lets people monitor how their site's being used, look at the different paths, partly based on technology that we acquired and brought in a great team for. And that idea of tracking what's going on in your web site is going to become the most powerful marketing tool that people have, the ability to immediately see who's coming in, what they're doing, try out different product offerings, see how popular they are and make adjustments very, very rapidly.

FrontPage has done very well; it's now in by far the front position in terms of being a web authoring tool. And it's through FrontPage that we're able to very rapidly get innovations in HTML out into the marketplace and get people using those very rapidly. And that's important to us because we're trying to make HTML a lot better than it is today. And so FrontPage, Internet Explorer and IIS really go hand in hand in letting people offer rich things, letting people serve those up and letting people browse them. They're a product line that we do the architecture not only around the standards, but around new initiatives that we can drive through those.

Internet Explorer 4.0, we just put out recently the second preview, and we're moving to have that product out by the end of the year, which will be very important for us. What we've done with various media companies, the standards that's been established through the channel definition format is very exciting.

Moving our browser share up from about 0 to 30 percent is quite an achievement in a year. Really, it's kind of unheard of in the software industry for a leader in a category to go down, say, from an 80 percent to a 60 percent share. It doesn't happen very often, and it takes a very innovative product and a concerted effort to drive up ours to a 30 percent level. And we do have a belief that next year will also be a very good year for us in driving up even higher levels.

You've probably noticed our share is a little bit higher outside the United States than inside the United States. That is true for virtually every product Microsoft offers. Our competitors are relatively less competent at doing business outside the United States than inside the United States, so we always manage to do very well. In fact, the worst thing that we can say to a subsidiary is that their share is the same as it is in the United States.

In terms of Windows I've got a lot here, both things on the desktop and things that are out on the server. The Server Admin Kit: another free product, but an important product because it showed people that it's very easy to control what goes on in a PC. If you want somebody to run a browser or a limited set of applications, you can always have their desktop set up that way. It's very easy to do. So the PC can be a very rigid device or it can be a very flexible device, and it's just a few drag and drop operations against group needs for the administrator to control literally all the PCs in a company.

Transaction Server we think is a big breakthrough product. This is really a requirement in the world of the Internet where you've got distributed applications and we've learned a lot from the Transaction Server from the past. CICS is the most popular transaction server of all time and nothing in the mid-range ever came close to that. Now this modern view of the transaction manager really is going to be used by an incredible set of software developers. And that fits hand in hand with the message queue capability.

We did decide to make these products part of Windows NT, not just NT Enterprise Edition, but all copies of NT server and that's so that the message to developers is very, very strong.

We got the Windows 98 beat out; we've gotten very good feedback on that. A lot of customer feedback went into shaping all the things there, things like the Support Wizard so that if you ever do get into a problem we help you diagnose exactly what's going on and also get you the information that you need about what is going on.

NetPC was this year. People are now just recently shipping those machines. And they're a significant advance, supporting remote booting and giving people a flavor of the PC that's very straightforward that you never have to open up or add or take devices out of.

Before the end of the year two big things: the Enterprise Edition of Windows NT and the Small Business Server. Now this is a business strategy to be able to price appropriately for these markets and to integrate products together in a very powerful way. Small Business Server includes the ability to do messaging and database, and the Windows NT Enterprise Edition has features that most people wouldn't be interested in, like the three gigabyte support, but that provide extra performance for those high-end configurations.

And so taking a very powerful product, Windows NT, and having somewhat different flavors for the market, you've seen us do that with Office, with Windows, and you'll see a little bit more of that because it's been incredibly successful in letting us address particular customer needs.

Some of you came to our Scalability Day that was back in New York. I think that was a very successful event in really getting people to realize that you take (Moore’s Law and multiprocessing systems and clustering and great software and you put it together. It all is just a question of when PC architecture overwhelms all other computer architectures in terms of the performance that can be delivered.

There are many elements to being the primary operating system in the data center. Scalability is at the top of the list, but it's not the only one that we, together with third parties, are working towards. And you'll see month by month more announcements where people are actually taking these things we demonstrated and using them in production applications, people doing hundreds of millions of transactions. We have lots of customer sites now that are putting many thousands of users against exchange servers, in some cases over 10,000, or web sites that are getting lots of hits.

The only one of the four demonstrations from Scalability Day that used unreleased code was actually having the single-mode terabyte database against SQL Server. That actually did use an early version, the beta one version of SQL 7.0. But those other three are things that are shipping today.

And of course Windows NT file system is almost the only system -- most versions of UNIX do not support 2⁶⁴ storage size in files. And so the product we've been shipping for a long time has much better upper bounds there.

I think this message is not one that people just take to heart overnight; it's through examples in their industries, really seeing it being done that this becomes common sense. And I think it will become common sense a lot faster than most people are predicting. Some analysts said, "Boy, this will be three to four years before this is clear-cut," and I think we'll have fun proving that to be a wildly conservative view of the time involved.

One other thing that's underestimated is clustering. Putting systems together through clusters is the way to get great reliability and to get scalable performance. And we've had great support from third parties in what we're doing here, not only with our BackOffice applications, but with other applications. In the next few months we'll put out phase one of this, which is supported in the Enterprise Edition, which is the high availability cluster. And then next year we follow that up with even richer clustering capabilities that have been part of the architecture.

I said our top priority has been manageability, and Paul mentioned that over a year ago we really decided to dig into this topic. You'll notice that I don't call it Total Cost of Ownership because that's a term that in a sense doesn't really capture the whole phenomenon. Although it contains the word "total" when people talk about that they're mostly just talking about desktop costs, and there are desktop costs: the hardware, the software, the support, going out and visiting the machines, keeping them up to date and dealing with the variety that develops over time because of the local storage that's in the machine; but when you think of costs, I take off what's going on with the desktop as just part of the picture; you've also got data center costs where all the servers are; you've got application development costs and you've got network costs, what it takes to have a high-speed network that connects all these things together.

And you have to be very careful if you look at reducing one element of this equation not to just shove the expense over to the other part. So if you have a strategy that says rewrite all your applications, that would tend to drive up your application development. If you have a solution that says, "no, go back to time sharing, require all service to be on the server," that would dramatically boost your costs of your network and of your server. In fact, your server costs would go up much faster than what you would save on the desktop, and the users would lose the flexibility, portability and responsiveness. When you work across the network you've got latency and it's a time-share type approach that just isn't as rich.

And so we take the term "manageability" in a very broad sense. This is a case where we put together a group of customers to be a council to really guide us on this, and a group of customers who are willing to try out new things, like the Server Admin Kit, and who could give us feedback on the approaches. We quickly broke the effort down into what we needed to do with our hardware partners and what we needed to do ourselves in terms of setting new software standards. The hardware side moved very quickly. As I say, NetPCs are out today. And Windows-based terminals are available from third parties and we'll have an enhancement to Windows NT partly based on technology like Citrix that will support that kind of capability.

The software side is the most important element of this. This is where you can really eliminate the complexity. Thirty-two-bit Windows is the foundation for that. The Server Admin Kit, it's a management server, takes us a long way, but really the shipment of Windows NT 5 is the big, big milestone. And we made sure that we weren't going to ship NT 5 until it was viewed as a monumental breakthrough in manageability. And so even though the features like the directory that we've been working on for a long time are really mature and we could ship those even sooner, we decided manageability really needed to be there, so there's been some heavy work going on pulling that together.

I want to explain a little bit this Windows-based Terminal concept. When you think in computing you can have your storage on your desktop or your storage centralized and you can have your computing on your desktop or your computing centralized. And so you can mix and match those things.

The best responsiveness and flexibility comes from having your computing and storage locally. Hence the success of personal computers, which have overwhelmed every other flavor of computing and led to incredible industry building lots of applications.

But there are drawbacks to putting those things on the desktop. If you have a very light user you can actually move those things to the center. You give up the latency, but it works just fine. In fact, you can do this with very little hardware on the desktop. You don't need to have a disk or a browser or anything all that powerful. Even a 386 machine is more than adequate to take essentially a video protocol and display it on that machine.

What's even more interesting is that you can run the software that does that on any kind of computer, so whether it's a UNIX machine, an MS-DOS machine or an out-of-date Windows-based machine you can run that display software. And so it gives you the ability to run applications that require the latest capabilities and software releases without necessarily changing the client. You might start out just running the two applications that way and then if you're making heavy use you'd want to update the device itself.

The world will not move to this overwhelmingly because knowledge workers will always want the responsiveness. But if you take a very light knowledge worker, somebody who's worked against the terminal, this is an interesting approach.

The biggest response we've gotten on this is as a cross-platform solution. You may want to run Microsoft Office against a UNIX workstation. That's great; you just run it on the server and you get anything you want there. So it's remoting our full capability. And you can do it in a hybrid sense where you run some applications locally and some applications remotely.

Now here when I show NetPC that's the case where you've moved -- you've changed and all the computation comes locally. So we give people the ability to continue to use the old hardware and migrate whenever they want, and yet they'll have the applications using the new software base. So it's an aid in helping people make transitions and dealing with cross-platforms. It deals with cross-platform in the ultimate way, because you don't have to wait for the release of the other platform or see if it has different features or different bugs; you just run the mainstream high-volume release by having servers available to do that.

Now I mentioned that storage can be local or centralized. And this has been a big debate in the world of computing for all time. Where should the storage be. Having it be at the center gives you certain scale advantages and lets you back it up all the time. You also have people with the expertise there to manage it and to make sure it's not duplicated and it's handled right and updated when they choose.

So that's a nice thing. It also means that no matter where you go and get your computer, no matter how you hook in, as long as you hook in across the network your storage is there. So you can go to somebody else's machine; you can go to a machine that's in a hotel that you didn't have to bring with you and there it is, all of your storage.

Now, unfortunately the drawbacks are pretty substantial. And so if you look at what's happened with storage in the world of PCs, overwhelmingly it's been on the local desktop. That allows you to be portable, flexible; it lets you run when the network is down, and it's very nice. And the cost of disk as a percentage of the machine is very small, even with the rather fantastic storage sizes that are available today.

And so most people install applications locally and keep their files there, and have those benefits.

Well, there's no way that either one of these approaches are going to win out. They both are very valid. But what we've been able to do is what we're calling this IntelliMirror breakthrough is to give you the best of both worlds. And that means that your storage is local; you get the speed, and your storage is central; you get the backup. Even going portable can work.

So this is one of the bigger breakthroughs in the PC space in the last decade. We make sure we don't overload the network performance, and it works with all of today's applications. And so really it's the benefit for the PC with the advantage of centralized computing.

Now, yesterday in the Windows Day Paul showed off a little bit of this where he showed the ability to roam between machines. I'm going to show one other aspect today.

[Demonstration of IntelliMirror technology. A user wants to use the server as his primary storage area but also wants to be able to work on the document on his laptop away from the office. When the user disconnects from the network, the file is automatically copied to his laptop. The user edits the file offline. When the user reconnects to the network, the server is automatically synchronized to contain the updated file.]

MR. GATES: One key point about this is that it doesn't require a new operating system on the desktop. It works with any of the 32-bit platforms, including Windows 95. The place you do need to get the new operating system is up on the server. You need to have some NT 5 servers that actually do this management. But the client side of this thing will be made available for Windows 95, Windows 98, as well as Windows NT itself.

Well, let's now take a little competitive look here and talk about the PC and the empowerment it's given people. The list of people against the PC has grown shorter over the years, and you can see it's down to a fairly short list now. The list of people who have benefited by the PC model and are in there really driving it forward is much longer than I possibly could have put on this slide. But it is notable that there are still people who really are against the empowerment. That's the idea that the knowledge worker ought to be able to get a spreadsheet and pivot that and create their own documents, and that a department in a company if there are applications particularly targeted to what that department does they ought to be able to go out and get that and use it without going through some central process.

Sun is very pure in their belief that the PC is an evil thing and that it will soon go away. The only command anybody should be allowed to use is the backspace key, and all those nice looking documents are kind of in the way.

In the case of Oracle it's much more an attack on the notion of the richness that's there and could we start from scratch and do something. And, of course, they have their flavor of that.

I put IBM with a question mark. It's not really fair to put them purely in one column or the other. They sell more PCs than any company in the world with the exception of Compaq, and they do that quite well and have had increased success in market share there over the last couple of years. So that's a substantial thrust for them. But at the same time, at least at a press conference level, they're involved in the things that are moving away from the PC.

And so it's an interesting battle, you know, to see do people want to mix into their environments with PCs things that are very different and incompatible or do they want to rip out and say to somebody who's got a PC, "Hey, you're no longer going to have that kind of empowerment."

Of course this all goes under the phrase NC. And it's a little hard to know what it stands for, because does it have a disk, does it not have a disk, is it $200, is it $500, and whose software does it have. I mean, the software of Sun's upon these things is completely incompatible with what Oracle does and what IBM does. The Sun device only runs against the Sun server. So you're locked in at both the client and the server level. The application models and resources provided are very different. And so the only thing that really stands out is this non-compatibility, and non-compatibility with each other and the PC.

So it's an unusual approach because instead of being really pure and saying, "Look, let's move the computations to the center," which is what happens with the Windows-based Terminal, they go halfway; they don't want to let you be undocked because then they'd be like a PC, but they go ahead and put the browser piece, which is the biggest and fastest changing piece of software, on the client. And so that means in terms of the processor and memory, whatever you download is going to require more memory in the same way that a PC does. And yet you've got to have the servers there and you're going to be loading them dramatically more heavily than you would have before.

For example, without a disk you're paging across the network. Paging, when you've got a browser it's very necessary because you're bringing in add-ons, and anything less than at least 64 megs of memory you're going to page a heck of a lot while you're running a browser. You can bring up the instrumentation to watch any of these systems. Doing that across the network is a dramatic degradation, I mean to really an non-useful level, without the local storage. And, of course, you give up the flexibility in terms of the applications and things.

In the case of these NC pilot programs -- even where the people have showcased that, you know, maybe they've really got some interest, we're not seeing any adoption, except in a few cases where there were dedicated terminals before. And there, whenever we go in and explain the Windows-based Terminal concept we're able to do very well with that and get that to be chosen.

One other element that, coupled with this, although I'd say it's loosely coupled, is the notion of Java and what is Java. Obviously Sun's view and Microsoft's view are somewhat different. In our view Java's a great programming language. It will be used. You can write very generic applications that stick to a very modest run-time, just like you can in C or COBOL or any other high-level language. The reason they're called high-level language is that they're not tied to any instruction set or any operating system; they never have been. We've done a good job of getting our Java tools to be the most popular on any platform and very widespread adoption there.

But in contrast to the use of the language is this notion of the middleware operating system. And when you have two operating systems on top of each other, it is fairly complex. There is history on this. You know, we tried to put Windows sort of on top of OS/2, and it was really a mess because the character set model and security model and user interface model were different. And so the user would often have to think, "Well, am I running utilities that relate to Windows or am I running utilities that relate to OS/2. Am I dealing with the naming convention of the underlying operating system or the naming convention of the thing that's built on top?" And so scheduling, user interface; just even event management, utilities, all are very complex for somebody who's basically got two different operating systems. And there are size and speed things as well.

When you're shifting an operating system with the application it's very complex because if one application uses one version of the operating system and another application's written to the other one, you can't really install those against each other. And so it's like updating an operating system with the incompatibilities that can create every time you install so-called pure applications.

At some point somebody will step back and look at the economic model. You know, doing operating systems, the testing, development, even mundane things like support, device drivers, is certainly something that requires a massive investment, the evangelism and efforts that we've put into that.

Here, for the developers, they have to look at either exploiting Windows and looking like a good Windows application or running against the middle layer and then getting very modest additional places to run against. And the people creating the systems at some point will have to get a revenue stream. Well, whose revenue stream is it? Is it a Sun revenue stream? Is it a Netscape revenue stream? Is it an Oracle revenue stream? And exactly how does that work in terms of the work that needs to be done? Who can afford to do the investment?

So it's something we take very seriously, but something that has enough of an unique way of going about it that we don't think there will be a substantial impact at the operating system level.

A few words about Microsoft Research. I'm always talking about the great things there because it is an amazing group that we've been able to pull together. It's organized into a number of areas and I won't go through them all, but some relate to improving our software factories: tools, compilers, testing. And that's just "magic" software, making sure that we find bugs automatically, that we can write in more concise languages, that we have massive optimizers that take user scenarios and reorder the code in a way that can optimize for the user. And they make a big difference to us; things that don't require any man-hours, just tools that we're using.

Systems and architecture: This is a group that had a big impact on carrying Windows NT forward. There are some very interesting challenges when you get video and audio data types into operating systems, so-called "quality of service" guarantees or demand scheduling reach deep into the kernel of the operating system in terms of what the requirements are. And NT 5 takes a significant step in that direction. But having this research group really get out ahead and think about these issues has been very important to us.

The thing that will have the greatest impact is the area I'm calling natural interface. And both yesterday and today you're just seeing a little more concrete about the vision of how this will come in and be a standard part of the operating system. What we're saying is that the linguistic system won't be shipped with individual applications. It's complicated enough and you need to share it enough that it will be at the operating system level and it will be a substantial percentage of the code in the operating system. And that's from speech, handwriting, vision, and learning type software. Only by sharing it across that volume and as a standard can you really get that going.

We did announce a new site for Microsoft Research in Cambridge. We decided that luring all those researchers over to our main research site just wouldn't be possible, and so we found a great person to head that up, and we'll be hiring aggressively there. And so that's simply going to let us tackle some additional areas of cryptography, network performance and some things that we wouldn't have gotten to otherwise in nearly that same fashion.

Speech is the thing that people seem most fascinated with now because you can actually go out and buy some of these commercial products and they work pretty well. And if you just imagine Moore's Law giving you eight times the speed, eight times the storage and memory, they would work even better. So we're getting towards the threshold where people are taking this very seriously.

There are a number of types of applications: command and control dictation and conversational interface. And some of these things are small vocabulary; some of them require training. So they're a little bit hard to compare. But everyone's technology is getting a lot better. We feel we're certainly state of the art in this. The thing that we're unique in is figuring out the interface that puts all the pieces together and how we're going to be able to go out to application developers and let them describe the context in their applications so that these probabilistic input systems can do the disambiguation. And certainly vision, speech, handwriting are the best examples of these probabilistic input systems.

When we think about linguistics, that's another one where there's many levels of expertise. And I show four examples here. First was simply spell checking, catching "t-e-h" versus "t-h-e-." That was very widespread in word processors six or seven years ago.

The next is grammar checking, understanding the matches of singulars and plurals and saying that each of the products is designed to help. That's state of the art today, although it's still very difficult for long and complex sentences to make sure you really are seeing a grammar error and it's not just that your word form list isn't totally complete, so you’re mis-parsing a sentence.

With the experience we've had with the grammar checker in Word, we're just constantly using that user feedback, and so the next big release of Office is a huge step forward not only in terms of English grammar checking, but starting to bring in other languages. You need a general architecture framework to deal with lots of languages here, but then you need specific work to approach all the world's languages, both in speech, handwriting and linguistics; that's true.

The third thing is understanding meanings, and getting the parsing right based on common sense. "I saw the Grand Canyon flying to Arizona." Well, that could mean two things, and it makes a big difference which of those, you know, in terms of whether you should be surprised or not surprised which way you happen to parse that. And you might suggest to the person that they say: "Flying to Arizona, I saw the Grand Canyon," which makes it a little less ambiguous.

And then you get into understanding deep semantics, which is what you have to have to be able to do real understanding in translation. And so understanding that a sentence that related to "pregnant women are not supposed to be in bars" that a translation like this doesn't make sense, you actually have to have someone who thinks in terms about language. And so that's the most ambitious to be able to flag that and translate it in a very different way. And we do expect to achieve that level; I mean, that's what this group is all about, but we'll productize at each level. And we have a plan that rolls out in its grand glory over the next five years.

Using this you can build conversational systems and you can do rich information access. Yesterday we showed how a search capability that's just based on string matching is grossly inferior to one that's based on linguistic parsing.

Of all the systems I've talked about, I think the one that may surprise people the most -- and it's really a fun one -- is vision. The cameras that people are going to be using for videoconferencing are getting very inexpensive. And we can take the processing power of the machine and couple it with that camera and do some very interesting things. And I've asked Matthew Turk from Microsoft Research to give us a little glimpse of where he is in some of this work and how vision will fit in as an integral part of the natural interface.

MR. TURK: The goal of many of us in Microsoft Research is to make interaction between people and computers much more natural and compelling. And the research goal of my group is particularly the visual part of that, to have computers watching people and understanding things such as: is anybody there; if so, who are they, where are they, where are they looking, what are they doing, what are their expressions, what are their gestures, etc.

So what I'm going to do is show you a few little examples of how you might use this kind of technology and little demos, but extrapolate and imagine how you can apply these things to computer games, to uses for disabled people, to all sorts of things.

[Demonstration shows that the computer recognizes that a person steps in front of it. The computer also recognizes a gesture that turns it off. The computer then is "trained" to play a game based on positions of the user’s head. The computer then plays the cymbal every time the user claps his hand.]

MR. GATES: Well, I didn't want to end without introducing my own dose of sobriety, since we all distributed that task very well this year rather than simply ask Steve to do it all by himself like he has too many years. There are some big challenges ahead in the business model, certainly in the costs. What we've done with the stock-option program is created an unprecedented amount of employee wealth, and certainly in terms of absolute, in terms of what that means and in terms of people making relative comparisons against that it puts pressure against the compensation model. Greg mentioned this. I'm not sure it's a significant issue in the next year, even two years, maybe not even in three years, but over time it is a significant issue that we need to think creatively about and could change the P&L somewhat.

Also we have been growing R&D costs very rapidly, and our plan for the next year was mentioned is to grow them around 25 percent to a little over $2.6 billion. And that's well ahead of what we're looking at in terms of sales increases. And so it is very interesting that we've chosen, partly because of the kind of management we are, we've chosen to make these investments in innovation. And in the long run you can't have the R&D expense outrunning any of the other items. So there's a real question of can we get to a point where we bring that growth well down. I'm not certain. A few years ago I thought we'd be bringing it down by now, but as we've found new areas, things that we think are important we do, we continue to grow quite aggressively.

In fact, the only thing that's really held us back there is our ability to hire. We did very well in that last year, but we were below what we had budgeted for R&D simply because of that challenge.

In terms of the top line thinking about market share possibility, we talked about saturation. There's the business challenge with the interactive brands. And so we're moving into these new low-end devices, are making quite an investment. The acquisition of WebTV is part of that. And so these are new markets: digital TV and handhelds, which today aren't very large, and yet in terms of the kinds of royalties you can get and in terms of the kinds of volumes that come out of them, they're challenging areas.

I do want to temper this a little bit by sharing some more thoughts about where we're going. I feel very good about what hardware is going to enable us to do. Moore’s Law for processors is a pretty safe bet for the next decade. I'm constantly impressed with the work that Intel is doing. They have competitors that are nipping at their heels, but they're staying ahead very well, driving up the Pentium, getting multimedia into that, really pushing the clock speeds up, bringing the prices down so they get into the mainstream.

We have a major cooperative effort with Intel on Merced, the 64-bit Intel chip. It's one of the many things we're doing with them on manageability, graphics, a lot of things related to multimedia. But Merced is one that you won't see in the marketplace for a little while, but a huge investment for them and a significant effort for us as well.

The goal is simply to have NT taking advantage of everything that you can do with Merced in both server and desktop applications when that comes out. And so the enablement to let software do new things, including natural interface through the chip, that's certainly going to be there.

Better flat screens, a lot of excellent progress this year. In fact, in my desktop display nowadays I just use a 15-inch LCD. In a few months I'll have a 20-inch LCD, although they're still fairly pricey; that certainly will come down.

Storage is important, because when you start caching a lot of information, you want to be able to include things like audio, video type data that require a lot of storage, and the improvement there has been even better than Moore's Law rates.

The digital peripherals are really coming along. Those will drive certainly both the home and commercial markets. Of all the things on this list, the one that is the toughest to predict is the high-speed connections. Being limited to 28.8K or even 56K doesn't enable the full-blown web lifestyle. If you want to have video data rates which, properly configured, cable modems can provide that on a two-way basis. We have now many hundreds of our employees hooked up to ADSL at six megabytes. It's a trial we're doing with GTE. And it's really phenomenal how they end up using their machines in a different way once you have that connection.

They key thing to understand is that with a high-speed connection you can from the center really guarantee a great experience for the person at the other end, because you keep your software up to date. If they ever have a problem you see it immediately and can take over and help them out with it. And all sorts of issues they would have to think about like managing their disk; is the disk full; what is this problem, all of those can be handled up at the center. So it's really a different world.

You're seeing a tiny glimpse of that with WebTV today where it's a managed device, managed over that phone connection. The software is updated. And we see every advantage. We can know whether somebody has been getting busy signals because that's all logged; has somebody been getting error messages; is there any problem; is somebody not dialing in. And so management is the big thing.

Certainly there's a lot we can do in user interface for ease of use, but managed from the center is a part that I think is greatly underestimated and it is just perfect once you get that high-speed connection.

We tend to be a little more optimistic, I think, than most people about how fast these high-speed connections roll out. Hopefully that's because we're more insiders in terms of getting involved and really putting the pieces together and hopefully for other people it's a little bit their emphasis is mapped to the interactive TV; overoptimistic predictions or still sort of getting over that.

This is the reason that we chose to invest in Comcast so they would be a partner in helping us roll this out and see what could be done. We're not in their business; we simply have a very strong relationship with them and an investment in them that we hope will continue to appreciate in value.

Part of that really is not just connecting PCs, but intelligent TVs, going even beyond Web TV to dealing with the digital video feeds. This is an area where you have a lot of companies who want to contribute, set-top box companies and some electronic companies.

Microsoft's coming at it from a software angle, and with some of the expertise through WebTV. And it's just developing. There's a lot of jockeying around. Even basic things like the video format, 1080i, 720p, you'll hear things like that. It's not actually a very big deal. I mean, we need to get these things in place so that there's a bootstrap, a reason for people to transmit high-resolution video. And it's not there yet because the displays and transmission systems aren't there. And to get those to critical mass there must be an economic incentive; a very difficult thing.

But I feel sure that through the collaboration with these industries, even though some of it appears like a standards' battle, that we're going to figure out the computer industry, the display industry, cable, industry TV and industry how to get high resolution out there and make sure that the PC can participate in that as a high-end device.

Portable devices. We really jumped into this late last year. Around the end of this year we'll have second-generation devices where our partners are doing even more there. And you'll start to see the pace of improvements, the size of the screen, the resolution of the screen, and glimpses of the digital wireless that makes that become real.

Windows CE has been unbelievably successful for us. We have a hard time keeping up with all the people who are coming in and asking us to use Windows CE in various products. I wouldn't say it's 80-million units multiplied by a Windows-type royalty, but it is exciting to see that the interest is there for an incredible range of devices.

I just wanted to close talking a tiny bit about the future of Windows, and it's the same general framework as the future of Office and our server software as well as content delivery. We see for pure software products continuing with the quarterly service packs. Those have been very successful. We're actually getting the size of those down a lot. They can be delivered over the web more easily. We will go to yearly releases, and so that's actually less releases than we've been having where we've been rolling out a number of features, sort of asynchronously you'd say. Yearly means things like the updates of Enterprise, Small Business Server, all get synched in to that once a year cycle. And probably only the release every two years will be a major release. But that will vary somewhat by product.

The Internet will be the primary delivery vehicle. Customers over time will be much more working this on an annuity basis. And that's why when you have all that discussion about the revenue model, you know certainly that idea of thinking of it more like a magazine where somebody signs up for 12 months to get all the bits we can send, all the support we can send and helping them out, customize what we deliver based on what their hardware and usage profile is, we see that as the future of the business. And so moving to 80/20 to that type of revenue recognition is just a step towards where we see things.

We are working on some big, big advances in Windows itself. We are moving to use Windows NT as our base technology. As we move to do very advanced things, evolving those on two different kernels would be prohibitive. And we're not going to do anything unnatural here, but eventually we see even the consumer part of the market by a flavor of NT. We're going to build onto NT what we call integrated storage where users don't have to think about address books and message stores and registries and databases and file systems and web stores as differently as they do today. They're all very complex in terms of their naming schemes, their navigation schemes, their maintenance schemes, their replication schemes, security schemes. It is way too complicated. And even if you just take Microsoft, there are really ten different storage systems that in some way users get exposed to. We'll bring that down to essentially one integrated storage system, although underneath you'll have intelligent record management and screen management -- screen management coming from the evolution of the NTFS file system. That's a breakthrough thing that's never been done in operating systems, and it's very important for simplification.

Windows "dial-tone" is this concept of using the high-speed connection so that that user is always taken care of, that even though their computation is local, logically where they're getting their software updates and help comes through a connection that is always there and the expertise is really at the center.

Using security to get rid of viruses. Going the full way in browser integration. And really going to a scenario-based UI. In the UI that we have today it's very difficult to map what you're trying to do onto the series of clicks to navigate through existing structures. The file system is too naked of a way, where you see all these files; you don't know what they are, you shouldn't really have to look at. You know, what you're trying to do is really very simple. Most people have a modest number of documents, a modest number of applications, a modest number of things that they want to step through. And both letting people do those things and helping them to deal with any problems that come up along the way can be dramatically more scenario-based. And several groups at Microsoft have gone down this path in prototypes and things which over a period of time, over the next several years certainly we see building into the operating system.

I'd put all those in a group, because I see deepening relationships between those things. And none of them requires a big breakthrough in terms of natural interface. And so those are very assured type things. The timeframe may be not so sure, but the fact that we'll do those and do those as a group requires no additional breakthrough.

We’re working on the natural interface in parallel, although I see the completion time as being significantly later and seeing it being built on top of these other things. So in a sense you can think of this as one big generation and then a big generation that comes after that. This is dependent on those investments and research, having linguistics, speech, handwriting, personalization and even see this ability to scan documents and call them up easily so that it's easier to work with your computer to store things than it is to file them in a paper way. Two of the ideas there: handwriting recognition and paperless office are currently discredited ideas. We love working on those ideas because other people really aren't investing in them even though the reason that they didn't succeed had more to do with the quality of the implementation than the fact that the basic concept wasn't sound, and very, very important and very exciting.

And so I'd just end on somewhat of an upbeat note in terms of where we see this business long-term. You know, we want to be very frank in sharing how we as the management are good at worrying about things and seeing those challenges in both the directional and a numeric fashion. But if you think of the importance of software and what it means as the world moves to a digital nervous system or web lifestyle, there's no doubt in my mind that software will be increasingly valuable and Microsoft is well positioned to take advantage of that.

Thank you.