Remarks by Bill Gates
Microsoft Corporation
Tech-Ed '98
June 4, 1998
New Orleans

MR. GATES: Well, good afternoon. I'm really pleased to have a chance to share with you some of the future directions that Microsoft is taking. This is a very exciting time in the world of software and hardware, and we're in a wonderful position to take your feedback and drive the products forward.

I heard from Steve Ballmer there's not only a large group there, but a very energetic group that's done a great job asking us questions and giving us feedback about our products, which is exactly what Tech-Ed is all about. So, it's impressive to us the sign-ups we get. It's always tough to find a venue that can hold the crowds where there's such incredible interest year after year.

One of the things that's been fantastic is being able to rely on the very rapid hardware advances. The Moore's law rate of exponential improvement is continuing and there's no end in sight. We can see this for the next decade, and probably for the next two decades we'll have the same doubling in a two-year period that we've experienced in the past. Intel, of course, is in the forefront on this with the speeds of their chips going up and including more and more onboard caching that allows the execution to be faster and faster.

Another big initiative that will be out in mid-2000 now is the Merced activity, which brings the address space from 32-bit to 64-bit. That four gigabyte limit actually is holding back some of the server and high-end work-station scenarios. But by going up to 64-bits, we'll really eliminate any sort of address space problems for many decades. And that will be an incredibly stable architecture like the 386 and up have been for more than the last 10 years.

We not only get the boost in performance from the individual chips, we also get the larger memory sizes, the increased storage capability, the faster networking, and we get to combine these processors together. Now, there's two key techniques for that, and both of them are very important. The first is symmetric multiprocessing, where you have multiple CPUs that see the same memory image. That's one copy of Windows NT that's running, but it's delegating threads and processes out to the different CPUs. Windows NT was designed from the beginning to take advantage of symmetric multiprocessing. Now, we have great eight processor systems out, and we're looking forward to 16, 32, and even 64 processor systems in the next couple of years.

The other technique, which is perhaps even more important, is taking logically independent computers and connecting them together on a high-speed network to create a logical cluster. This is a technique that can scale arbitrarily, allowing not only for increased performance but great fault tolerance. Since last year, we've had base level support for the clustering in. It's a huge focus for us because it's the only way we can get up to the performance levels that come with the Internet and all the transactions which take place there. Those demands go far beyond what any mainframes have had to provide in the past.

So, Microsoft has a lot of activities going on, a lot of different products and initiatives. They really fit into a few key areas, though. We have two big themes. The first is that businesses get immense value by having a digital nervous system. That goes beyond electronic mail, beyond eliminating paper, so that people can work together. If somebody needs to get news, good news or bad news, it's immediately available. And we're working with companies of all different sizes, different industries, to see how they can take the investments they're already making in PCs and networks to get far more out of it by applying this concept.

The other theme is the Web lifestyle. This is talking about how individuals will be turning to the Internet more and more for all the things they care about, whether it's travel reservations, sports scores, buying different things, staying in touch with their friends, there's no doubt that that's on its way. It will take great devices, cheap communications, lots of new elements will have to come into play before everybody is using this, but it's simply a matter of time, and there's a lot of great software opportunities that are created by that trend.

Our key products are reasonably straightforward: Windows, including Windows CE on smaller devices; Office, our biggest product line that we're constantly creating new major versions of; BackOffice, that's our fastest growing large product with SQL and Exchange, our Visual Studio tools that are fundamental in terms of how we let people take advantage of the entire platform. And I couldn't be more enthused about the new work coming in Visual Studio 6, and the work that's coming even beyond that. We have a lot of great things to make those tools better and better. Finally, our interactive media work is important to us because we're learning about the Internet by building sites ourselves and making sure that our various products are going to make that simple for everybody who wants to be out on the Internet.

Across all these products, there are a number of initiatives. The Internet was one that we made very visible at the end of 1995, after Windows 95 had come out. We've come a long way on that. We've got the leading Web server, we've got very popular tools, and our browser is doing very, very well. So, we feel good about that initiative. Really, I think that's well-accepted inside Microsoft. So, much more I'm driving the other initiatives, making sure everybody is think about those, scalability which includes not just high performance, it includes incredible reliability.

Interoperability, so that even though Windows makes up a higher and higher percentage of all the systems out there, we do a better job reaching out to the data and the existing applications that are on UNIX and mainframe-type systems which customers have in their environments, and we want to make sure we let them take full advantage of that.

Finally, I'll talk today a little bit about simplicity. That's really kind of a new thing for us, and it's one where there's a lot that can be done, whether it's the cost of ownership reductions, or simply making it so that users don't have to see error messages, don't have to be confused about what's going on in their systems.

It's easy when you're involved in this industry to take for granted the very rapid improvements that are going on. In particular the level of power that's built into the Windows platform has changed very radically. It was only back in 1990 that Windows first started to catch on. Up until then, people really discounted graphics interface as something that was going to be too slow, too hard to develop for. But around 1990, people saw that there was some real flexibility, some real merit, partly because the machines got fast enough, the graphics resolution got better, and so that became a key foundation layer for most new development. It was separate from DOS, so there were a lot of confusing things where you booted up the system, and where the device drivers ran and a lot of duplication between the two systems. Multitasking was weak, the communications support was very, very limited, adding devices was very complex. There was no networking built in. And so, all these things I've got listed here we now take for granted. And that's in less than eight years, a very, very dramatic change.

Now, the next eight years will bring even more of that, particularly in the areas of linguistic richness, Internet support, and the power that we build in to make it easy to write applications. The interface has gone from graphical now to Internet integration. There's still quite a bit we're doing there. And the next stage after that is having the computer remember what you've done over time, use natural language, speech synthesis, and make for a far more natural interface. We've been experimenting with these techniques in products like Office, where you type text into the answer wizard and it analyzes whatever you're interested in. We've also tried various agents and wizards. And we've been very enthused to learn which of those techniques make sense, and use the extra power to make them even better.

To me, Internet integration is an obvious thing. All the different operating systems out there from all the different competitors include a browser and are doing things to integrate it in. The Internet is a major part of using a computer today, and it should be as seamless as possible. If you want to get help information that goes beyond what you've got on your local machine, you just click on a link, and if you're connected up, the information should appear. Likewise, if you want to update your software, all the latest drivers should be out on the Microsoft Website, and it should be easy to analyze where you're not up to date, and bring the minimum set of information down to your system.

We've surveyed developers, and gotten very strong feedback on what we're doing with these very high-level Internet APIs that allow their applications to reach out to the Internet. Today, over 95 percent are building applications that way, and I have no doubt that that would go out even to 100 percent.

I mentioned simplicity. Everybody's probably got their favorite story about some way that the system is just still too hard to learn, whether it's forcing you to get down into the registry, which should never be necessary, or trying to figure out why one system is different than another system, or running into some files, in a file system that it's not at all clear what the file names stand for or what you're supposed to do with those things. Even the experts find themselves getting lost when they're doing things they don't do very often.

Likewise, the error messages that come out of the system can be extremely cryptic. The one I think is the worst, although there's lots that compete very well for that title, is the error I've got here. In fact, when I first showed this, people thought I got this error while I was giving the presentation. I didn't make it clear enough that this is, in fact, part of the Powerpoint slides. There it says: The DHCP client could not obtain an IP address. If you want to see DHCP messages in the future, choose yes, otherwise choose no. Well, in one way this message is very helpful. If you've gotten confused about the difference between yes and no, it really kind of clarifies that. The thing it doesn't make clear is what the heck is a DHCP message, and what does it mean to want to see them in the future versus not wanting to see them in the future, since it doesn't really make any sense. There's nothing actionable that you're seeing here.

So, those are the kinds of things we've just got to make sure people never see. They shouldn't have to manage their disks, they shouldn't have to deal with the error conditions, particularly in the area of communications that they have to right now.

Now, we've gone out and talked to business users and home users about this quite a bit, and there's really a mix of attitudes. People are so excited about the new things, they turn on every TV show, and they see those things, and yet they're worried about whether they'll fit in and they'll be able to take advantage of them. We've actually made a little video that shows some of the challenges people have in using these devices. And let's go ahead and take a quick look at that.

(Video shown)

MR. GATES:Well, so it's clear we still have a lot of work to do, to make all the scenarios easy for everybody who wants to use them. One of the interesting things is how the standards of the Internet are helping us move forward very, very rapidly. Microsoft has gotten very involved in all the standards committees, and in the last year has brought excellent progress. In the years ahead, even with the very tough issues of quality of service for video and audio transmission, and advanced security requirements, I feel sure we'll get the standard protocols we need for all of those things.

If you look to the developer, though, the explosion in the Internet is creating a lot of uncertainty about which programming models to go after. HTML has some great capabilities, but the scripts that are embedded in those pages really aren't rich enough to let you do full-blown applications. You'd like the best of both worlds, the best of the typical Windows programming model that lets you mix and match different languages, and have different applications, and the one-click deployment that is associated with navigating the Internet.

You want to have rich business logic, you want to have rich database source that manages the information. You want it to all be transacted, load balanced, and highly reliable. Today these really appear to be two different worlds, the Web and Windows 32. We've bridged that in some ways with the new forms models capabilities in Visual Basic and Access, but we have further we can go to really pull these together.

In fact, we created an architectural framework that we call DNA, that completely brings these worlds into a single, advanced Windows programming model. Now, this requires innovation coming from both sides. Our transaction server is a very powerful component container. And you can get at it through your Web pages, through the Web server, or through Windows 32. We're also taking and coming up with rich new graphics capabilities, that you can get at from HTML, Windows specific extensions there. So really the two do meet in the middle; you won't have to think of those as two different things.

The specific initiative we have for taking advantage of Windows graphical powers is called Chrome. It really makes sense, because of the fantastic advances in the graphics chips that are taking place. And it uses a very straightforward programming model. Site developers can easily take advantage of it, while continuing to work the same way they've always worked on machines that aren't as powerful, because we use HTML and XML tags in order to do this. And this basic thrust that we're showing here is one that we think is very exciting. So now I'd like to turn it over to Eric Engstrom, who is the manager of this project, and ask him to give you a quick look at what he's doing with Windows Chrome.

MR. ENGSTROM:This is my computer, running on a machine version of Windows 98, with Chrome added to it. What we've basically done with Chrome, is we've taken our Direct X multimedia technology and made it accessible from an HTML space, as well as from a normal business application space on Win 32.

So as I move over here, I'll show you just first-off some of the UI enhancements we've been able to do with this. This is a normal live piece of HTML. You notice when I press seek, it works, all those hot links and everything you'd author the normal way you'd author a piece of HTML. And then there's a few lines of Chrome that allow this to actually work.

I'm going to close this and go to the rest of my demo, because we don't have a lot of time. Here's a Web page I'm sure you're all used to, and you'll notice that as I move over it, I have a fully functional pane here, as I move back here I get a much broader working area to view my Web page in.

Here we're dreaming a little bit about what a future desktop might look like. This isn't actually in the product. It's just doable with the product. As I move over these, you'll notice that they come to the front. And again, totally live HTML. These are normal web pages, we've simply put them inside a Chrome working space. So that's how Chrome interacts with Windows applications and Windows business apps.

Now, I'm going to show you an example of how it improves the Web page. This logo MS World here was actually copied off another live Web site we found on the Internet. This was previously done with an animated GIF for the 3D text, which took approximately 70K. You can imagine how long that took to download to get this GIF. It's 355 bytes with the Chrome. And the frame rate is 60 Hz, as opposed to the 15 frames a second you get with an animated GIF. This is really the second feature, if you will, of Chrome. The idea is to bring broadcast quality graphics to a PC over the Internet, things you're used to seeing on the television, that if we tried to download it over a 3K a second wire would be intolerably slow.

The other thing I'm going to show you here is a thing we've done in Chrome to make it easy to do this sort of thing. This is the help system for Chrome, the way it comes on the CD. You'll see this 3D spinning text. We have a little wizard built into the help system, where I've typed Tech-Ed, and I've changed the font to Times New Roman. I'm going to say, update the text, and you'll notice over here that we've bolded the changes made to the HTML to allow this to happen. You can literally go into the help system, and interact with it until you have a piece of HTML that does what you want. And then just mark it and paste it into your page. We have that interactivity --

MR. ENGSTROM:We have that functionality for everything you can do in Chrome. And I'm going to show you an example of using Chrome on the Internet to update a news broadcast. You'll notice we have the globe spinning here. As I click on Moscow, you'll notice that we highlight Moscow, where it's at, and then the news comes out at you. You might think of this as totally gratuitous graphics, but really, that's not it at all. We still have the pipe problem, and one of the things we can do that solves this small pipe on the Internet is, we can attract the user by providing them with more information, or some sort of transition that actually seems part of the content. So the Chrome team spent a fair amount of time watching television, a really hard job for us, and tried to take advantage of the things that they do, even though they have all the bandwidth they need to make the news interesting, and try to put those in a way that didn't take a lot of bandwidth, but would allow us to hide the time it actually took us to get the real information. And the result is an example page like this.

Finally, I'm going to show you what you can do with just straight Chrome. Most of the pages I've shown you so far have some sort of down-level storage. Obviously, the frames bending or the spinning cube in the My Computer panel have a way to function on a non-Chrome machine. But here's some things you can do if you just use Chrome by itself. This particular graphic here we tried to make look like a regular television advertisement, though we're not quite as creative as those people. The download size for this is around 2K, everything else is rendered locally on the client.

MR. ENGSTROM: And here's another simple example. The text you see on the sign is regular HTML, and you can add these kind of Chrome things which are ignored by down-level browsers. You don't have any compatibility problems. You can put your normal HTML content on sort of this thing and spin it around for an extra graphical appeal.

Okay. Finally, I wanted to show you that Chrome even works on the active desktop. Hopefully you can find something more interesting to do than a fish tank, but we're programmers not artists. So, this is Chrome running in the active desktop. And that's pretty much it. I'll turn it back to Bill Gates then. Thank you.

(Applause)

MR. GATES:Well, thanks, Eric. It sounds like you got a very good response. We ought to get that one done and get it shipped.

In terms of DNA, one of the key elements, of course, is the object plumbing, making sure that components are self-describing in a very rich way, including things like an entire application describing what it needs in order to run. We're already pretty far down the path there. We took a lot of learning from the Transaction Server and our early COM work, and really pulled those together. And so this year we'll have out the full advance of COM+. That will be fully specified, and people can start to work with that. There is more to come in terms of the synthesis here, particularly in the area of forms, pulling those things together, and having very rich storage capability. The forms work we've done with our Trident capability in the browser that you can access through data pages in Access and also through Visual Basic. That is a start in that direction.

Likewise, in the storage area, we have a very rich API called OLE DB, it's the follow-on to the original database API, the ODBC, but here it understands things like hierarchy, or arbitrary properties that are necessary when you start to deal with information, including mail messages. And so by enhancing that API and building a storage capability that has all of that richness in it, we'll be able, again, to bring the Windows 32 capability to a new level.

As we charge down this path of making Windows richer and richer, we are guided by a lot of things. We have our research work to show us some of the new possibilities. But even more importantly, we have the input that all of you are able to provide. It comes in many forms. Our field group out there meeting directly, the advisory groups we put together. The specific forms we have for developers, including the open process where we have design reviews of new areas. We've got a lot of online capability that fits into this. And some broad reach programs, including the surveying we do, and the feedback forms that come from an event, particularly a large event like Tech-Ed.

A good example of how we get out early when a product is still being shaped so the feedback can really make a difference is what we've done with Windows NT 5.0. We actually started this process many years ago. We talked about what we thought made sense in directory. We had a meeting for management services. We had a broad developers conference a little over two years ago. And so, as we constantly refine NT 5, we've had the benefit of this guidance.

We are reaching the final stretch. We're not too far away from having the major beta release, the beta two release of NT 5, and we'll get it out to the marketplace based on that. Any major feedback we get at this point will probably have its effect on Windows NT 6, because we feel like we're pretty far along with NT 5, although in terms of making sure it's totally reliable, we're still going to take whatever time it takes to make sure we get that right.

We did a special exercise here at Tech-Ed '98 where we had a survey form, and we got several thousand people to go through and pick some of the items that are top-of-mind for them in terms of how we can do a better job. I've got two slides here that go through those things, and around Microsoft we're taking a hard look at each one of these things to see where we can do more.

Better job of fault isolation, detection and resolution of mission critical products is a very key thing for us. It's interesting in our support organization, we're actually getting less phone calls than we got in the past. We think a primary reason for that is that people are using the Web as the way of getting information. Our support and information up there is getting an immense amount of traffic. However, because it's really the simpler issues that are handled by that Web information, the average length of the calls we get into support are still going up quite a bit. And, as we look at those, we realize that this idea of having diagnostics that can immediately tell what's going on, determine what might be going wrong a lot faster than people have to do today, that is going to be critical, certainly for us as well as for customers to reduce the turnaround here.

Fix bugs faster in key tools and platform products. That's a message we've had for a long time, and we're doing much more with service packs, and identifying service packs to be very, very focused on fixing problems that people are running into.

We started with NT doing that on a discipline basis, or as to the old products what's the next. Now we've spread that so that Office and our tools are also involved in that, but it's something that we can always do better on, and particularly in terms of driving these things in a timely way.

More comprehensive documentation. That's one where, particularly with sample applications, I think we can do more. This is a case where being online will be very helpful.

When discussing complex theories, like three-tier, provide more examples. This is a case, where with three-tier, we're just coming out with some new examples. But I think the general point is a very strong one there.

Make it easier for developers to get technical support. Again, we'll have to drill into that one to see which kind of technical support they'd like us to focus on there, because the online and the phone are both areas where we're putting more and more investment in.

And, finally, be more open about the directions we're taking. Events like this are a big part of it, our Website is a big part of that, but we understand that feedback. We can never be too exact in terms of the exact timing because we don't prioritize the schedule over the quality of the work, including the deep integration and the frontiers that we need to tackle there. Things like unified storage for general objects, things like the linguistic work we're doing, we understand our direction very, very well, even though the exact timing we don't. So, we're going to have to be careful to make that distinction while being very clear about where we're headed.

We've set up a Website so that it's got a lot of Microsoft employees going through these items and understanding them a bit better. We can show what new activities are going to be focused on in improving these areas. So, you might want to take down that URL, it's http://www.microsoft.com/msdn/news/punchlist/ , which is the term they gave for this feedback list.

As we charge forward, we are investing more and more in R&D. This really means one simple thing, it means hiring lots of great people, both developers and testers. The portion of this investment that goes into testing has gone up even faster than development, and in many products it's equal or even slightly greater than the development investment. And that's an area where research improvements really are needed to have much more automated high fault isolation in order to allow these rich products to get out into the marketplace in an extremely reliable form.

The goals we have in our research are quite ambitious. The computer today doesn't see, doesn't listen, doesn't learn. And as we reach those milestones, people will look back on what we have today and wonder why it was so limited. Underneath, there will be a lot of rich software that makes this happen. Things like the graphics capability, decision theory, rich database advances, particularly in the object-oriented area, all of those, we're going to have move a long ways in to build this promise. As I said earlier, the hardware people are doing their part in terms of memory size and storage size. So this is going to happen. When in the next decade we get to all these things is tough to say, but I feel very comfortable in saying that everything here will be a standard part of Windows within that next decade.

We thought we would show you a piece of progress we're making here, and this is a great demonstration because it's got many different aspects to it, including linguistics, parsing and automatic reasoning, and using feedback from users to adapt to what they're interested in.

So, let me now turn things over to Greg DeMichillie who is a program manager on our intelligence assistance work.

MR. DeMICHELLIE:Thanks, Bill.

What I'm going to show you here is a prototype of sort of the next generation of agent technology that Microsoft is working on. What you see here is Microsoft's Outlook 98. This is our most recent version of Outlook. But what it has added to it is a prototype of an agent that is able to read messages for me and help me respond to them. Now, back in our Office 97, we first introduced this concept of an agent, and that was a great first step. But this agent is a step in a new direction because, since it has been installed, it's watched me as I've responded to hundreds of email messages. It's learned things like, what are the typical things I do in response to a message, do I read it, do I reply to it? Do I schedule an appointment? And it also learned how long I tend to read a message before taking action. So, it's able to tune itself, and it provides feedback when I'm most likely to need it, and provide that feedback in a very focused way. And, as Bill mentioned, it combines many of our different technologies. It combines our agent technology, along with our text-to-speech, speech recognition, decision theory altogether.

And to see how it works, I'm just going to double click a message here that I got called Project Update. And you'll notice this says, "Perhaps we should get together sometime soon, how does your schedule look next week?" And it popped up and it asked me, would I like to see my calendar. And using voice recognition, I can say, yes. It pops up my calendar. And, you'll notice, it gives me a calendar that is appropriate. It was able to read the message and understand that it was talking to me about next week, not this week. So, it was able to provide that feedback.

But it can not only do simple things like that, it can actually schedule appointments for me. So if I open up this message that says OEM get-together, it says, "It would be nice if you could attend this meeting with the OEM team this evening."

You bet. So what did it do? Well, it created an appointment for me, it would put the text of it in, and it verified that that time was available for people. Now, if any of you have scheduled appointments, you know that something like 15 or 20 clicks that are required to create an appointment, look into everybody's schedule to see if the time is free and send it, and this was able to do that by intelligently analyzing the way I work and the context of this message.

MR. DeMICHELLIE: Finally, let me show you one where this holds a start and stop time, so you'll know if it says we'll meet Thursday from 3:30 to 4:00. So now I've got a meeting set for 3:30 to 4:30. And it's intelligent about things like next week, or two weeks from now, or next June. So this, I think, is a really great example of how the interactivity that we now have in our products is going to take a quantum leap forward, as we get things like this agent, that are able to learn from my personal way of working. And with that I'll turn it back over to Bill for the rest of his presentation.

MR. GATES:Thanks. Let me just wrap up by saying it's a time of great opportunity, for Microsoft, innovation in Windows and the tools around it. That's our top priority, and we're seeing there's a lot you're asking us for, and we think we can make very, very rapid progress. Looking out a little further into the future, computers will be a much better tool than they are today. And that has profound implications for how we do business, how we learn, or even how we entertain ourselves. There are almost no businesses that won't be touched by this way of getting their activities done, and this way of reaching out to their customers. We're very excited to have the relationship we have with you. We appreciate all the time you're taking to come to our events. And the more feedback you can give us, the better job we can do.

Thank you.

MODERATOR:We'll begin our Q&A session now. If you have a question, please step up to one of the four microphones in the front. Step closely to the microphone and speak clearly, please. We have a question at number four.

MR. CHRISTIANSEN: My name is Bart Christiansen , I work for Chicago MBD Bank. And I had a question -- I know that my SQL and my SNA servers, they're all year 2000 compliant. But, what UCL (sp) is going to be -- some of my colleagues were talking at lunchtime today and they were saying, well, lousy systems aren't up to par, the BTN (sp) the CICS regions, and how do you feel that that factor is going to start to affect us, and also the global economy?

MR. GATES:The year 2000 problem is a really major problem for the computer industry. And it's one that's forced a lot of computer managers to go back and look at the old applications they have, and how easily available are programmers who understand those applications. If you really know the application, going and finding the place where date comparisons take place is not that difficult. But, if it's an application that nobody is all that familiar with, making sure you've found all those things can be very, very difficult. Now, the PC industry is very lucky that for the most part, dates aren't stored in a two-digit column. And so the kind of major problems that you'll have in large machine applications, typically you don't find in any PC software. We've created a special Website to address the year 2000 issue, and it goes through the status of the in-depth testing we're doing. We do have some older products that require some patches, mostly in terms of cosmetics, how they display the information. But, the site goes through that in great depth.

For the personal computer industry, I think there is both a plus and a minus here. The plus is that these problems probably will accelerate people's willingness to say, it's time to go out and get new packaged applications, and use the new, more efficient PC approach. But, as we get into late '98 and '99, there will be companies that do less in terms of new PC-based systems, because they'll be distracted going through their year 2000 issue, and we'll only get past that as we get out to mid-2000. We're trying to do our best to get information out to customers. I think the people who talk about this problem in an extreme way go overboard. But, it's probably mid-range between the people who view it as a minor problem, and the people who view it as major. It's right in the middle there, and therefore, something quite serious. And the sooner organizations get a plan to get that behind them, the better off they'll be.

MODERATOR:Thank you. We have a question at microphone number three.

MR. MILLER: Yes, I'm Larry Miller, from St. Louis, Missouri. I'd like to ask, when do you think we'll have 100 megabits per second bandwidth to the home? What are some of the obstacles, and some of the intermediate stages we'll see?

MR. GATES:Well, that's a very good question, because it's tough to predict. And if we had it, we could do some fantastic things, helping people maintain their systems at home, and letting them get at very rich, high-quality video. In the business environment, the move from 10 megabit to 100 megabit has been taking place for a couple of years. And so it is very typical to have 100 megabit, and now in some cases, one gigabit access into the network. So it's quite a contrast. Here, in the business environment you're going to have this high speed connectivity. And that's why breakthrough ideas like the Intellimirror capability in Windows NT 5 can work very well, because we mirror all the states up on the server and on the client. So you never miss the information, it's always backed up, so you don't have to overload things, or you're able to work offline.

In the home environment, you've got the cable companies going in with cable modems, those are not 100 megabit. And also, that's a shared bus approach, so even the top rating, you don't get the full 10 megabit that they talk about there. Another approach of the phone companies is various kinds of DSL. And none of those are 100 megabits, even the DDSL, which is the best, which is about 56 megabits, is still a little bit short of that. There's other flavors, ADSL is typically around 6 megabits, and a new thing called UDSL or CDSL is about 1 megabit capability.

So my view is that bandwidth to the home will speed up. cable modems and various forms of DSL will be in 20 to 30 percent of American homes within the next 3 to 4 years. There are some places around the world where fiber has been installed all the way to the home. Singapore, Hong Kong, parts of Australia. There they could have the 100 megabit pretty rapidly. But, so far they haven't priced it in such a way, and gotten enough services together that that's become a mainstream thing. That's why you see Microsoft reaching out to work with phone companies and cable companies, really to encourage them to make those investments, because we think there will be very high demand for them.

I'm afraid that it will be a two-stage process, a stage where we move to DSL or the cable modem, that will take some time, particularly on a global basis, and then a stage after that where we move up to actually running the optic fiber all the way to the home and we get that 100 megabit. So it could be more than a decade before a high percentage of even U.S. homes have that 100 megabit.

Now, I don't want to paint too bleak of a picture here. When you have a connection that's, say, even just a megabit, you can do quite well, in terms of maintaining files and information, and providing high-speed Internet access. The main thing you miss at one to two megabytes is you miss the ability to do video. So you have to get your video either at low resolution, or get that as a broadcast transmission, and then mix it in with your Internet capabilities.

MODERATOR:Thank you for your question. We'll take a question at number two.

MR. WOODHILL: Jim Woodhill from Houston. At conferences like this, we're encouraged to develop applications, commercial applications, that use all the cool new features coming down the pipe, like the Active Directory in Windows NT 5. I'm a developer of enterprise-based applications, and I need to know when I'll have a market. What are Microsoft's internal projections for what year and what month Windows NT 5 Server will start outselling Windows NT 4 Server?

MR. GATES:Well, I'll give you as much guidance as I can on that. We're a little conservative about doing numeric forecasts. The day that NT server 5.0 becomes available to the marketplace, it will certainly out-ship NT server 4.0. But, for your question, you're probably also interested in the installed base, and how quickly we're able to move over the installed base. NTS 5.0 is a huge breakthrough, compared to NTS 4.0, the directory, the clustering, there is just so much there I can't imagine somebody installing a new NTS 4.0, after 5.0 comes out. Some people will, but the vast majority will move to 5.0.

In terms of upgrading, we've tried to make that very smooth, and the benefits are pretty strong there. People move a little bit slower to upgrade servers than they do other machines. If the application is running well, then often they'll wait until there is a new version of the application, before they do the upgrade.

I would think that not only will NT Server outsell 5.0 -- outsell 4.0 in '99. But, within a year of its shipping, we'd have half the installed base up to 5.0. And particularly people who are in the market for new applications, I think we'd have an even higher percentage of those people moved up to 5.0. It is the biggest set of advances we've had in NT, ever since the very first version of NT came out.

MODERATOR:Okay. Let's take a question from number four.

MR. TOM : Hi, I'm Tom from Baltimore. And I'm curious as to how the actions of the DOJ recently have affected or changed any of your plans or strategies or plans moving forward?

MR. GATES:Well, I'm glad you asked that, because I think it's very important for us to make clear that nothing that's happened there in any way has changed our plans. What we're doing to improve Windows, what we're doing with our products, it's all the kind of thing that the laws and the system are designed to encourage. In fact, there's no industry, I think, that has a better track record of improving things, and even bringing the cost down, like the personal computer industry does.

So although it's a distraction for the lawyers who are involved in it, it's very important for me as the leader in the company to make sure no one else is at all distracted by this, and that we just keep being guided by what it is that our customers are asking us to do. There's a lot of press that goes along with a lawsuit. But, the fact that we're putting the browser in the operating system, it's a good thing for users and developers and we have no doubt that the courts will find that's a great way for us to do things.

MODERATOR:Okay. We'll take a question from microphone number three.

MR. MIKE: I'm Mike, from Baltimore, also. Bill, I'm just wondering, as your company reaches higher and higher levels of success, how are you still able to keep it motivated and keep the drive like it was when -- pre Windows 95?

MR. GATES:Well, every time Microsoft has reached a new milestone, it's always a challenge to make sure we aren't getting lazy, and that we're not resting on our laurels. I have a lot of management views about how we need to focus on the bad news, not the good news. We need to really listen to the customers who aren't happy, not the ones who are. And that will guide us to new levels.

The real vision of what software can do for people, we're nowhere near exhausting that. Even if you go back to the sentences that Paul and I -- Allen and I wrote down when we started the company, a computer on every desk and in every home. We're nowhere near to that. And so by picking a vision that was breathtaking, you know, I know there's at least 20 years of good work that we have left in improving Windows and all the other products, to get where we need to get.

We do have a lot of competition, whether it'sIBM, which is still four to five times our size, or people like Sun, or Oracle, or many, many others that are out there. We've got people who can constantly remind us where we're falling short, where we need to do new things. And there's no shortage of ideas of how to make the products better. Large companies do have a tendency to slow down and miss what's going on. In the computer industry there are a lot of famous examples of that, and so I come in every day saying, I'm going to make sure this is not the day that we've reached our peak and, you know, make sure that in our hiring, and our excitement about the things going on, that the message is there to keep people really doing their best.

If I look at the people we've been able to hire into the company broadly, particularly in the development groups, and in the research activities, we have hired more great people this year than any year in our history. And that is a great leading indicator of whether we're driving ahead or slowing down. I take a lot of good feeling away from the fact that the wonderful people are still joining us and coming together with the people who have been here for a long time.

MODERATOR:Thank you for your question. We have time for one more question, we'll take it from microphone number two.

MR. : Hello, Bill. Let me start by saying, we really missed you here this year, and wish you had been here. But, my question is really about CE. By the way, thank you for making those available to us at the show. I bought one myself. It's multithreaded, okay, has plug-able feature set, runs on different types of processors without problems, appears to be very scalable, and also it doesn't crash. Is this the OS of the future?

MR. GATES:It's very interesting, we have a three-level Windows strategy today. Windows CE, which does not require a hard disk, it's quite small, and very portable. Then we have Windows 98 in the middle, and Windows NT on the high end. The most popular product, the highest volume product by far is Windows 98. It outsells the other two put together by a large amount. However, our strategy is very focused on Windows CE and Windows NT. We're being pretty clear with business customers that Windows NT 5, as a desktop and a server product, will be the best choice for them, even though they'll continue to be able to buy Windows 98. And as we get out into the future, we'll do a version of NT that's focused on consumer machines, particularly as 32 megabytes becomes typical on consumer machines, which we're almost there already today. We can bring NT down into that space.

NT has taken time, it's had to mature in terms of the breadth of device drivers that are out there, the compatibility, the power management, and we only get there as we get NT 5 into the marketplace. Well, all those nice attributes that you mentioned are common to Windows CE and Windows NT. There's no doubt that Windows NT with the rich file system, the rich graphical capabilities, the rich Internet support, for anything where you've got a full-sized screen, whether it's a portable full-sized screen, or whether it's a stationary full-sized screen, you're going to want to be at the Windows NT level. For things like a pocket device, an intelligent TV, your auto personal computer, your intelligent camera, that's where Windows CE will match up.

Like every family product line, you want to make sure that you're covering the entire space with the things you're doing. And I think with Windows CE we've been able to achieve that. So although Windows 98 is the big thing today, I do expect, say, when we get together two years from now, I'll be able to say that Windows CE volume and Windows NT volume are as great, or greater, than the volume of the product in the middle.

OPERATOR: Okay. Thank you for your questions. I'm sorry we couldn't get to all of them.

Thank you, Bill, for joining us today. And thank you everybody for coming. We'll see you at the party tonight.

(Applause and end of event)