Speaker:
Bill Gates
Chairman of the Board and
Chief Executive Officer
Microsoft Corporation
Event:
1998 Microsoft Financial Analyst Meeting


The following are highlighted excerpts from Bill's presentation.
Good afternoon. It's exciting to be here and to have a chance to focus on what I'm going to spend most of my time on now, which is the breakthrough product innovations that will really drive our future.

Before I get into that, though, I thought I'd deal with a concern a lot of people have shared with me, which is, you know, asking, how do Steve and I work together? And in fact I can give you some pretty clear evidence that we work very well together. We got away from campus a few months ago, and were working out some of the key issues. And we took some video of that. So, I'd like to show that real quickly, and you'll understand how we work.

(Video shown.)

So, it's exciting for me to put Steve in the driver's seat in terms of a lot of the customer issues, and a lot of the business issues, and let me spend a lot more time with the product groups.

I just have one slide that looks backwards in terms of the innovations, because there have been some big milestones. We were the very first company focused on microcomputer software, there was nobody else who had thought about that because the microcomputer was just coming along. But perhaps most important was our business model, to be very specialized, not a services company, not a systems company, not even a vertical applications company, strictly a software platform company working with all the different hardware companies who came in with PC-type equipment.

Microsoft Basic was kind of an amazing product because it had to run in 4K bytes. Memory was very expensive back then, and just getting anything to run on these machines was pretty hard. But that was really our biggest product for quite some time. It ran on the Commodore Pet, the TRS-80, the Apple II computer, every personal computer that came out in the '70s was standardized around Microsoft Basic. And, of course, we built applications on top of that.

In the 16-bit era, we broadened the company a lot. We did the first word processor that had a graphical approach. Obviously, we benefited from and hired a lot of the people who had been at the Palo Alto Research Center. You know, we bought 50,000 mice, because we needed a mouse to make the thing work well. And, of course, those wouldn't sell at all. And it was a slow start, but the technology that was there over time has gone to give us a strong leadership position.

In the '80s, we focused on the Mac, and we started building Windows. We advanced with the 386 version, the first 386 offer that came out, and we did our spreadsheet breakthrough, Excel, and we also started the integration.

The idea of integrating the applications was a fascinating thing because in the early '80s people were doing what we thought of as premature integration. They were doing Symphony. There were about 50 different packages that came out, and all basically failed because there really wasn't enough power in the machine. Later, we came along with a different approach to integration, which was the Office approach, and that did super well. We started that on the Mac, but then, of course, it's big on the PC.

One of the interesting products that's on here that you might not think a lot about is Microsoft Project. We have about a dozen businesses like this that are just amazing businesses, where we've gone in and revolutionized the size of the business, and the customer set. Our project management software in total in 1990, when we first came out with our product, was about $200 million a year business, and that's accounting across all computers, PCs, minicomputers, mainframes. Nowadays, that's grown to be about a $500 million business. And all of the growth, the $300 million has been our product, Microsoft Project. There's been other people coming and going, but the way we use the graphical interface, we brought in a new class of users. We gave you flexibility, immediacy, that's made that a huge category.

So, for us, at $300 million of our revenue, it's not something people look at a lot, but it's the kind of business we're very proud of, incredible growth, great customer feedback loop, very synergistic with the other things we do. We love having businesses like that. We look forward to coming up with more of them.

In '91, a couple of big things there, a real push to have object orientation. So-called OLE was the first time that you could put video, audio, arbitrary objects from other applications in, and some of the very tough issues about how do you make that efficient, how do you define the name space, is the object linked, all of those we tackled back in '91.

In some ways, like Windows itself, it was ahead of the market. The machines weren't big enough or fast enough. And so it took some time for the machines to grow into it. Today, people use this without even thinking about it. They just put things -- lie things on the clipboard, move them around, cross-link by clicking on the different objects. But when we first came out with that, that was a big contribution.

Visual Basic is by far the most popular development tool in the world today, built into Office, built in to Access. And there were some very big things there. The next biggest development tool, of course, is C, and C is bigger than the next biggest thing by some large amount. But Visual Basic now has over 10 million active developers worldwide. So, evolving that on top of the platform is a very big deal for us.

And then, of course, Windows NT. The same way we bet the company on graphics interface with Windows, we really said the next era would involve moving up and bringing in the kind of power and robustness that people have thought about mainframes and minicomputers having, and Windows NT was aimed at that. And it really has been the foundation for a lot of our growth. Graphic interface drove an era of growth, Windows NT has driven an era of growth. Now, I'm going to talk about some of the things that will drive our opportunities looking forward.

None of these things happen overnight. Take Windows NT, you know, we hired Dave Cutler, and it was many years before we even shipped the product. When we shipped Version 1.0, the broad press perception was, who cares. It's ridiculous that a PC piece of software could grow up to be something that would be used very broadly. And see, I've got a lot of patience and belief in these things, and really persevere with them. And for every one of these we have that type of commitment.

I'm not going to hit all the different areas of innovation, but I'm going to hit a few that I think are very high impact that we're making good progress on. 
Here's one that's a little bit like that Project business. It's an area of our business that people don't look at very much. In total, it's on the order of $500 million. It's the work we do in peripherals, the mice, to start out with, but now, as you saw today, some new things. And now looking at some things that aren't shown here at all. In fact, we have a couple of products that come out in the next year that we're very enthused about.

The basic spec here is, we design these things, we don't build them. They're very software intensive. They're outside the PC and so they're visible. The design quality shows through to the user, and there will be many more of these things in the years to come.

They also enable new scenarios, they let people do things that you wouldn't have thought of doing with the PC previously.

We're also involved through partnerships now, not just in the PC improving, but getting into the wide variety of devices. And all these devices have to work together, because you don't want to take any effort to move your information around. If a friend's phone number changes, you should change it once, it should show up on your cell phone, your PDA, your PC at work, your PC at home, your Web TV, anything you work with, all of that should be there automatically. And it's not just name and address lists. It's your messages, your schedule, your files, your preferences.

Say you find a neat new Web site, and you'd like to navigate to it, why shouldn't that be there wherever you are. All of that information has to be seamless. And so there's a very rich object model that encompasses user preferences, and user state information that will migrate around and through these devices. You see that a little bit today with what we're doing with the directory. You see it a little bit today with what we're doing in Hotmail, but fundamentally there's more to be done there.

The volume of these devices' potential is very large. The systems software royalty will scale with the price of a device. So if a PC is $3000 these devices are $300. You'll have a substantially lower royalty, the only way to make up for that is the volume. Also, of course, it's no a substitution phenomenon, it's a complementary type device.

Even in areas like set-top boxes, if you take a 5 to 10 year view, eventually those things will be sold as retail products, and will have the access control scheme that still determines what material you can get that the cable operator will have control over. But, the technology just moves too fast, and the variety of things that people might want to hook up to the high-speed pipes into the home, cable or phone, push for the retail model. And that's partly why we're working on this early design wins with so much vigor. Those early design wins are interesting, but we wouldn't put as much priority on them if it wasn't the opportunity to see that fitting in with your PC, fitting in with the other devices, and also being, themselves, retail products over time.

Simplifying user interface is the top priority. That is the thing that is most important. And there's many ways we have to tackle this problem. There's the mundane level of just looking at things we've got that are confusing or wrong. There's the level of looking at the fact we have too many mechanisms. 

Simplifying those mechanisms, and then there's the fact that we have too many concepts and actually getting rid of those concepts.

We did take a step towards that with the browser integration, and we're just going to do more and more of that. The forms capability which has always been native to systems, the navigating of the local storage, will make it so you really don't care whether it's an application form, whether it's local data or remote data. The same way you pick up the phone, you don't care if it's a local call or a long-distance call, the way you search, navigate, annotate, things that are actually very hard to do on both local and remote information. That will come through with the way that we drive the browser.

We also have to stop adding new things. You know, the menus, when you get about 30 menu entries, it's too many. And although people don't use all those commands, the interesting thing is, every one of those commands somebody does use. We only put commands in that are there for a reasonable-sized portion of the user base, the legal command, footnote commands, a lot of users don't use them, but enough do that it's important to be able to exchange documents with that kind of richness.

So how do you deal with that dilemma? You've got to have the capability there, but how do you expose it? There's a couple of ideas. One of course has to do with using a sentence type approach to navigation, that I'll talk about later. The other has to do with an adaptive interface that learns over time. So that you can take actions to simplify it, or the machine itself can help you do that.

Now, a milestone in this area is going to be the next release of Office, that is out next calendar year. 

[demo]

Office is fascinating in that it's become a vehicle to actually take research work and get it into shipping products, starting with natural language and grammar, now with that stuff extending to all the key languages we're in. And that will drive a lot of the new -- the new versions. An even bigger challenge than unifying the interface is getting the concepts underneath so they just aren't as complex.

We've been talking about this for quite a while. In fact, it's an area we've been working on for some time. What it really requires is essentially an object file system, something that takes the characteristics of a database, a Web store, a mail store, and can handle these things.

A file system today just isn't good enough. It doesn't understand all the different properties and attributes, and searching capabilities you have on, say, mail messages, or project files. It doesn't relate to the richness of what's in there. And so you don't have nice views. You know, you certainly wouldn't want to take your mail messages and have each one be a file, and just use the shell as your way to go around and look for those messages. Likewise, you wouldn't want to use it as a way to manage a database.

Now, if you can get a store that does all those things, then you get a seamless approach. Anything you annotate, anything you move, it's all done I one place. You have one set of commands, and it's not just an illusion that somebody built at a user interface level. This is something that we'll build into the next major rev of the operating system, past NT 5.

We actually have a lot of code running today, but it's not something that makes sense until we get out to the next big release.

One part of this, though, is to take replication and have it work pervasively. You hear a lot about replication today, you know, mail products talk about replication, file systems talk about replication, database guys talk about replication. But, really, you shouldn't have to think about that at all. And so not only do we need the bits to move around to all the devices, we also need a common understanding, what's called a schema, of all these objects. What is the user object and their properties. So I never want to go to a Web site and have it ask me things that I've already answered. But, then again, when I go to a Web site, I don't want to necessarily automatically reveal all that information.

So the ideal here is, I tell my machine my zip code, my credit card number, my address, my preferences, and that's stored in my machine under my control. And then, as I go to various sites, I pick the level of disclosure that makes sense for that site. Maybe my zip code I make widely available, when I'm about to do a transaction I give out my credit card. And there's ways of making these things very anonymous, that avoid there being any type of privacy problems. So on your local machine, or anywhere you log on, your business card, your payment information, your calendar, we can store all those things and they're available to you and whoever else you say that those things should be available.

And so we can take simple scenarios that should be possible, that just aren't possible today. You know, we were sitting down talking about why don't people use the electronic calendar anymore? Well, the fact is, virtually all the key scenarios are just way too difficult, even if you got the handheld devices to be inexpensive, and got everybody you interacted with to use them, there's way too many steps that you have to go through.

Moving to another area that we do make a huge advance in the NT 5 time frame is this management thing, the state of a machine, and where it is. You know, it's too hard to know not only the user's information, but also the PC specific information. There's a tendency for stuff to accumulate over time and each machine becomes idiosyncratic, and you can't set policies without actually going out and visiting the machines.

You know, can you tell what the difference is between these machines? Absolutely not. It's only by having directory policies that you get with NT 5, that you can say, look, every one of these machines I want to be the same, and therefore, I know when I deploy applications I'm going to get what I need.

So there's two different things I'm talking about, there's user state, that should roam with the user, and then there's machine state, that goes out and, you know, controls things like the applications that are available, the fonts that are available, how up to date the drivers are. Both of those should be independently controllable. And today they've been mixed around so much that if you tried to take all the bits on a disk and move it from one machine to the next machine, it won't work, because even though your user state is there, the machine specific things that rely on that hardware configuration are also there, and so you'd spend literally days sorting all of that out. So, it's just way, way too complex.

Well, finally I get to natural interface. I do put natural interface at a level of importance below simplicity. Just putting a speech user interface on the PC we have today really wouldn't help that much. I mean, so what, you have a speech user interface to go find out what's wrong with your registry, great. You can try and pronounce all those long registry keys instead of typing them in. It's not that much better. You know, your computer can read to you the error message you don't understand no matter what is going on. So, the conceptual improvements are actually the hardest part, and the most important part.

Now, as we do that, we're able to broaden the audience, let people do more ambitious things to their machines. And then we put on top of that the hardware innovation of a tablet-like device. You're starting to see some people do electronic books. I don't think they're adequate today, but the screens will improve. You're starting to see Web TV that actually does come at the problem with a very simple user interface, and so the hardware enablers will be there. 
The performance of the chips will be there to let us do these things. So, if you're ever confused using your PC, you just state a sentence, or type that sentence in, and it can work for you.

Now, this is another case where it's not a big bang, you know. We have been making progress in this area for many years. We started over five years ago with our natural language work. One of the great milestones that's coming up will be a key feature of the chips with SQL 7 towards the end of this year, and it's the SQL 7 English query.

[demo]

The amazing thing about an application like that is, you know, when we started working with a group called MLI, it really wasn't practical in terms of size and speed to do it. But we knew that picking up the work they'd done and putting a few years more in, that we would be able to intersect with the better hardware and have it as a standard feature. And so, I think you're going to see all the database applications working that way, as well as your ability to navigate all the crazy things on your PC using that same type of technology.

I'm not going to dive into any of these other areas, but these are areas where I'm very enthused about what will come out within the next few years, collaboration, you know, key areas. When you're navigating the Web and you have a reaction to something, how easy is it to edit that, annotate it, share it with other people. Likewise for the documents you have in your company. Can you go back and find a document with 60 seconds that somebody might have done?

It's all still way too hard. Defining a structured process just by doing that graphically, we have to have that as a built in way of managing information. Graphics, the world of graphics and video really are coming together. Digital video lets us do some fascinating things in terms of interaction, resolution, even taking different viewpoints and using a lot of compression to have that in what might be a normal video type channel. So, the world of video watching is going to change dramatically. We're getting enough capacity that storing video and moving it around is within the realm of possibility.

Database, there's a lot that comes out with SQL 7, but there's a lot more that has to happen in the world of databases, shared, clustering, and even more rich data analysis. Imagine the data pivoting you saw earlier today and this English Query, and how those two will come together in a very rich way.

Finally, distributed computing, our industry has talked about that for 10 years. It's still so difficult to write distributed applications that there's only a few cases where people really have done it. Typically, they're just writing one end. They write one end, the database runs on the other piece. You have a few people, like SAP, who really have done three tiered applications, but with an immense amount of work. And so making it easy by building in a lot more to the operating systems, starting with the transaction manager. But that alone won't do it. We are really going to fulfill that original dream, and not just do it in a narrow sense where you have to go out and spend hundreds of thousands of dollars to buy some run-time for a distributed computing application, we're going to do it so that every copy of Windows NT we ship is a platform so customers of all sizes, across all their servers can have these applications run seamlessly, scale and get the reliability that the distributed vision has always contained.

Just two quick thoughts on Microsoft research. Of our R&D growth, the fastest growth is the research part. You know, we've given them essentially authorization to grow as fast as we can bring in the people, and we've done very well at that. We're just over 300. Today, we added the U.K. as a site, we'll add a site in Asia in the next 12 months. We've got some incredible people. We're even doing some stuff that's mind-blowing in terms of looking at physical models of computation as we have some pure mathematicians in playing around with that. 

We're very pleased with the publications we've got out. If you've looked at SIGGRAPH for the last few years, we've been the biggest contributor in some really neat stuff that we're sharing.

So, it all goes back to a vision of changing programming, making programming different than it is today. It's still far too manual of a task. Testing is far too manual of a task. Optimizing programs is far too manual of a task. Our factories, we can do secret soft type innovation and the whole process of how we build these programs that allow us to strengthen our leadership over similar people who are trying to do equivalent things.

At the same time, we can put a lot more intelligence into the product, you know, and get out there and make devices that really start to get to the potential.

So, I'd just really close by saying what Steve said, is that innovation is the key to driving this industry. We're willing to take the long-term view, like we always have. If we had a short-term view, we wouldn't be growing R&D like we are, because this is the first year we don't have those nice offsets in terms of other things going down that will essentially bury the R&D increase. You know, this year it will show up because it will be a stand-out in a big way as we move well past $3 billion in the spending there. We do see the breakthroughs that make that make sense, and we are absolutely still at the very beginning.