MicroTimes

On March 26th, 1992 Mary Eisenhart of MicroTimes visited Autodesk and conducted the following interview with John Walker. The following is a complete transcript of the interview, which lasted almost two hours and covered a wide variety of subjects relating to Autodesk's history, present, and future. A shorter version of this interview appeared in the May 11th, 1992 issue of MicroTimes. Although this is rather long, it's the best distillation of how Autodesk looked ten years into the adventure in the eyes of somebody who'd seen it from the start.

Q: It would be more standard, at this point in a company's evolution, for somebody in your shoes to have washed their hands of it and walked away. You seem to be doing rather the opposite—while not being chairman, not being president, and categorically denying that you're calling the shots, you nonetheless do not just leave Autodesk to its own devices and let it sink in the mud. At the recent press conference, you were a pretty forceful advocate for the new products and their importance. Why are you doing this, instead of going back to Switzerland and doing something else?

Well, I guess you have to ask the question “Why do people wash their hands of companies?” If you get fired, that's a pretty straightforward case. There's the case where you conclude that “the company wants to go this way, I want to go that way.”

Those are all things that I've certainly thought of many times. I'm always constantly reevaluating everything.

Basically, where Autodesk should be going is where I want to go personally. And as long as I can persuade Autodesk to go in that direction, I consider that my efforts are more useful, more effective, leveraged by having the resources of a company like Autodesk going my way, than they would be off by myself in my pump factory in Switzerland.

I'm in Switzerland because that's where our European technical center is. That's why Switzerland. In fact, about this time last year I thought our European technical center was going to be in Nice, in France, and so I expected to be there.

What I intend to spend the next five years of my life doing is working on automated manufacturing. I'm certainly planning to be doing that in my capacity as a programmer at the European technical center of Autodesk. If Autodesk recruits a new CEO, and the CEO's first act is to fire me, I'm going to spend the next five years of my life automating manufacturing anyway, okay?

It's what I want to do, and I believe it's the most important thing I could do for Autodesk. And that whether I do it or not, it is something that Autodesk has to do if Autodesk wishes to remain the strong force in the CAD business that it now is, because that's part of where the CAD business is going.

Coming over here entailed sacrificing a large amount of what I've spent the last five years of my life trying to achieve, in terms of having a low profile, in terms of not being identified with the company. Throwing that away, I've lost a large part of my life, and it's not clear that it's retrievable, at least not without another four or five years. I guess that's an indication of how much I believe in what we're doing here, in what we're attempting to do, and how much I believe I could contribute to achieving that.

Q: You've defined what you think Autodesk should be doing. What other likely directions do you see for it, whether or not they ultimately turn out to be pitfalls?

Well, I think you have to take an appropriately long view of the context we operate in. We are in several businesses, which I don't think you could find any competent thinker who would not believe are going to grow very, very rapidly over the next twenty, thirty, fifty years. If you are in a business like that, then you have to be a competitor who is constantly concerned about how much of that business is yours.

I believe that however you define market share, having the largest market share in the CAD business is an asset that is measured in the hundreds of billions of dollars over the next fifty years. I believe that having a commanding market share in desktop video is an asset that will be worth hundreds of billions of dollars over the next fifty years. Similarly you can just line these things up.

Now if a company focuses too much on the short term, on the margins this quarter, the margins over the next year, and sacrifices market share, that's a bad business decision. That is squandering an asset of the company for a near-term benefit. That's not in the long-term survival interest of the company.

And if you look at companies that are extraordinarily successful over periods of a century, you'll find, I believe, that they achieve that by taking the appropriately long view. The company I would cite who takes that long view is IBM. If you look at IBM, they are willing to lose essentially unlimited amounts of money in technology development efforts that are in their long-term strategic interest. That's what I think you have to do. If you look at the position that Microsoft is in, they don't believe this industry's going to go away next year, and they want to own as much of it as they possibly can, and as broad a base as possible. That's what any competitor has to do.

I think we have a tendency in this business—perhaps it's educated in business schools—to view competition as some kind of a zero sum game. It's not a zero sum game at all. This industry didn't exist in 1976. It all came from nowhere. And a lot of additional things are going to come from nowhere over the next three years, five years, ten years. The products that you're going to be writing about in 2002, probably none of them have been started yet. They're the exciting non-obvious things.

When you have a success, it's easy to assume that 1) it can go on forever, and 2) that that's all you need to do. I look at Autodesk, and I look at our rate of growth, and I look at Microsoft's rate of growth. I look at Borland's rate of growth. I look at the other players in this business, and I prefer not to think of it as a predator/prey relationship, but rather colonies of bacteria in a Petri dish. If one is growing at 75% a year, and the other is growing at 15% a year, then before long you tend to end up with all of #1. Not that #1 even bothers to devour #2, but rather that #2 is sufficiently irrelevant now that it just doesn't matter.

That's what really happened with Autodesk in competing with the major mainframe CAD vendors for the drafting market. We didn't really take seats away from them—most of the companies that had those mainframe systems still have them, in fact have more of them. What happened was, we created an entirely new market for CAD that was the other 98% of the business they weren't selling to, and that's where the growth came from.

I believe that that's what you have to do. I think it's crazy to say that the CAD market as it exists today is all there's going to be in CAD. In the long term, there isn't going to be drafting. In the really long term there isn't going to be drafting. But I think getting rid of it is going to take a lot longer than a lot of people believe.

Q: Especially since, as you say someplace, a lot of people are still doing it with pencils and paper.

It isn't even that so much. When we talk about eliminating drafting, we're talking about eliminating a language, a form of communication between the designer and the manufacturer, that goes back to the Pyramids. Certainly the Parthenon—there are places in the Parthenon you can see reference lines cut in the base.

It's a highly effective form of communication. It's a form of communication that is so evolved after thousands of years that it challenges the designer of the CAD system to better it. I can literally get a napkin, and about that fast, make a drawing that I can hand to a machinist who can hand me a part, with tolerances of a thousandth of an inch, to my specifications. Block, circle, just draw the dimensions, it doesn't have to be precise—that person will manufacture it for me. If I were to sit down today using the best solid modeling systems in the world, it would take me a comparable amount of time to build a model of that object as it takes the machinist to make it. Now if I'm the designer, which am I going to use? If I can scribble down a couple of lines and give it to some guy, and he goes off and does it, why should I manufacture the part myself so that now it can be manufactured?

That's the challenge we have. It's really as if we had word processors that worked with proofreader's marks still. As if we had all this computational capability, but basically you printed out a paper document, marked it up with proofreader's marks, fed it back into the scanner, and then an hour later you got another draft out, or something.

We've changed the way we write because we have word processors. We haven't changed the way we manufacture because we have CAD yet. That's not going to happen in five years. It's going to start to happen in five years. In isolated places you can see it now. In the very high-end military systems they're building fully faithful 3D models of the objects before they're ever assembled, they're doing automated manufacturing takeoff from these models, and on a project the scale of the B-2, they save money doing it. But for that technology to filter down to the machine shops, most of which are very small and don't have a large capital budget, is going to take a long time. And it's going to take pushing.

Those were the very people who, twenty years ago, were putting in CAD systems for drafting. So that's the filtering-down process, and I don't mean filtering down passively, I mean people like Autodesk pushing that technology out into the world, and creating opportunity for these people. Because it's fundamentally an issue of productivity gain. It's productivity gain, and that's survival in the manufacturing business.

Q: Is this one of the areas where the Cyberspace project would tie in?

It's central to why I encouraged Autodesk to start the Cyberspace project.

I recognize that all kinds of people say different things about this. To me, Cyberspace isn't William Gibson's vision, it isn't the global net. When I use the word, I'm simply talking about a 3D user interface. That is all I'm talking about when I use the word.

Q: I haven't seen it for about three years.

It hasn't changed much, in what you see in the demo. What underlies it is a lot more real now than it was then—real, and portable across hardware, and so forth.

It's a user interface technology. One of the things I wanted us to do two years ago, and maybe when we get the product finished we'll actually be able to do, is to have the world's first cyberspace race.

What I want to do is to take a simple mechanical part, like a universal joint for example—two Us and a couple of pins—and set up three workstations, as they were, side by side. One has a very conventional 3D CAD system with a tablet interface and a flat screen. The other has a nice shaded 3D fast rendering engine on a flat screen, and a dial box and various things you use in high-end mechanical CAD. And the third has the Cyberspace station, whatever technology we're using. The guy fires the starting gun, and three people, from all the parts of this thing, assemble it in space, which is basically positioning operations. Position this so it aligns with this, put this piece in here. This is something you can do on any CAD system in the world.

If we've got Cyberspace right, the guy with the Cyberspace rig is going to be able to do it in about ten seconds. It doesn't take long to put a U-joint together—you just pick it up and [whoosh] you're done. With a great graphical user interface—click, drag, position, pull down menu, dialog box, drag, align—it's ten times longer. And if you're working on a tablet with a 2D, you're typing in all these numbers and angles and aligning this and verifying that, and looking at it from a different angle….

That seemed to me to be the big win in Cyberspace. To me, the relationship between Cyberspace and CAD was the same as between drafting CAD and high-performance graphics. In the early days of the graphics business, back when we had Adage displays and huge Tektronix storage tubes and other things that cost large amounts of money, they were bought by the very first high-end CAD users. Because as expensive as graphics were in those days, there was a clear productivity gain in doing CAD with graphics over doing CAD without graphics. CAD was done without graphics. There were people that did computer aided design with FORTRAN programs and plotters. There was a whole generation of CAD that was done that way, that we kind of forget about. The graphics screen was an obvious huge productivity gain.

Similarly, I think the early adopters of Cyberspace are going to be people doing 3D modeling and 3D positioning. And ranging all over the design spectrum, certainly not just mechanical; architectural walkthroughs are an obvious application. Really, across all of 3D design. That's an application that can really bootstrap this hardware to the point that it becomes accessible for lots of other people.

I guess time has proven me wrong, by the way. That's why I thought [virtual reality] was going to happen in 1988, and a lot of other people were saying “No, it's entertainment that's going to push it out first.” And entertainment has. There's much more VR hardware both being developed and being used in entertainment applications than in design at this point. The design applications tend to be fairly arcane, and certainly very rare. You see some people doing some things like maintenance accessibility and so forth.

I guess one could also say that the VR hardware we have is so bad that a designer can't really use it now, and that's part of the problem.

It's like, you can play a game with a Nintendo, but you need better resolution than that if you're going to do serious design. And I guess we're kind of at the Nintendo level.

Q: If you're manufacturing sensitive parts, Nintendo is not what you need.

Right. And it's also a question of how complicated things can be. Part of what people miss about CAD versus other applications is—a 12-, 15-, 35-megabyte CAD database is large, but not huge. Designs are very, very complicated things. And the CAD user, in fact, always wants to have more and more speed and complexity, simply because they don't want to have 50 separate models that are subassemblies, they'd like to have it all in one model. They'd like to have one faithful model.

So the CAD user is always trying to get more CPU power, more graphics resolution, more speed, simply because of the productivity win of being able to get larger models in there. When you start talking about VR hardware that can't track in real time when you have 2,000 polygons—2,000 polygons is nothing in the CAD business. We're talking millions. Tens of millions.

Q: What is Alvy Ray Smith's definition of reality being a certain number of polygons?

Eighty million comes to mind, but I'm not sure. I think that's based on the human visual field, and basically if you assume one polygon per pixel of your human resolution, you end up with something like that. The intriguing thing is how close we are to that.

But of course, Alvy's reality is 80 million polygons at sixty frames a second or faster. We're a lot further away from that. (laughs)

Q: Where does multimedia segue into all this? Where do sound and motion become important?

Well, I never thought of sound and motion as media. I don't use the M-word. I prefer to describe what we do in our division that bears the M-word as desktop video.

I think desktop video is a huge business—again, it's one of those businesses that nobody is going to take seriously until the hardware appears that really does the job. And when the hardware appears, it is going to be so obvious that it's going to change the world. Vast amounts of money are going to be made, people are going to do things in different ways, people are going to learn new skills.

Today we're really at the Altair level of desktop video, just about. The hardware is something that an enthusiast can get very excited about, and with a vast amount of suffering can do amazingly impressive things—if you're another enthusiast.

We all remember back in 1976—“Wow! Look! It plays Tic-Tac-Toe!” “Wow! It computes the first eight digits of pi in six hours! That's really cool, isn't it?” “Yeah, yeah, that's cool!” (laughs) It took a while, and everybody who “got it,” to use Ted Nelson's phrase, understood that in five years these things were going to completely change the world. It would change the way everything was written—when you have a computer, you don't write the way that you write when you have a [dedicated] word processor. They understood that someday, modems would connect all these things together, and there would be a global community.

But that [laughs] took a lot of imagination. You had to be plugged in to what had been done on the high end, to see the potential, and you had to believe in something that I think for a lot of people is very difficult to believe in.

There aren't many times in human history that you can draw a straight line on semi-log paper over forty years, and have that trend continue. If you talk to people in a normal business and say, “Well, okay, yeah, a car today goes 55 miles an hour, but in ten years it's going to go half the speed of light,” they kind of say, “What are you talking about?” But here we say, “Well, yeah, sure, we design something [assuming] everybody's going to have 32 megabytes in their PC in three years.” I have 32 megabytes in my PC today, and it cost me less than my Altair.

It's really mind-boggling. And if you really accept that, you can buy into technologies that today don't look terribly promising at first glance.

Desktop video—how big is the market for this? Well, how many camcorders are there out there? You own a camcorder for what, an hour, before you realize, “Gee, it'd be really nice to be able to edit this stuff.” “How do I do that?” Well, you can't. “What do you mean I can't?” “Well, you can buy all this special hardware and it still doesn't work.” “Why can't I just do that on my PC?”

Well, why can't you, after all? You can go to Ted's houseboat and use his Aavid system. Clearly everybody's going to be able to do that on their PC. The Aavid's s a Macintosh, after all, with some special hardware in it. That technology is going to be everywhere.

It's a new art medium. We're talking about the ability to do Disney, Lucasfilm—choose what you like—3D animations, 2D animations, multi-plane animations. You don't just make a new art medium available to a lot of people and suddenly expect a million people to run out and start using it. I'm sure it took things like frescoes a while to catch on also—the technology has to get diffused….

Q: “Wet plaster? Are you nuts?”

[laughs] Egg gesso. Talk about low tech…. When you make a technology like that available, you can't expect suddenly to have everybody start using it. You sell the technology, and every trained animator in the world buys it. Okay, well that's a hundred and fifty. What about the next million? Well, you've got to get it into the schools, you've got to get books written about it, you've got to show people what they can do with it. You've got to give them something to start with so they aren't faced with a blank sheet of paper. After all, you tend to start with a coloring book rather than a notepad, and that's a lot of what we have to do to get it primed.

But in the long term, how can it not be something huge?

But I don't think there's going to be a multimedia market any more than there's going to be a graphics market. In the '60s and early '70s, there was the computer graphics industry—the National Computer Graphics Association, the SIGGRAPH show. Analysts would tell you how much market share each of the players had in the computer graphics industry—Adage and Sanders and all these people were out there, Tektronix owned a huge piece, CalComp owned a huge piece with their plotters. And that was the computer graphics industry.

There isn't any computer graphics industry any more. It's gone. It's in the box. Plug in the PC, and there's graphics. If it doesn't have graphics, it will in a year. So there's no graphics industry. Those people that sold that stuff in the graphics industry, for the most part, are still around, selling what's now the high-end stuff. But that's not the graphics industry. Graphics is part of the PC.

Pick a number—how many years is it going to be? Every PC is going be able to read and write sound and read and write video. It's not going to be as useful if it doesn't have that capability. And I think we're really there. Every Macintosh has had sound since Day One, and I think really every Windows machine is going to have sound very soon, except in large corporate applications or something like that. It's just a matter of time before the chip's on the motherboard.

We spent the first three years around here trying to convince people in this grand and glorious world that you needed a floating point coprocessor. We used to have a version of AutoCAD that ran without the floating point coprocessor, for longer than I either remember or would like to tell you. You plug in the coprocessor, and the product runs five times faster. Not until the 486 was floating point a standard part of the microprocessor—and what's the first thing they did? Take it out! Make a 486SX that doesn't have floating point, and then sell you this floating point that's the whole chip! It's crazy!

It's amazing to me that you can do anything at all without floating point. In the CAD business, basically every pixel you see on the screen has gone through a 4×4 matrix multiply. And how patient are you? Particularly when the cost is so low.

So we saw that with floating point. We've seen it with graphics being in the box. From the very early days, we sold very well on machines that came with graphics. For the first year of Autodesk, Victor 9000 outsold IBM, because the Victor 9000 came with 800×400 graphics right out of the box. And the IBM—hey, remember the CGA? First you had to buy one, then you had to get it to work, then you had to put up with four colors—

So really, once something is in the box, the software will appear. People will start to use it. And I think we're really getting to that point on the hardware you need for desktop video now.

But again, we're clawing our way up a very tall ladder where the rungs seem to be very well greased. We're going to get sound hardware on the motherboard, I expect, within the next generation of motherboards. We're what, maybe two or three years, from having the JPEG chip as a standard component. All these little pieces have to come together. But they clearly will.

I never quote numbers in these things because I'm worse than Reagan on numbers—quotable numbers, I mean—but one of the things we actually did market research on was what percentage of people who own a camcorder own a PC. And it's a huge percentage. If you told people that you could do useful editing, useful production titling, with your PC, a large number of people would, and that's a number in the tens of millions. That's the size of this thing.

There's this feeling that all the money in multimedia is going to be made in titles, in CD-ROMs. I'm not sure that's true. I think we may be trying to apply the TV model of the passive receiver of information, to something that's a creative medium. I'm sure it'll exist, but I'm a lot more interested in getting the creative tools out into as many hands as possible, because I think that's going to cause this upwelling of software.

Again, we've seen it before. How can you talk seriously about selling tens of millions of Pascal compilers and C compilers? Nobody wants to be a programmer! And yet people have bought tens of millions of Pascal and C compilers, and it's changed the world. Because anybody can program now. You don't have to work for a big company and do what they say. And look at all the software that's been created by these people.

That's what we need to do with these video communication tools. And we're just at the beginning of a terribly long cycle here.

Q: Do you see that dovetailing with CAD at any point?

Again, I think the CAD user will be the early adopter of many of these technologies. We're seeing it already. Clearly, architectural walkthroughs are a major application of 3D Studio. Product illustrations, product concepts, are places that CAD and multimedia/desktop video fit together very well.

A substantial number of our installed base are people who either already have AutoCAD or use AutoCAD within an organization, and I think that's a toehold that can make us successful in this market—that we can then broaden out, I trust, to the larger market. I think the parallel with Cyberspace is very similar here.

To get very far with this stuff, you've got to understand 3D modeling. And there aren't a lot of people in the world who understand 3D modeling, except CAD users. And so you don't have to bootstrap them up through the whole concept of building models, thinking in 3D, working in 3D, handling lights, handling rendering. They already know that. They're doing it. So to a CAD user, you're simply saying “All you're really doing is adding motion and dynamics” to something they already understand. If you're talking to somebody who watches MTV and would like to do stuff like that at home, you have a much longer learning curve.

Clearly the people that first bought PCs were people that already used computers at work. It took a while to broaden out. But it's inevitable that it will.

Q: You've said something I thought was intriguing—that there's a perception that CAD is a vertical nichey market, but that after all, there was a time ten years ago when spreadsheets were considered a vertical nichey market. How do you see the analogous transformation taking place?

Well, CAD is about building models of real-world objects inside the computer. That is what CAD is. I believe that in the fullness of time, every object in the world, manufactured or not, will be modeled inside a computer.

That is a very, very big market. This is everything.

Look at the huge successes in digital technology. It's basically been making things that used to be analog digital, right? In some finite number of generations we won't have to suffer through this [photo shoot] because there will be digital cameras that are good enough to do that. Sound went digital in the 1980s; it was one of the huge consumer electronics successes. Musical instruments went digital, with MIDI. I think we see on the horizon that television is going digital, with whatever technology they adopt for HDTV.

And every time you go digital, whatever the original motivation may be, you end up with a tremendous number of ancillary benefits. The fact that the data are in a form that one can process them fundamentally changes things.

What CAD is about is that artifacts are going digital. In Arthur C. Clarke's book Profiles of the Future, in, I think, 1963, he had this timetable of the future for the next 75 years. One of the things he talked about, I believe in the early 21st century, was encoding of artifacts. In other words, storing artifacts in digital form, and potentially replicatable form. That's what CAD is about. Today we're in very, very narrow areas of the whole field of being able to design and manufacture anything using your computer.

Again, it all happened before. We had a word processing business before we had personal word processing, and it was in a very vertical market. Large corporation typing pools, legal offices, insurance company client writeup—they had a word processing industry. All the analysts analyzed all the pie charts of market sector, and they missed the fact that there was a market for about another half billion word processors that weren't in the word processing market.

That's what's beginning to happen with CAD. We've seen it over the last ten years, but we haven't really seen anything like the takeoff yet. If you have lots of money and are very patient, you can get a stereo lithography system, set it up on your desk, build a 3D model of something, push a button, and after an embarrassingly long time, get a piece of plastic that is a 3D, faithful-within-the-resolution-constraints-of-this-device technology. That's desktop manufacturing, and again, that's worse than the Altair level right now in terms of cost and time and plastic that you use and so forth.

But I expect in the ultimate destiny of this that there's going to be an appliance that sits on your desktop that can make anything. I don't expect that to happen within the lifetime of AutoCAD Release 13 or 14. But that's where it's all going.

This week we're shipping HyperChem, which lets you design objects at the molecular level. The technology is going to go on scaling down until we have the same level of fine-grained control over the structure of matter that biology has been using for the last billion years or so. And at that point, we're going to be able to make anything from a digital model.

That's going to really change the world. It's going to change the world beyond comprehension, I think. Beyond, perhaps, the comprehension and imagination of the people who are spending a lot of time thinking about that technology.

And central to all of that happening is computer aided design. You can't do anything without building a model of it. If you're building a workbench, even if you're doing it in your head, you have a model of it in your head. Something tells you how long to saw off the plank, something tells you how many supports you need. You have a model of it.

Well, doing models in a computer is going to make the whole process much simpler, eventually. When we figure out how to communicate these things, when we have the right user interface technology, when we have enough compute power to throw at these problems.

As the microcomputer has evolved, we've been picking off the easy things. Word processing, the personal database, the spreadsheet. I think the thing that kind of holds what Autodesk does together is we're doing stuff that's going to use up all those compute cycles over the next 20 or 30 years.

If you go on drawing that line on semi-log paper, and you assume that in 1998 you have the power of a Cray 2 on your desk—what are you going to use it for? You're going to use it to do things that are fundamentally different from what you use your Macintosh for, and are the kind of things that very, very few people are doing with Cray 2s right now.

That's what we try to identify—bet on the technology going on, start to build these ambitious things that can take advantage of the computing power as it arrives. That's what we've done constantly in CAD. I think we see it in desktop video. We certainly see it in molecular modeling. Xanadu's great today, but it's about 100 times too slow. Well, time will solve that. And those are the kinds of technologies, to me, that are a lot more interesting than going out and competing for market share points in the spreadsheet market or the word processing market or the database market.

These are things that don't exist at all today. It's hard to keep focused sometimes, as the years pass, on getting these things finished, and to keep the belief that it's going to happen.

But we have an existence proof. AutoCAD was not viable at the time we started working on it. There simply wasn't enough compute power around to do the job. The graphics were laughable.

Time solved that. And if time hadn't solved that, we wouldn't be around. So those are the kinds of things that I prefer to look at

As a strategy it can't fail. The strategy can only fail if you don't believe in it or you don't execute it properly. Do we really think that this evolution of technology is going to stop in the next two or three years? That suddenly machines aren't going to get any faster or have more memory?

If that happens, you'll be able to see it coming because you'll be able to look at the fundamental technology and say, “We can't go any further without a fundamental breakthrough.” But nobody sees anything like that on the horizon for at least 20 years. There's a point where you have to look at quantum electronics and so forth. But even just the linear progression of the stuff we've got is going to carry us into the 21st century.

To me, plotting the strategy for a software company isn't very difficult. Particularly once you have a success like AutoCAD, because you have an installed base that is beating you over the head—“We want this, we want that, we want this other thing. We want you to do this. We also want this.” So you don't have to do an enormous amount of market research when you've got 600,000 people out there asking you for things that you don't have. It's basically a question of deciding what to do first.

When you have the financial strength that we have, if it costs you a quarter million dollars to do a product, and one in a hundred succeeds the way AutoCAD has succeeded, we can do a hundred failures for each success. The ratio of what it costs you to build and launch a product, even today, to what you make when you hit, is the thing that makes this business totally unique.

This isn't like the chip business. It isn't like the steel business. It isn't like the auto business. It's like the movie business. You make lots and lots of movies in the hope that, if it's a great year, you'll have one blockbuster, and if it's an unbelievable year you'll have two blockbusters. But you probably made 18 movies that year. And you may go two or three years making 18, none of which is a blockbuster—they may be profitable.

I think that's the kind of business we have here. It's a business that evolves rapidly, where preferences are difficult to predict, and yet failing to innovate is not the way you make a blockbuster. Making the 900th Western doesn't give you the kind of success that making Star Wars did.

That's what I always look for. I think that every company that tries that—unless something exogenous happens to blow them up or they lose their vision—they all seem to succeed.

Q: How does a company keep its edge, or its soul, or whatever you want to call it, in the face of overwhelming success?

I haven't solved that problem yet!

Q: I know. That's why I'm asking you, because you're one of the few people I know who seem to think about it at a meaningful level.

Well, no, I think if you look at companies who are successful over a long period of time, you find that there are people there who think about those things. I'm very sure you find it at Microsoft. I'm very sure you find it at IBM.

It's a terribly difficult challenge, because when you build a company, you have to be faced with the reality that as you grow rapidly, at any point in time most of the people you have working for the company are people who arrived after the initial success that created the company.

What is really necessary is to provide the message and the history of the company, so people understand it's not like it was always there. There was a time at which it wasn't there, and there was a time in which it appeared rapidly. That's why I did The Autodesk File. It was not really intended to be circulated outside the company—I seem to have that problem a lot [laughs]. It was really just a collection of stuff that people who were interested in where the company came from could use to find out, with the goal of leveraging that experience to people that arrived too late to have lived through it.

The other part, though, is really self-fulfilling, in the sense that if you go on having successes with new products, then you have cadres of people who have seen it happen. And so it is essential that you do what is necessary to make successes of your new products. That's what I'm trying to do right now—guarantee that every product we invest in will turn into a success, if it has the fundamentals to be a success. Again, I'm willing to fail. I'm not willing to invest in a product and then never give it a chance—pushing it off the loading dock, as I refer to it.

So this product [HyperChem] is going to be one of the biggest rollouts we've ever done. The AutoCAD Windows rollout obviously got us a lot of attention. And all of the other 29 products that we're rolling out this year will be rolled out in a manner distinctly different than the Autodesk tradition, and I believe that more than one of them will be a success.

That changes the mindset of people. You get into this depressed “everything's going to fail” mode. Everything doesn't fail. Companies have lots of successes. But you have to believe in it, and you have to make it happen.

It's not like the product comes out on the end of a conveyor belt and God looks at it and if a lightning bolt comes down you get all this money. [laughs] You take it off the end of the conveyor belt, you go out in the world, you start jumping up and down and waving it in front of people, and finally convince them that it's worth giving you their money to make you go away [laughs]. And then they take it home, and “Gee, that's really cool!”

But you have to keep pushing. At Autodesk I think there was a mindset that the machine runs all by itself. That it's just magic. If you look at the companies that advertise and promote most heavily, they're the people that make this [can of Diet Coke]. They're Kellogg's. You're talking about products—in the case of Coke Classic, the product has not changed in over 100 years, and the one time they changed it they regretted it. Kellogg's Corn Flakes—how often do they re-engineer that product? Release 3.7? No, no, no, sorry, it's Kellogg's Corn Flakes. And yet they spend billions of dollars promoting those products, because they believe that in the year 2700, if there are still human beings, they will still be eating corn flakes, and they will still be drinking Coke. And they want to own that market, or as large a part as they can.

It's not a self-feeding process. You have to explain the benefits of your product to people. And you have to do it on an ongoing basis as you improve the product.

Q: A lot of it seems to be mutually exclusive loops—you've got the people developing the product, you've got the people marketing the product who are completely clueless about the people who are developing the product, and so it goes all down the line.

Well, you have to, at some level, just have some basic rules, Okay?

I feel kind of embarrassed because I often use the phrase “you don't have to be a rocket scientist to figure this out”—and here I've got a product [pointing to HyperChem box] you really do have to be a rocket scientist to use [laughs].

But I never used to understand, back when I was a big-computers kind of guy, the concept of the salesman who didn't understand the product. The salesman who couldn't demo the product. The programmer who didn't know the names of the competitors of the company, who wasn't familiar with the competitor's product.

To build a company that's a competitor, you have to have everybody be a little microcosm of the company. You've got to have the programmers be intimately aware of what causes the purchaser to choose their product or a competitor's product.

Now that may be marketing, or that may be sales, but if you're a developer and you don't know that, how do you know what to develop? How do you know what your target is, what your priorities are? It's not looking at the competition to be your guide, it's knowing what's in the customer's head. Knowing who the customer is. Knowing how the customer obtains the product.

Similarly, if you're out selling the product, the people you're talking to who are buying the product know about your product, they know about the competitor's product, they know about where the market's going, and if they ask you questions about things like that, and all you know is how to fill out the sales form, how are you going to sell?

So I think that you really need to imbue in the whole organization an in-depth knowledge of what this company does. I think it's got to be across the board. That is part of the reason that Information Letter 14 was so long, and that I wanted so many people to read it. It was because everybody who works here, here in Sausalito and around the world, needs to understand this business we're in, the industry we're in, and what's happening to it over the next ten years.

Information Letter 14 was not intended to be a secret—no more than the first twelve were. It got out in an earlier version—in fact the one that most people have read is a draft, not the final one, because that's the one that started self-replicating. Some things changed—there were some things that I thought weren't right in the first draft.

But it was inevitable that it would get out, and I intended it to get out. There wasn't anything that I said in there that I had not heard from analysts who covered the company. There weren't any secrets in it. It was really just a view of what the company had to do, that I think you cannot communicate to five people and change the outlook. You have to communicate it to everybody who works in the company, or else you have no chance.

But I certainly learned a lot about publishing. I mean, I sweated blood over this thing [The Autodesk File] to get this book put together. Hey, I just leave something on somebody's chair, and before you know it it's all over the world! I guess I'll do that with my next book, The Hacker's Diet: How to lose weight and hair through stress and poor nutrition.

Q: Where does AMIX fit in?

AMIX is, again, I believe, one of these little hard-to-define, hard-to-explain things that, once people experience it, is going to be so obvious that it is going to grow very, very rapidly. I really believe, of all the things we've done, AMIX has as much potential to change the world as anything we've done. And even having watched it be built, heard the rationales, watched it be put together and grow—experiencing it is fundamentally different than what you expected. It's like you can imagine what it's like to be on the WELL, but it's very different to be on the WELL.

For years I tried to imagine what it would be like to be able to read netnews. We were never on netnews, and then finally we were—and it was nothing like I expected. And that's really what AMIX is.

I've sold stuff on AMIX—actually I have a net profit in my account now—and it's very intriguing to realize, “You know, I do things very differently, even if I'm going to ask two dollars for it, than I do when I'm just posting something to a newsgroup.” I put an entirely different amount of work into it, I'm much more concerned about my reputation and the comments that are going to be put on this thing. I make tremendous effort to target the introduction to be both persuasive and fair, so nobody will buy it that isn't going to be satisfied with it.

The anonymity of AMIX is very interesting. Like being in a bookstore and knowing that you've sold 22 copies of something, but not know who bought them. They can communicate with you, but the fact that you bought something does not generate a mailing list in the same sense as a bookstore.

And similarly, I think, when you're on it as a buyer, you recognize that the sellers are thinking that way. The sellers that do not think that way are going to be selected out very rapidly, because if somebody's asking you for $15, a book-type price, you're going to read the comments file. You're going to look at the resumes of the people who put those comments there.

Again, the “vision thing” is being able to imagine what something is going to be like before it exists. I've heard that from Ted over and over again, that that's a very rare talent. I really think that Phil Salin knew what AMIX was going to feel like, at least after he'd built the prototype. I think only a very few people really knew what it was going to feel like.

And then once it's built, after that vision, now I think people will evaluate for themselves the benefits of it.

But I think part of it is just fundamental economics. Phil was an economist. He understood economics. He understood transaction costs. He understood the way purchase decisions are made. The trend in price in the software business has been down, down, down, down, down. And we have a means that you can sell software for $2, $5, $10, at a lower transaction cost and a lower distribution cost. The rules limit this, but you could sell software for $50,000, three copies a year, and have a very low transaction cost and a very low support cost. Companies that sell software for that kind of price tend to have large organizations that could largely be eliminated if they could deliver their updates and support and consulting electronically.

These are all things that AMIX is going to be getting into. We're getting a lot of interest in AMIX purely as a fulfillment mechanism, of being able to sell an update. Your update to Product X, Release 4.0, for $10.

Part of the problem we have to climb up against—and I think Prodigy actually helped a lot in this—is that surprisingly few people have modems. We looked at some of the demographics, looking at AMIX, and it astounded me. It was a shockingly small number.

Slowly people will realize the benefit, but they're not going to buy a modem until there's something that they want to talk to.

Q: And by the same token, a lot of people won't buy Macs until the Photo CD is out and the Mac reads them. All these years I've been telling my parents they don't really need one, because they're not into the toy aspects of it at all. But as soon as they can put three generations' worth of family pictures on a CD, then everything changes.

Mm-hm.

Q: Are the AMIX guys also Xanadu guys? I always see them in the same place, so I guess I've got that perception.

No. No. They share a parking lot, of course, and they share a lot of the same intellectual groundings and ideology. But there is no connection between the two companies. They are two independent companies in which Autodesk has unrelated investments.

A lot of the connection was that when Phil started to think about AMIX, he became very interested in Xanadu, and saw Xanadu as the ultimate system that would underlie AMIX. It would really be the first system that would have the strength to do all the referencing and linking that he would like to do.

In the early days of Autodesk's involvement with Xanadu, he was a very strong force in causing the deal to come through, getting Xanadu organized. He probably spent more time on Xanadu the first year than AMIX. But we concluded that AMIX did not need Xanadu beneath it. An AMIX with Xanadu beneath it will be something very, very different and much better than what we have today, but what we have now could be built as a conventional program, and it needed to be built as soon as possible without waiting for Xanadu in order to succeed.

Q: I take it you think Windows is pretty important.

I think it is of surpassing importance.

What is happening is that workstation-class operating systems are becoming available for the masses. Workstation-class operating systems I really define as a system in which graphics, and soon 3D graphics, are tightly integrated within the system. It has a modern user interface. It has true multitasking, and it networks well. That's what you really get today when you buy a workstation. You have to suffer a tremendous amount when you buy a PC, to stack pieces on top to get all of that stuff to hang together.

When that solution becomes available, the software environment is going to change in a very substantial way. You can certainly say, “Well, what about the Macintosh?” Well, before System 7 the Macintosh was not a workstation-class system. Only with System 7, which I guess a lot of people still aren't using, did the Macintosh become a workstation-class platform. And with Windows NT, we will see the PC become a workstation-class platform.

Suddenly all this capability that people have not been able to afford is going to be available. And it is going to change the way everybody builds their applications, buys their applications, and uses their applications.

I don't know for sure, and at some level I don't care, who wins—whether it's Windows NT, or, should Microsoft stumble, an OS/2 32-bit version. Suppose both Microsoft and IBM stumble, and Sun Solaris for the 486 ends up filling that niche. Or all three of them stumble, and OSF/Motif or something like that becomes a standard. Whoever wins, it's going to happen.

It's as inevitable as anything that I've ever seen in this industry that's going to happen. The benefits are simply too huge, and the costs are going to be negative, really, in the sense that when you go into the stores a couple years from now, there isn't going to be much raw DOS software for sale. It's all going to run under the winner of this operating environment battle.

Any software company that doesn't really understand the true significance of this is not positioned to be a survivor across this transition. You have to anticipate both what's going to happen—make some bet on who's going to win, certainly, in terms of where you spend your time—but most of all recognize what the expectation of the user is going to be. Because the user who has a multitasking, multiprocessor, high-resolution system that's networked to lots of other people is going to be using your application in a very different way than the guy that's looking at a black bozo screen one application at a time.

That means that you have to start making fundamental changes in how you build your product, in what you put in it, in what capability it provides to the user, that are bets on that occurring. Because if you don't start now you're going to be too late.

Q: So you think that, nice interface aside, the main advantage of Windows is that there's so much of it out there? Or does it have other particular strengths as an operating environment?

Well, I think that one cannot ignore the impact of something that has sold nine million copies. Market share counts for a lot, and it counts for a lot when the incremental cost is virtually nothing. If you have a PC with appropriate hardware, it costs you almost nothing to get into it. You're not making a purchase decision like going to a Mac or going to a NeXT or something like that. You almost can't afford not to.

After all, if you want a really nice word processor today on your PC, you've got to put Windows on it. Because all the word processors for Windows are a lot better than any one that doesn't run under Windows—that's my experience. All the nice spreadsheets run under Windows. So you have to do it.

Now when that standard is created, we're going to have software that can move upward from all the software that's being built for that environment to systems of workstation class and above. NT is going to run on symmetrical multiprocessors. Sun is just getting out a system that runs on symmetrical multiprocessors.

I don't make predictions about hardware, because I'm always wrong, and I don't know very much about hardware, even though I used to have a hardware company. But it's not inconceivable to me that in 36 months a machine with fewer than four processors is considered as passé as a machine with no more than 640K of memory on it. After all, the chips don't cost very much, and they're plummeting in price. And if you have software that runs six or eight times faster if you put in six or eight processors, you're probably going to put in six or eight processors, if you're interested in going fast.

This software [HyperChem] will run eight times faster on an eight-processor machine than it runs on a one-processor machine. The SGI version demonstrates that. We're going to be shipping that one later this year.

Dr. Neil Ostlund, who designed this software—he is a professor both of chemistry and computer science; he learned computer science in order to parallelize computational chemistry. You don't just sit down and do this stuff; it isn't easy. It's like you've been writing COBOL business applications for 45 years and then “here's a Macintosh, design a GUI.” Well, it requires a little head-scratching and a little learning.

But application vendors who wish to play in this environment are going to have to learn to do it. Particularly if, as in the CAD business, you can use any amount of power. We could saturate a machine that was a thousand times faster than anything you can get today, just doing the same kind of stuff we're doing now. So we'd better be able to take advantage of it.

Don't forget the multiprocessor/multitasking aspects of this. I think it's really a very important thing. And all the major players are going to provide it. It's going to be essential.

Q: In niche markets, or everywhere? For what I use my Mac for, I don't see the need for multiprocessing, but will that change?

Well, are you happy with how fast it runs now?

Q: Depends on what I'm doing. For text, who cares? But for PageMaker, I go crazy because it's so slow.

Okay, well, if you go crazy about how slow it is, and the Mac toolbox were running on one 68040 and the application were running on the other, it would immediately run twice as fast, assuming that the breakdown were about half and half.

I just got my 486 this week. I've been using the same 386 20 MHz machine for, I guess, four or five years. And for everything I do with the machine except CAD and HyperChem, I was quite happy with how fast it ran. Yeah, it took about seven hours to compile all of AutoCAD, but I only do that every couple of weeks. Besides, it can do that overnight.

But except when I was really beating Excel to death or something like that, or editing a 175-page document, I was fine. But as soon as I start to do CAD—“Oh, this thing is so achingly slow!” All the floating point-intense stuff, huge databases, lots of operations where you're searching through a lot of stuff to try to find things….

Q: And not even doing 3D motion yet…

That's right! It's basically 2D CAD for the most part.

Our company sold a million copies of AutoCAD and lower-cost CAD. That's not a niche to me. If it is, I hope there's a lot more! [laughs]

I think, yeah, people are going to be doing things that are fundamentally different as well. But they're not things that you can't think of. It's simply things like photo composition. Photo composition and retouch from your Photo CD. You can use a lot of compute power when you get into the really serious image composition type stuff. Video throws another factor of 1000 at it. When you have a machine that lets you cut and paste moving video, you're going to probably use all the processing you can get, and all the processors you can get, to do that.

The other thing that's happening with these workstation-class environments is that for the first time, applications can live beyond the hardware. I have now had pieces of C code that have run on three different processor architectures from Sun, to which I did nothing other than simply recompile. I started on a Sun 3, 68020. I moved to a Sun 386i, 386 architecture. And now I have a SPARCstation. And I did nothing. I just use the code. I don't even think about it. It's inconceivable to me that as long as there is a Sun, I won't basically be able to take my code and move it along.

Lacking that is something that always seemed to me so foreign about the PC business. Because I grew up as a UNIVAC mainframe programmer. I worked from 1967 through 1978, and actually I did consulting into the '80s, on machines that ran programs that were compiled in 1963. Think about that. Basically children running code that was compiled by their fathers. IBM mainframes have been upward compatible through the years—you have to look at the DOS to OS bump in the '60s as an aberration. Yet here it seems like all your investment is a wasting asset that disappears after 24 or 36 months.

You talk about your parents putting all their pictures in the Photo CD—I hope they keep the pictures, because in the year 2700 you're still going to be able to look at the pictures, but what's going to be around that reads the Photo CD?

It makes me sick to think of all these people putting their 8mm movies onto VHS tape and throwing the movies out. And what is there that's going to accept a VHS tape in ten years, or fifteen years?

If you look at the heavy CAD systems, the CADAM, CATIA, Computervision, and the rest of the design software that isn't what you'd call CAD—the structural analysis, NASTRAN, ANSYS—these are packages that are approaching 25 years of age, that have massive investment every year over 25 years, of doing more and more and more. Larger models, better user interfaces. When you make a commitment to something like that, that's a long-term commitment. You and the vendor are taking a very profound step.

Because in engineering, we talk about things that span decades. The Boeing 777 is going to be flying in the year 2050. It'll probably fly till 2075 in the secondary and third-tier market. And they're going to be manufacturing replacement parts for it; they're going to be upgrading it; they're going to be stretching it; they're going to be replacing the avionics and all of this stuff, but that's a design that spans a human lifetime from conception to obsolescence. And you don't want to buy into a design system that's going to disappear somewhere during that cycle.

So there is a somewhat different perspective. This is engineering that we do here. Engineering is unlike a lot of other businesses because when you get a wrong answer people get killed. And when you're educated to be an engineer, you take a very different view of quality and accuracy and so forth, that's kind of indoctrinated in one.

If your spreadsheet gets a wrong number, you may be embarrassed. If your word processor crashes, you're going to be furious. But if a CAD system produces the wrong strength number for a load-bearing member in an aircraft wing, airplanes start falling out of the sky. And that's really, really bad.

Sometimes we test to death. We're accused of really bashing on the quality of the product. But it can't get wrong answers. It's a fundamental thing. You can't do it in this business.

And so yeah, we do tend to take a view where you have a long-term relationship with your customer. Your customer is sitting there having, in some cases, many employees spend their whole working day putting stuff into your system and storing stuff in your format, and accruing capital. And as long as there is a CAD market, as long as there's an AutoCAD, it's going to read every AutoCAD drawing back to Day One. However different the environment. However different the user interface.

Windows and these other things give applications that sense of immortality. In the sense that once you buy into it, clearly there's going to be an upgrade path. When we get full-motion video integration, when we get 3D viewports of the models, it's going to be done in device-independent way.

I think we need only look at the integration of sound, that you can see both in MIDI Manager on the Apple first and in the Windows Multimedia Extensions. It used to be that if you wanted to do anything with sound, it was all hardware-specific. Anything you did was obsolete in six months because the board that you designed it with wasn't made any more. Even if you had three systems, they were all different, because you simply couldn't buy the parts that you put in the first system, even if you were happy with something that was obsolete.

And now you write something for MIDI Manager on the Mac, you write something for Multimedia Extensions, and you know it's basically going to work for a very long period of time. Because the drivers have been decoupled from the application.

And that's really what these systems do for us as application vendors.

Q: Are there areas you'd like to be in that you aren't?

Oh, hundreds. Hundreds. There's no lack of opportunities in this business. There's just a lack of the time to do anything about them.

Q: These damn finite carbon units….

Well, yeah. It's lack of attention to do it right, I think, that's the most obvious thing. I think we could define 25 products that our existing AutoCAD installed base would love to have that we're getting around to one or two a year at this point. If you look at opening up new markets, there's absolutely no problem coming up with ideas.

Most of them are bad, but again, it doesn't cost very much to find out that they were.

We'd like to do another 35 titles in the science series. We're trying to make our science series like the Time-Life Books science series, except dynamic living things that you can do real things with, rather than presentation in a book that becomes dated. We can only do about one a year right now; you don't find guys like Rudy Rucker standing in line at the unemployment office. [laughs]

So there are those constraints. But there's never any lack of ideas in this business.

Q: One of the good things about Autodesk as a company, it's always seemed to me, is that there's always been the perception at key levels that the company's crown jewels weren't some code sitting on somebody's hard disk somewhere, the crown jewels were in the heads of its people.

Mm-hm.

Q: Do you feel like that's an accurate perception, and if so, how do you sustain it in the face of corporate inertia?

I think, in a software company, you only really have two assets. You have your position in the market and your installed base, and you have what's in the heads of the people who work for and with the company. Because what you're selling today is not going to exist in 36 to 48 months.

I think everybody in this business totally rebuilds their product, even though it has the same name, over that period of time. You don't start from scratch, you simply replace. If you start replacing 20–30% every release, before long you have a totally new product. And what determines what that product is in the heads of the people who are working on those new releases right now.

It is of surpassing importance to make sure that you have the best people you can possibly get. Information Letter 14 was written assuming that people had digested an earlier 25-page memo that never got out that explains this.

The nature of a business is inextricably related to the constraint upon its growth. In other words, what is the constraint that keeps a company from growing faster, doing more, whatever. And that varies from industry to industry. In a capital-intensive industry, it's how much money you have and how much money you can raise. Lots of businesses are like that. Some businesses are natural-resource-limited-how many trees do you own? How many trees can you cut down a year, if you're a paper company.

The software business is constrained by talent. Human talent. How many talented people can you get to work for your company? How many talented people can you develop from people that come into the company and wish to be trained? And that's not just programmers, by any sense. It's people throughout the organization.

You need talent everywhere in the company—it's that broadly based talent, of being engaged in the industry. In the software business, you're not investing heavily—we didn't even own a forklift for the first five or six years. There's no capital equipment. Even the PCs don't cost much—they cost like a guy's salary for a month.

You're investing in people. And if you don't invest in developing the people you have, in getting the right people, you're going to fail. It's just that simple.

This isn't Marin County airhead mush, it's just good business. If your asset is people, and you don't invest in the asset, you're crazy. It's like you're running a railroad and you don't invest in the rails—after a while, the trains start running off. It's just stupid.

So I think you simply have to do that. And if you make that investment, then the company can grow over a period of time and continue to succeed.

I think a lot of the problem that you have in business, and also in politics and so forth which I'm totally not interested in, is that it's always easy to neglect maintenance right now. The impact is always deferred well past the current horizon, or perhaps even somebody's retirement date.

If the railroad starts to get in trouble, “well, let's just stop fixing the tracks.” And you know, for the first three years nothing really bad happens. And then, unfortunately, that guy retires and a new guy comes in, and the trains start running off the rails, and “oh gee, it's going to cost 4 1/2 billion dollars to replace all the tracks on the railroad, and we don't have that much money.” And then you're stuck.

Similarly, if you take a typical software company, you can fire all the programmers, and probably for 24 months nothing goes wrong. You make all kinds of money. And then all your competitors come out with new releases and you say “Oops! Guess we should have had a new release….” And then you're stuck.

It's just sound business, in a business that is talent-constrained, to do everything you can to attract and retain the talent that you have. Because that's really all you have.

Q: Given that you can find them, what tends to keep talented people interested and involved? I love your thing about “Make the best product. No bullshit. Reward the people that do the work.” But what does that mean in real life?

I think one thing that people don't pick up when they talk about Autodesk—because Autodesk kind of had this image, at least, of being a very far-out-on-the-leading-edge kind of company, in some ways—is that of all the guys who founded Autodesk, we were all maintenance programmers. We weren't Xerox PARC researcher type guys. We were guys that worked on these big operating systems that live for 25 or 30 years, out in the field, fixing bugs.

We were maintenance programmers. And the culture of the maintenance programmer is very different than the culture of the research guy. The maintenance programmer works directly with the customer. Maintenance programmers recognize that someday they're going to retire and their children will be maintenance programmers, who are going to have to inherit this thing.

But really, it's more of an ethos that is that of engineering rather than that of a writer or of a moviemaker or something. It's recognizing that the thing has a life cycle, that there are lots of people involved in it, that if you write something and don't adequately document it so somebody can pick it up, you're not doing your job. You're fundamentally not doing your job.

And we try to take that approach to everything we do. We certainly rely on individual talented people, and one of the things that's unique about this business is that there are individuals who are hundreds of times more productive than other people who are productively employed in this business. When you find those superstars, great. But you can't build a company out of all superheroes, because there aren't enough superheroes to go around.

Q: And besides, they don't necessarily get along when you get them.

[laughs] That's right.

If you're fulfilling the needs of real customers, you have to be able to do it with a product cycle that fits the customer's expectations. You have to meet the needs of the marketplace, you have to evolve onto the new hardware that's coming along. And that means, in a large developed market like the market for AutoCAD or for spreadsheets or for word processors, you have large teams of people doing lots of things at once, doing lots of things in parallel. And that's management. You need to have management that understands how to run big engineering projects.

It's not the most exciting thing in the world to talk about, but if you want to be in this business you've got to know how to do it. You can't design and manufacture cars with one bright guy that knows a lot about cars. It takes a lot of people in that cycle, all the way through. And that's the way we have to build software, when software begins to become as complicated as cars.

People say, “Well, wait a minute. A car isn't very complicated.” But these are people that don't cut metal. I remember years ago I heard George Morrow speak, and he said, “You know, I thought this computing stuff was really complicated, but you've never really done anything until you've punched a hole in a piece of metal.” He was talking about his adventures making the cover for the first one-megabyte disk drive that he sold, and getting all the holes to line up right, then going to a second supplier and they punched the holes in the wrong place…. And recognizing that there's no Undo command once you've punched a hole in a thousand pieces of sheet metal. [laughs]

It's a perspective you have to have. Particularly as the user expectations grow. When you go into Egghead, and pay $35 for a piece of software, you expect to have something that's got a large number of man-years of work in it. I may say your expectations are ridiculous [laughs], but they're yours, and you've got the money, so I better conform.

I think part of what's happened is that the kinds of things we vendors had the audacity of selling for money four and five years ago are what you expect out of a public domain program now. So if we intend to go on asking people to pay money for this stuff, we're going to have to really give them something that's worth it.

And that's a big job. Again, when you talk about the big CAD systems, the mainframe CAD systems, those are systems that are 50 times bigger than AutoCAD. We have a long way to go just in that line, and our users can use every bit of that capability just as soon as we can get it to them.

But those are big engineering projects that have to be done in a rigorously professional manner, or you're going to fail. If you talk about building a system that has conceptual design on the front end, drafting in the middle, manufacturing on the back end, it integrates to your manufacturing control system, it complies with the ISO 9000 specs for quality control that are being enforced around the world now, that communicates to other CAD systems in the PDES product description language that is now becoming standardized—you end up with a couple hundred people working on these things, and all the pieces, and they've got to plug together when you put it in a customer location.

Q: How do you avoid the design-by-committee problem when you do that?

Well, at some level I'm not convinced there's a problem with that. You certainly look for inspiration, but also you have to get there from here. In the engineering world, many things are designed by committees whether you like it or not. The IGES standard—you can't sell a CAD system unless you can read and write IGES—was designed by a committee. You may be on the committee, but all your competitors are on the committee, and government and industry. People are on the committee, and they hand you this thing and then you go and implement it. We don't operate in a domain that affords us unlimited freedom. ASCII was designed by a committee. Unicode was designed by a committee. These are standards that you have to conform to.

When you go to build your own application, I think you avoid that by having those Wild Talents, and by doing a lot of prototyping, getting a feel for how something is really going to look, and then go out to the customers. Companies that aren't out there with the customers are doomed, I think. The customers have a lot more wisdom—when you've got half a million customers, there's a lot more cubic centimeters of neurons firing outside the organization than inside the organization. And it's up to you to really be out there communicating with the customers, looking at what they're doing and learning from them.

Q: Some people would have it that being engineering-driven and being customer-driven are diametrically opposed.

I think that's nonsense. I think that any business that isn't interested in meeting the needs of people in the real world is somebody's personal playpen and not a business. When I say meeting the needs, I mean meeting the needs of the people that are currently buying your products and the people that aren't buying them yet, or aren't buying the stuff that you aren't building yet.

You've got to be customer-oriented. I think every business is a service business. I believe that firmly, that fundamentally when you ask somebody to give you money, you have to be giving that person something that is more valuable to them than money.

Think about that. That's the decision you make. Even when you go into the video store, you're saying “Being able to watch Terminator 2 tonight is worth more to me than having $1.50 in my pocket that I can do anything I want with.” Or even waking up tomorrow with $1.50 in my pocket, and then I get to decide “I want Arnold tonight.”

That's what you're doing. And in the software business, it's even more of a service business, because when you're buying a piece of software, you're not buying something that's disposable. You're buying, generally, a commitment to something that you expect to be updated and upgraded, that you expect to preserve the investment you make in hammering parts of your life into this thing.

A software business simply has to respond to those expectations.

Q: It seems that a lot of things you're talking about—long-term strategy, diversification, keeping your edge—pretty much flies in the face of the way business is done in the United States.

That's why I don't do business in the United States any more. (laughs)

Q: So do you think there's hope for business in the United States?

I am uninterested and uneducated in the issues of politics.

Q: I was less interested in politics than in what I've heard coming from the mouths of Mitch Kapor and John Roach and others, that the obsession with the quarterly bottom line is killing American business. Do you find that to be true, or not?

I guess that has killed a lot of companies. But I think it's been, perhaps, used as an excuse by other companies for killing themselves in other ways.

Meeting quarterly figures is an expectation of the United States public stock markets. But the responsibility for the long-term destiny of the company does not rest with the Wall Street investors, it rests with the senior management of the company. If the company has an asset value of a billion dollars, say, like Autodesk used to, and its management squanders that by focusing on the next ninety days, they are not doing their job.

There's a point at which you have to go to Wall Street, and say, “We're going to disappoint you. Because if we don't invest in this business, we are not doing our job as management. We're abdicating our fiduciary duty to the shareholders, to invest in the long-term future of this company.”

If you look at companies, you will find times when IBM has gone to Wall Street and said, “We are going to turn in disappointing numbers because we need to develop this new technology. We need to build a 16K RAM plant. We need to have this technology—we need to invest in Josephson junctions or whatever we're doing.” That's a decision that they make based on the assumption that twenty years from now, thirty years from now, there will still be an IBM, but there won't be if they don't make the investments they need.

You have to manage the business. You have to sometimes make Wall Street happy, and sometimes you have to make them unhappy. But you're not doing your job—where you fail is when you let the Wall Street analysts make your decisions for you. You have to make your decision, and sometimes, if you're a public company, that means you have to go out and take a lot of heat. But it's better to take the heat up front than it is to take the heat down the road, as surely it is coming if you don't make the investments that are necessary.

On balance, certainly I wish we'd never gone public, looking back on it. It would have given us a lot of freedom, and we didn't fundamentally need to go public. When we made the decision in '85, it was the right decision, because in 1985 we had a company that could run just like a couple of weeks if our sales went to zero. And we expected, within a year, to be competing with IBM, AT&T, ComputerVision, and Intergraph all at the same time. And doing that with no money is kind of a sobering prospect.

So from that context, we really had to do it. The fact that in retrospect we never really needed to is not something that you can undo at this point in time.

I think you do have to ask the question “Is the environment that you're in the environment that is most conducive to long-term success?” And for Autodesk, we're not really coupled to the US, because substantially more of our sales come from outside the US than come from the US, and I expect that to continue. The US is less than a third of the global economy, and you should not be doing a huge percentage of your sales in the US. And we don't.

We do more business in Europe than the US. Particularly since our CAD business is very closely tied to the manufacturing business, you expect a large part of your revenue to come from those portions of the world where manufacturing is basically done, and that's not the US right now. Certainly Europe or the Far East are the meccas of manufacturing, and I don't see that trend reversing in the near future. So being strong there is very, very important to us.

But I think you also have to constantly reevaluate where you're going to make your investments, where you're going to do things. Up until last year, we did all of our software development in the United States. Now we have a software center in Neuchâtel, where I work, and we're moving a substantial amount of our new software development and localization there. I expect before too many years pass we may also have a software development center in the Far East.

Q: Are there qualitative differences between the European and Far Eastern markets, as to who your main customers are?

Yes. The Far East tends to be much more of a very-large-company, very-large-account market, whereas in both Europe and the US we have been selling in a very bottom-up fashion to individuals and small customers. Not that there aren't a tremendous number of individual customers in those markets—it really isn't five big companies running the whole thing. But being in those big companies is much more important, because they tend to set the standards for the people that subcontract for them, and the rest of the economy.

Q: A lot of vendors go into, say, the Japanese market, where there's about twelve different operating systems running on PCs and have a hard time getting past that fact—“Do I write for the NEC, or DOS/V, or AX, or…?”

NEC has 70% of the market. That's a pretty easy decision right there. (laughs) NEC is in a stronger position than IBM ever was.

I think—I get nasty after two hours—I think a lot of companies fail because they're arrogant, ignorant American companies that think they can do business in somebody else's culture the way they do business in California. If you go in, and you're trying to sell somebody a word processor, and it doesn't use their character set, you're probably not going to get very far.

Hey, we're just a CAD system vendor, but we got a bunch of guys to sit down and draw 80,000 characters. If you want to play in Taiwan, particularly, you've got to have traditional Chinese. You've got to have your commands, and the vector fonts that draw on the screen for CAD, and it's a big job. Think about what happens when all those little 256-character tables now have 80,000 characters in them, and there's a whole disk full of just fonts, right? It's a real mess, but if you want to play the game, you've got to be in it. You can't sell cars with the steering wheels on the wrong side, you know? [laughs]

You have to make that investment. Typically the way we organize is that there's a country manager for each country, and that country manager is running the business in that country. They decide which of our products they will sell, what price, what distribution channel, how we do our support—they're the president of a company. Autodesk happens to have a headquarters in Sausalito, but Autodesk is an Italian company, and a French company, and a Russian company, and a Swiss company, because the people in those countries run the Autodesk business there, and they basically make the decisions.

They have to do that. I mean, the thought that you could sit in Sausalito and decide what the price and distribution strategy for Thailand is going to be—it's just crazy. Even if you could do it, it would be ten times harder than having somebody there who speaks the language, who knows the dealers, who visits the customers, who knows who the competition is. Because typically when you go into other countries, you're not competing against three other American companies; you're competing against a whole bunch of indigenous products.

Again, you just have to be sensitive to what people want. The European manufacturer does business in a very different way than a lot of American manufacturers do. And if you want to sell to them, you simply have to solve their problems.