An interview with John Walker - Part 4
This is the fourth and final part of the series documenting my recent interview with John Walker. Where the first three parts looked at the early history of the company, the architectural evolution of AutoCAD and its APIs and Autodesk's strategy and eco-system, this part takes a look at past and future business opportunities for the company. I have not linked to terms previously linked to in the first part of the series, which should also be referred to for typographical conventions, if they're not clear.
- Part 1 - Autodesk's early history
- Part 2 - AutoCAD's architecture & APIs
- Part 3 - Autodesk's eco-system and strategy
- Part 4 - Past and future opportunities
Past and future opportunities
What technology trends do you monitor in the industry?
Well, I don't even really try to do that, since I'm almost always consistently wrong. I just kind of wait for the technology to roll over me like a wave, and then try to get up and surf on it. I think the next big trend is the one that you haven't seen yet. Because if you've seen it then it's the current big trend.
I've always taken this view that was so discredited, the George Gilder view, that was some completely discredited during the dot-bomb: that there really isn't any technological limit to how much bandwidth we can deliver. And how much we can connect. There isn't any technological limit - at all - to how much computing power you can put on the web, as long as you're willing to have it broken up into little pieces that talk to one another. That's not visionary - that's here. Because Amazon and Google both have platforms that do that. And on the 10th of September, the LHC is going to start taking data and sending it out to the grid. And doing their computing - of their unbelievable data-stream, terabytes coming out of that thing - storing and distributing it all over the world and doing their processing wherever there's computing power available to do it.
So that's here, now. What are the applications for that? Well, the Large Hadron Collider is an application, but that's kind of a one-off - you're not going to have the PC collider, or something like that. But what are the applications that you can't do today, but that you can do with, say, 300,000 machines of super-computer capability? Well, I can name one: spam! That's how it's being done today, it's being done by the botnets, running on corrupted Windows machines. So there's one. But what's one that - well, that's one you could make money off of, actually - but what's another one that you can make money from that's legal and ethical, and so forth? Particularly for the Autodesk niche, what can you do - I would say probably not so much on the CAD side, anymore, because you've got more compute power than you need for CAD, right now - but what can you do in the video- and movie- production environment, with a million times more computing power? Well, there are probably a lot of things you can do, suddenly. If you could imagine rendering Jurassic Park 3 in real-time, that's interesting.
Have you monitored the products Autodesk has developed or acquired over time?
I've never even seen them. I know them only from what's written up about them in the annual reports, which is basically the only communication I have with Autodesk, anymore. I think they've - certainly from a business and financial stand-point - done superbly. It's something that Autodesk absolutely needed to do. At the point that I basically got disgusted and walked away I was saying "you guys, you've got too much money, you need to do acquisitions. Because you need to do acquisitions in things that have a growth in front of them. Because when you've got 95% of the market, where's your growth? You're not going to grow any faster than the general economy." And they were originally very tepid in doing these little acquisitions. I think Discreet was the first big acquisition, and that worked beautifully. I mean I didn't see the messiness from the inside, I just looked at the financials, but it worked beautifully from the financial standpoint.
I think they've done very well with that. That's an industry that was - very much like AutoCAD - poised to profit from every increase in compute and graphics capability. And that was something that was - again - at the right time.
But again, it's a lot harder than it used to be. When we were sitting there in 1982, how difficult was it to think of products to put on the PC? Well, about as hard as looking at the products that we'd already worked on on mainframes. And making a forecast as to when that capability would be present on the PC. That's all you had to do. There were dozens of product categories that ran on mainframes that you could basically port to a PC. And plotting out Moore's Law you could be confident that by the time you were done, you're going to have the compute power that you needed to run that. Well, if you were wrong, that was a problem, but we never were wrong! In fact, most estimates were actually less than we got in the '80s and '90s.
And today, you can't do that anymore. Because the cutting edge is on the PC platform.
You're plotting out what can be taken from the PC and put in the cloud.
Maybe. Or what can you do with the cloud that you cannot do on the PC. But you've got to think of that. You don't have a place that you can look and say "well, people are doing this with supercomputers". No, because the cloud is far more powerful than a supercomputer, and it's a different kind of beast. And so that's something that's much more of a challenge. I have an almost unlimited optimism in technological progress but I have an even greater optimism in the ability of creative humans to think up things to use all of that technological progress to do. Those applications are going to come, but I have no idea what they're going to be.
Who would have imagined Twitter? I mean, come on!
I'm curious - do you use Twitter?
Do you have an interest in those technologies?
I'm interested in anti-social networking. I'm interested in protecting private data and one's own history in this environment of unprecedented disclosure.
I remember you developed an early, encrypted VoIP tool.
I spent half of the 1990s doing that. That was Speak Freely. And that was launched in 1996 and I finally drove a stake into it on January 15th, 2004. It's the first one that really worked, basically. And it not only worked, it was cross-platform from Linux and Windows. It was a bit of a reach. It would run on a 50MHz 486 with a 14.4K modem Internet connection. That wasn't easy. And it ran on Windows 3.1, too. By, say '98-'99, at the point the Internet - the consumer Internet - was exploding, at that point the infrastructure was such that it just worked. There just wasn't any problem with it and people weren't even using the most restrictive compression modes or any of the redundancy and recovery and so forth. People just routinely used it. There was a software company in Berkeley, California that had a bunch of programmers working in Moscow, and they would have their weekly teleconferences on Speak Freely, it was just a routine thing and it just worked.
It got killed by NAT. When the great broadband pile-on happened, suddenly people who had unrestricted inbound-outbound connections were behind this NAT-box, and they could just not initiate connections to the outside. So that meant that the only way that you could make it work was with a central server, and nobody had the bandwidth to - I mean there were a million downloads of Speak Freely - and nobody had the bandwidth to set up a server to handle that kind of traffic.
What do you consider the most promising Autodesk products - that were under development or made it to market - that you believe should have been successful, but were not? I'm thinking along the lines of products such as HyperChem or projects such as Xanadu, which I remember reading about during my Computer Science studies.
Well, I'll give you the number one, and that's probably one you didn't expect to hear, and that's AutoSketch. Which we developed in '85-'86. Autodesk spent a year and a half on that product - I don't know how much money, particularly when you count QA people working on the product, and so forth, and spent absolutely nothing to promote that product in the United States. I mean zero. After it was shipped it was maybe six years until the next release came out, there was no-one doing any development work on it. And how was the product received in the market? Well, in Europe it got about a 90% market-share in about 18 months, because the people in Europe did promote it. So I don't really believe that there's a cultural difference that meant it wouldn't have had the same results in the US. But the Autodesk dealers didn't want to sell it, because they thought it would cannibalize sales of AutoCAD. The marketing organization didn't want to put it through the mass-market channel, because they were afraid the dealers would show up with pitch-forks and torches, if they thought we were shifting to the mass distributors. And it was only $79.95, I mean we only got twenty bucks, or something, from it.
So they ignored it. And they consequently had to spend $10M buying Generic CADD, which should never even have existed. Which they then ran into the ground and obliterated, through - may I use the word - incompetence. Incompetence and lack of interest. I mean, Pete O'Dell went up to Generic CADD, and he was a fire-brand, I mean he was doing everything in the world: the development, porting things, application developers, developing a new distributor channel, he was getting it into like Home Depot, with kiosks, and so forth. Coming up with products for individuals who wanted to redesign their kitchen or bathroom. And the Autodesk organization just strangled the resources, didn't give him anything he needed, even though he was generating profit and contributing profit to the bottom-line. And just ran it into the ground. And so twice - we had two products, well a product and a whole product family - in the low-end sector and it was just a classic "California blinders", so we had to do it a third time with LT! Having blown the opportunity, and particularly blown the opportunity to own 90% of the market from the outset, the way we did with AutoCAD. There should have been at least four releases of AutoSketch in those six years.
And the thing about AutoSketch - and I was the principal developer of AutoSketch, so I have both some knowledge and emotion in this - was that it was deliberately designed not to cannibalize AutoCAD. It was architected not to cannibalize AutoCAD. You could add all kinds of things to AutoSketch, but because its interface - AutoCAD was basically a command-line interface, this was a pull-down menu/dialog box, albeit MVC text-based - and AutoCAD was inherently programmable, AutoSketch was inherently interactive. So it just wasn't going to happen, you weren't going to see people abandoning AutoCAD for AutoSketch, whatever you put in AutoSketch, because it's just a different beast.
And that was just thrown away. And there wasn't any button that I, or anybody else who cared about it, could push that would change that marketing and sales organization, which was generating so much money just sitting there and writing orders from dealers selling AutoCAD, that it didn't want to do anything else that wasn't as profitable as that. And it didn't want to do anything at all that could possibly imperil that. And that was just lost. And that, I think, was probably - at least during the epoch I was there, which is all I can talk about - the biggest strategic failure of Autodesk in its core market.
Now the things that we did in the '80s, those were all opportunities that really had nothing to do with AutoCAD, in a large part. Those were attempts to get in the next rising big market, so that we could ride that one up and not simply grow as fast as the general economy.
Well, I think Xanadu... it's really very simple: if you had visited the Autodesk lab in the summer of 1991, and asked for a demo of Tapestry and Xanadu, but you had in your head what you have now, you'd say, "oh, that's a web browser!" So we basically had the web - Xanadu - and the browser - Tapestry - running as a prototype in 1991-2. And killed it in 1993. Nobody saw the Internet exploding the way that it did. And the fact is, of course, that the Internet - mass-market Internet - is a prerequisite to having that technology work.
Now I've always taken the approach that I think Bill Gates takes, which is: if I see something that I just know is eventually going to be an unboundedly large market, I'm willing to lose a modest amount of money for however long it takes for that thing to arrive. Because when it arrives, I want to be the one there with the product. And I think you could see that in the '70s and '80s with the CD-ROM. That he organized the first CD-ROM conference, he made all of his products available on CD-ROM, he put CD-ROM drivers in Windows. Nobody had CD-ROM drives! It took years before that really happened. But when it happened, he was ready. Now that wasn't a huge money-maker for him, but it was in the sense that it enabled him to deliver products that were vastly larger than 320K floppies, or something like that. So that was strategically important to him. I counseled Carol, "you can afford this. This is going to happen. And you want to be in a position when it happens to be the person that everybody comes to for the technology." And that was not the decision that she made.
Now there were a lot of problems with those projects, and there were a lot of problems with the people on those projects. And, whether, in fact, if you had gone on funding them at the level we funded them, they would have ever have gotten anything done, that would even have been compatible with the first release of Mosaic, is not clear. Because remember, really, Berners-Lee was already working on the web in '92 and '93, so it was going to happen pretty soon, anyway, and at the point that the web appeared, the growth was so instantaneous that Xanadu would have just been forgotten instantly.
Although, of course, you could say that it was a lot better than the web, you don't have any broken links in Xanadu, for one thing. But whether it would have really... because it was, in some ways, more centralized. Whether it would have been able to fit the model by which the web grew is not clear.
But then there was also AMIX, the AMerican Information eXchange, which was killed the same day as Xanadu. If you'd seen a demo of that, you'd have said, "oh, eBay!".
So there's a couple.
HyperChem was more a problem with the very crusty developers that were responsible for the package. They just didn't fit with Autodesk, and the way that we'd licensed the product was such that they could essentially pull the plug on us. And they essentially did: we were not going to get any new releases and development was not going to go in the direction that we thought it was going to. And we told them, "if you go this way you're going to have a very specialized, very high-price product. It's never going to have the mass impact." But they were not willing to lower themselves to the commercial hucksterism that we used to promote AutoCAD to get there.
And that's fine: it's their product, it's their work, it's their company. I think they missed an opportunity by doing that. Would it have ever been a great, mass money-making product? No. I mean, it would have been a very prestigious product, and it would have established Autodesk... it would have been a substantial amount of revenue, I think, if it had been done right. It would have established Autodesk in a completely different sector, which was when all the analysts were saying, "you're a one-product company, you're a one-industry company!" OK, well architects, and mechanical engineers and... chemists. Now is that different enough for you? Thanks, right.
And it was also a superbly Windows-integrated program, and they were just on top of getting something like that to work in Windows 3.1. So I think we would have been able to suck a great deal of Windows expertise that would have helped us with AutoCAD. And that wasn't a reason to do it, but it was another benefit that you might get.
Hewlett-Packard doesn't make caesium atomic clocks primarily because they make so much money selling like eight of them a century to the Bureau of Standards. They make them because knowing the company that makes the most accurate clocks in the world also makes your PC is a benefit on the marketing side. I think perhaps HyperChem would have helped in that way, too.
Other than for the prestige, do you see it at odds for a volume software company such as Autodesk to invest in - whether buying or developing - low-volume, niche technology?
I think the first thing to do is look for the niche that becomes the next volume market. That's where - again - you play the game with Moore's Law. The thing you want is the niche market that's constrained by the fact that the hardware that it takes is too expensive. Because wait a few years and the hardware it takes won't be expensive. Now that doesn't say that every niche is a potential mass market. But remember, that there was a time, in the 1960s, when there were maybe eight computers in the world that could play video computer games. And that's a substantially bigger market, now.
So they do exist. And I think a lot of people would have said: professional-quality video-editing - back in the days of Avid - or the Avid hardware... workstation - that's never going to be a mass market, because even if it were free, the skills that it would take to do that are so specialized that it'll never be a mass market. Well, wrong!
There is a cultural and commercial problem, though. Because when you acquire a company you also acquire their sales force. That's used to writing $100,000 orders, and suddenly you're talking about $1,000 orders and they're not interested, thank you.
Well that's a very famous quote from Jim Meadlock, when he was speaking... it's the only time I actually met him in person, we were both at - I think it was - the Silverado conference in 1986 and I believe I had given my presentation about how were going to simply take over the whole CAD business, because within a year, every PC was going to have what you guys now call workstation capability. Who's going to need the workstations when we have that? The workstation people can go higher, than what they've got now, but we can do that on the PC. And Meadlock got up and did his presentation, and I think that's when he started flogging the object-oriented stuff, but one of the Wall Street CAD analysts got up and said, "Jim, what do you think about this AutoCAD phenomenon? The guy before said he's going to take your whole market away." And Meadlock said, "when I figure out how to make money selling CAD for $1,500, I'll do it." And I wanted to get up and say, "well, ask me! We're doing it!" :-)
I loved that speech from the Silverado conference, where you got up and started the speech by pointing at a photo of a 1960s computer and said "remember when computers looked like computers?". And then you finished the speech with the comment that in twenty years' time, someone would point at a picture of an IBM AT and say "remember when computers looked like computers?" That, for me, painted a very clear image of technological progress. Did you end up getting the chance to say that again in 2006?
In a way, though, has there been that much progress? I mean for a lot of that stuff, people really do, it looks a lot better but it doesn't even seem much faster to me. An AutoCAD, sure, but word-processing, the stuff that you did back then in '82...
Then that's not the right device - maybe you should point to, say, "remember when mobile phones looked like mobile phones?"
Right! Like bricks with huge battery packs.
Maybe the real innovation has really shifted somewhere else.
Nobody saw the Internet 18 months before it happened. It's like the classic example. You can read all the science fiction from the 1920s through the 1960s! The 1960s! And they had asteroid mining, and they had faster than light travel, and they had terraforming Mars, and nobody saw the PC. And in the same way you can read through almost all the science fiction through the 1980s, and nobody saw the Internet. A couple did, John Brunner, and obviously Ted Nelson, if you call his stuff science fiction and I think we can now can call it science fiction.
But I don't think anyone imagined that it would happen, and yet the funny thing was, the people that built the Internet in the 1970s, they all knew it was going to happen. They were believers. They didn't necessarily know when but they knew what. And thank goodness, because if that hadn't made some of the design decisions they made, back in those days, we would have just had horrific problems getting to where we are now.
[Portions of this interview have been elided both at the discretion of the interviewee and of the interviewer.]
[So ends my transcription of the interview. I'd once again like to extend my sincere thanks to John for sharing so many valuable anecdotes and insights. While he may attribute much of Autodesk's early success to luck, I think it's clear that John's leadership was just as significant a factor.]