Betting on the Net:

Eric Schmidt interview

By Educom Staff

Sequence: Volume 31, Number 5
Release Date: September/October 1996

In his role as chief technology officer for Sun Microsystems, Eric Schmidt oversees all aspects of Sun's core and emerging technologies including the new Java programming language. In this insightful interview with Educom Review, he discusses the future of the "network computer" and the "Internet appliance."

Educom Review, Volume 31, Number 5, September/October 1996

Educom Review: Since the time Sun started using the slogan, "The network is the computer," in 1987, a lot of other companies have come to the same conclusion. So in, 1996, how does your strategy really differ from the strategies of your competitors?

Eric Schmidt: Well, the good news is that they are coming to our party. It's not like we have to change much. We have the benefit of having been preaching about the Network-centric model for a very long time. And in a competitive environment there's a lot of benefit from having done something for awhile. So typically what happens is that if you want something that actually works, you buy from Sun. Yeah, everybody else has some competitive offerings, but we end up surviving, partly because we've just been in the game longer. That's sort of a pedestrian answer . . . but it's a truthful answer. Sun is in the fortunate position of being one of the companies that is perceived as a leader in a very dynamic environment, where there's always a lot of new stuff being invented. Java, for example, is a technology that just emerged, popped out of nowhere. So as long as there are a few of those great ideas that survive, we'll continue to do fine.

ER: Are you saying that your main advantage is being first?

Schmidt: Well, yes, there is certainly an advantage to being first. But Java has had a number of things going well for it as well. It has had an important set of partners, starting with Netscape; it's also a language derivative of C++, so it's very easy for people to use and it works with all existing operating systems. So those reasons, which are not particularly technical, but more business-oriented, are key elements to Java's success.

ER: If the network is the computer, the question arises: will the network be robust enough?

Schmidt: Well, it's not the telephone system. You know, a telephone system is designed with a certain basic level of reliability that is guaranteed end-to-end and its quality of service is based on a long set of rules - such as "What does it mean if you drop this?" and "How long do you have to keep connections open?" It works remarkably well. It's an amazing computer system in itself. The networks that are being built today do not have that level of reliability and will not have it for quite some time. On the other hand, they offer best-effort service at a much lower price point, because you can just sort of sling these networks together and don't have to worry much about how they are really designed - and yet they end up working pretty well. And that's why the adoption rate is occurring so much more quickly than it is in telephone systems, which have to be centrally planned and thought through. So there is always a tradeoff between a perfectly planned, perfectly uniform system and something that can be put together very quickly. And given that we have these networks which are hodgepodges, we are now developing protocols that will provide end-to-end guarantees of service on top of an unreliable structure.

ER: Does that imply that the big players will become bigger players - the AT&Ts and so forth?

Schmidt: Well, I actually think that in the infrastructure game, size matters - and is positive. So I think a well-managed global network communications company is a good idea.

ER: Do you mean one? Or several?

Schmidt: Several. But I think eventually there are a lot reasons the economies of scale favor global operations. One of the problems of the Internet today is that the quality of service is terrible, so now you have vendors who come in and say, "Hey, I'll offer you good connectivity, and what I will do as part of that connectivity is I will give you better service if you stay on my network." Think of it as being like a hub-and-spoke system on the airlines, where as long as you stay on United you get (at least in theory) better service because you don't have to trade terminals and the schedules are aligned. The airline industry evolved the hub-and-spoke model partly as a defensive maneuver to keep customers by keeping them satisfied. The same thing will occur in networks. Once you stay within the UUNet network or the Netcom network or the BBN network, you will presumably get better service.

ER: Then are the small Internet companies going to be blown away?

Schmidt: I don't think they will be blown away. I think they will be aggregated. They are providing an important function today. My guess is that somebody like Craig McCaw - some clever person like Craig will come along, buy them all up and eventually they will all be bought by AT&T or MCI or somebody like that.

ER: So how many companies do you think will be left in the business of providing Internet service?

Schmidt: Well, one thing people don't know about the telecommunications industry is that in fact there are 470 long distance companies in the United States, yet we only think of three. So it is probably the case that the data networking providers will come to a similar structure. Let me pick a number - five - as the number of large, global ones. And then there will be another 400 boutique ones, which will provide a premium product or a premium service. They offer something special that's unique.

ER: What do you see ahead in the evolution of the Internet, and what different roles will be played by PCs, TVs and NCs (or "network computers")?

Schmidt: Well, many people get confused about whether they are talking about the device or the network itself. To me the two are completely separable questions. I believe there will be a very large, highly interoperable broadband data network, and it will connect a lot of traffic, from a lot of different devices. Let me tell you some of the devices that are going to plug into it. Obviously PCs; and corporate nets which are PCs, Macintoshes and workstations and stuff; and cellular phones; and pagers, which are just a cheap form of cellular phone. (A pager is a bad cellular phone but it has the same properties of mobility and ubiquity.)

ER: What about TVs?

Schmidt: I think that TV will be a very poor-quality browsing device. My guess is that the TV-only solutions won't succeed because resolution is so bad on a TV. TV wasn't really designed for digital data.

ER: And then there is the network computer?

Schmidt: Yes. Absolutely. I think eventually there will be a market for this network computer idea. I really do.

ER: Well, how do you answer those people who say you're wrong - that the network will simply never be that robust, or that there will never be enough really powerful programs available over the Net?

Schmidt: It has historically been a mistake to bet against the Internet. Any rational person who followed the trend lines would bet in favor of the Internet. The Internet as it's grown has faced a great many challenges - legal issues, connectivity issues, bandwidth issues, scaling issues - and yet because there are so many players and because there is so much at stake these problems get fixed, and right now there are many people working on protocols that will give you a guarantee of service over an unreliable fabric. So I think betting against the ingenuity of people on the Internet is a mistake. And my feeling is that the biggest barrier to the success of the Internet right now is not technical, but economic, and is that the people who use the Internet only pay for their first connection. It's the equivalent of being able to make a phone call from Oakland, California and not being charged any difference if that phone call goes from Oakland all the way to Sweden or just to San Francisco on the other side of the San Francisco Bay.

ER: And you think there should be a difference?

Schmidt: The fact of the matter is that the world does have distance-sensitive pricing. Regardless of whether we like it or not, it costs more to get on that cable that goes underneath the Atlantic than it does to go across the Bay area. So here you have a situation where you have people who are building an entire infrastructure on top of a model that does not reflect the true cost of transmission. So what you have is rationing of a different kind. You know, if you don't have price rationing, you have time rationing - you have queues, and that is why the service is so terrible. You can't pay for better service, so you pay in some other way. It's the same principle as in the Communist system, where you waited in long lines because you couldn't pay to get out of the queue.

ER: Changing the subject now from cost of transmission to cost of information, what do you think of the ability - or inability - of companies to make money selling newspapers or magazines on the Internet?

Schmidt: Well, first of all, there is this complete deluge of publications coming onto the Internet, but in almost all cases what they are doing is essentially treating the Internet as an equivalent mechanism to their current method of distribution. So if they used paper, now they just papering over the screen.

ER: Well, but they pretend there is more to it than that.

Schmidt: Yes, but realistically, it is pretty much the same thing. And my guess is that it is not going to be that way in the long term. In my view the way to win is to recognize that the Internet and network computing create a new medium that is much more hyperlinked and much more interactive. And one of the facts about the traditional print media and broadcast media is they are non-interactive. To me the medium is not like that. It is not the boob tube, watched by people lying on their sofas. It is a very different medium.

ER: So you don't think it is enough for the established newspapers and magazines now coming online to have chat sessions with the editors and journalists and with fellow subscribers of the publication?

Schmidt: I think that's a very small component of what they could do. My belief is that every news story should come with a program which in some sense diagrams or explicates what happened. For example, if there is a story about a crime, why is there not an interactive equivalent of a hologram picture of where the crime scene was? In the case of a presidential proclamation, why is there not a dynamic analysis of whether this is true or not - with me being able to choose the truth vectors? There is a lot more data out there than is being properly massaged. But it's a new medium, and it takes a long time to develop these things. So we need to give our newspaper friends a break: they just figured this out in the last year, and the economics have just begun to work. The Wall Street Journal and the New York Times have just begun to offer interactive editions, which so far are basically text with a few hyperlinks and, as you mentioned, a few chat sessions with editors. It will be a couple of years from now before the media begin to exploit this stuff.

ER: Well, while we're on the subject of things-to-come, what do you see in the future?

Schmidt: It's always extraordinarily dangerous to try to predict, but I think that there are a couple of things going on that are really fundamental and that are likely to be true for 20 years. One is that the semiconductor price performance continues to improve on what is known as a Moore's Law curve - and although there's some evidence that it may eventually hit a limit within 10 years, let's be generous and say it will continue for 20 years. If that happens, then we are talking about supercomputers on your desktop, because we get the compounding of doubling every 18 months over a 10- or 15-year period. The rate at which everything is going to become digital and the ubiquity of digital transmission is one of those underlying changes. Here's an example. My guess is that you will carry around five IP-addressable, Internet addressable devices, on your body in the year 2000 - maybe more - and every device that has some kind of a computing or battery function will be on the Internet.

ER: What do you think those five would be?

Schmidt: Well, I could actually give you more than five, but let's try just a few. First, my pager. Second, my cellular phone. Third, my watch, which would include a global positioning system so that I could tell people where I was. Fourth, my wallet, and in this regard it's clear that there will be digital versions of cash - which will be smart cards, roughly the equivalent of a credit card with a computer inside of it. And fifth, obviously my portable, which you may or may not consider plugged in to me.

ER: What about tennis shoes? Some people are talking about computerized tennis shoes.

Schmidt: Yes, there are a lot of body kind of athletic things you can imagine, but in terms of devices that would get advantage out of being connected to a vast global network, I think that knowing where I was, being able to be paged or talked to, being able to compute, the ability to spend money - all of these are Internet uses which will be inevitable.

ER: What kinds of people will make all these things happen? You know, it sometimes seems that there are two kinds of people who are leading the industry and they are different types. We won't name names, but some are very colorful characters - and some are not.

Schmidt: Well, I think every one of them adds value in some unique way. What I've noticed in our industry is that an awful lot of the most successful people are self-made founders. I'm thinking of Bill Gates, Larry Ellison, Scott McNealy. It's really quite a long list. And I think that part of the reason the founders tend to win is that unlike normal traditional corporate bureaucracies, high-tech requires a genuine bet, from people who just don't care, they just want to be right. And it is very rare to find somebody who has that kind of vision and can be right. And it is often a founder.

ER: Is it often a young person? Is there a youth culture involved?

Schmidt: Well, there certainly is a youth culture. But that may also be just a one-time effect. Remember, computer science wasn't even taught until the generation now in its mid-40s went to college.

ER: What's your own educational background?

Schmidt: I went to Princeton, undergraduate in EE; then I went to Berkeley and got a Ph.D. in computer science. But I have observed that the most successful people at the top of these companies are typically not computer scientists. They are typically just really clever businessmen. My characterization of it is they are people who are extremely smart, work all the time, have tremendous passion, and have insight. I'm thinking of people like Steve Jobs. The successful ones spend most of the time trying to keep their cultures entrepreneurial and progressive, because the biggest problem with these big and successful companies is that they are in fact big and successful. Of course, the reason they got to be big and successful is that they were little upstarts that were scrappy and aggressive, and it's very difficult to maintain that.

ER: What do you do at Sun to try to remain scrappy and aggressive?

Schmidt: Well, we do a bunch of things. We have a research lab at Sun, and the idea is that the research lab is a place where people can do crazy things. Java originally came out of the research labs. Scott McNealy, as our CEO, has pushed very hard for new initiatives in new areas of business. We pay our executives based on business success and a certain percentage of their income is based on profits of their business. So we pay for performance. People are incentivized to perform better. The biggest danger to any of these things is a sort of stifling corporate bureaucracy which is characteristic of these large companies 10 or 20 years ago.

ER: And how does Sun keep in touch with the noncorporate world, such as the world of colleges and universities?

Schmidt: Well, let me give you some examples. Sun's technologies originally grew out of work done primarily at Berkeley and Stanford. We are fortunate in that the majority of the technologically sensitive graduate programs in universities use Suns for computer science, engineering, things like that. Colleges and universities have a fairly well-understood set of problems. Many of them share software that we have to make sure runs on our platforms. As an example, the astrophysicists all use the same particular set of codes which they do their basic research on and then they build stuff on top. We make sure that stuff runs on Suns. So we have more than five hundred million and perhaps a billion dollars in educationally oriented programs at Sun.

ER: How do you see education being affected by the Internet?

Schmidt: The new data are that the universities are all now on the Internet, so you now have instant virtual collaboration around the world. My guess is that the quickest way to improve things in the university is simply to use the data that's out there that you haven't seen on your campus yet.

ER: What kind of data?

Schmidt: Well, let's start with source codes being developed in programs that are shared but not widely distributed. There's a tremendous amount of government funding in universities - with good work being done that people just don't know about. I'm thinking now about anything that involves any research curriculum - physics, chemistry, you name it.

ER: Let's end where we started, with the topic of network computing. A lot of people now seem to be saying, "Well, the network computer is all very well, but it's just an ancillary device; it will never replace the computer. They will both coexist quite happily." Is that the way you feel about it?

Schmidt: Well, I certainly do not believe the network computer will displace the personal computer. I think the personal computer is going to do just fine because networking makes it more valuable. To me, networking makes everything more valuable - right? You can see more things! If you're sitting there in front of some machine that is not connected to a computer, you are lonely, because there's a lot going on that you are not participating in - you are the person who didn't get to go to the dance. So my feeling is that the network computer creates a new, additional category of users who either are put off by the complexity of PCs or who have a relatively dedicated use.

ER: Is there going to be a real reason for a PC 10 or 20 years from now?

Schmidt: Oh, I think so. There will always be a split in computing between general purpose and special purpose. The general purpose computer - which is the platform on which you can build the next great invention - will always be there, and the PC at the client end in a corporation is playing that role by providing the place where people can do things like play around and do interesting things with software they have found at some store. But the dedicated device is also going to be very successful because corporations and some universities will see that it is much cheaper to have something that has a dedicated use. This is the kiosk approach, especially useful for those people who find the complexity of general purpose computing too much. There will always be that struggle - it will never go away. There're about three million dumb terminals in use today around the world - basically VT100 and 3270 terminals. People think that PCs have taken over the world, yet there are almost that many dumb terminals connected to mainframes still sitting around running some dedicated application. There's no question there will continue to be a big market for that.

ER: Compare network computers to dumb terminals.

Schmidt: There is a similarity. Let me tell you the difference. The network computer has a huge CPU, lots of memory and a very fast network; whereas a dumb terminal has none of those. The dumb terminal doesn't run the application, so the issue with the dumb terminal was that the mainframe was always bottlenecked. And if there is an issue with the network computer it will not be that the mainframe is bottlenecked; it will be that the network is overloaded. So the criticism that I think is valid about the network computer is that it presumes that networks will continue to get faster and faster and become more generally available. I happen to believe that's true - and I think the trends will suggest that I am right. So the network computer is a bet on the network, which is good. You need to make bets and I think that is an appropriate bet. Somebody like Larry Ellison is making a big bet - is betting his company on that. I think that's great.

Take me to the index