from www.wired.com

Text of Wired's Interview with Google CEO Eric Schmidt

On March 23 I spent an hour interviewing Google CEO Eric Schmidt in a cramped conference room 50 feet from his even more cramped office. (It's so small that if you spread your arms you can almost touch both walls.) We talked about everything from Google's competition with Microsoft and its partnership with Apple to all those data centers it is building.
- Fred Vogelstein

WIRED: When you joined Google it was just a search engine. Now it's redefining the way the world thinks about computing. Explain.

ERIC SCHMIDT: It's pretty clear that there's an architectural shift going on. These occur every 10 or 20 years. The previous architecture was a proprietary network with PC clients called client-server computing. With this new architecture you're always online; every device can see every application; and the applications are stored in the cloud. It means that your servers are professionally managed, so you can actually have a weekend and not spend all your time trying to manage your servers. It's like having banks manage your money rather than you managing your money. And the networks have become secure, and the computers have become fast enough that this is mechanically possible - it actually works.

The other thing that's interesting is that the new architecture brings in other voices. The earlier model was pretty proprietary. The protocols, which were typically Microsoft-based, didn't allow for other (interface) choices very well. Now, with the Internet protocols you can pretty much plug in your own interpretation of how email should work and your own interpretation of how voice over IP should work.

This point about anyone being able to enter the market is a big deal. Photo sharing, social networks, all of them have this property. And what's interesting is that Google, although we're one of the companies, we're, by far, not the only company that's doing this. Yahoo is an example of a company like this. eBay is a company like this. Amazon is a company like this. And each of the companies I've named makes money in a different way.

Right.

We have talked about this network, or the cloud computing model for years, but we were beholden to the old software selling model - the one where the salesperson is making a million-dollar sale. I used to be in this business [when I was at Novell and Sun]. That model doesn't scale for Internet users. You just can't get that kind of money out of the average citizen. So the new model allows you to have free services with advertising. And this targeted advertising thing is a really, really big deal.

Isn't it more likely that we'll have a hybrid model - with some applications in the cloud and others on the desktop?

It depends. There is not a middle ground when it comes to protocols. In order for this vision (of cloud computing) to work, the protocols have to be open. They can't be proprietary. Everyone has to have access to them. So that's a clear, binary answer.

With respect to the user experience, which I think is your real question, a hybrid works depending on how it's architected. It makes sense, for example, to have graphics computing close to the end user because that's where your frame buffer is and your computation is. Games are a good example because it's very, very hard to imagine games that are network resident only. They're so highly interactive.

Right.

But it's perfectly possible to have most of the other computing being done on the server, so that's an example of a hybrid model. If it's something (like a video or a document or a spreadsheet) where there's relatively few changes (to the file), you can put it on a service (in the cloud) and then you can cache it locally.

All these features don't exist yet, though.

True. Google docs and spreadsheets don't work if you're on an airplane. But it's a technical problem that is going to get solved. Eventually you will be able to work on a plane as if you are connected and, then when you get reconnected to the Internet, your computer will just synchronize with the cloud.

Here's another way of saying this - and these are not my words. People call this an Internet operating system. And by "this" I don't mean Google, I mean the sum of this vision. And if you think about it as an Internet operating system, the Internet operating system will have to have all of the normal features of the older versions of operating systems. It will have to have security, it will have to have caching, it will have to have replication, and it will have to have performance.

Why is it taking so long?

Well, one answer is that the systems they're replacing are very complicated, and people have very high standards for interactive services. So everything has to work; all the features have to be there; and they have to never break. We used to think that the enterprise was the hardest customer to satisfy, but we were wrong. It turns out, consumers are harder than the enterprise because the consumer will not give you a second chance.

And by the way, I would argue that we in the industry forgot this. We became as a group - certainly I did - consumed with the complexity of the systems that we were building for powerful corporations, and we forgot that there's a much larger market around consumers for simple solutions.

Online calendars are the perfect example of this. Sharing a calendar in the older (client server) model was hard. Now it's easy because the model says the calendars are stored on professional servers, and they are visible everywhere you want them to be. Making this happen reliably and securely is complicated and technical, but it is ultimately justified by delivering on a very simple concept.

When you joined Google it was just a search engine. It has grown into much more. How should we think about Google today?

One is as an advertising system. Another one is as this end-user system (the search, email, and other applications Google delivers to users through an Internet browser). A third way to think of Google is as a giant supercomputer. And then a fourth way is to think of Google as a social phenomenon involving the company, the people, the brand, the mission, the values - all that kind of stuff.

How powerful is the supercomputer?

There's never been anything like it, so we don't know how to express it. We build our own data centers, and we do a lot of the work ourselves because the commercial solutions do not have high enough performance.

What do you mean you do a lot of the work yourself?

Well, essentially, we do all of the software. So the computers that we're running start off with Linux as the base, but after that it's really custom software to move all the data around. The Web services, all the identity management, all of the database activities, all of the indexing, all the searching, all the ranking, all that kind of stuff in the cloud we do ourselves. This is a great core competency of the company.

And we have not only data centers, but we have fiber that interconnect those data centers, and connect to the ISPs. At Google, speed is critical. And part of the way we get that speed is with that fiber.

How many data centers are there?

I don't actually know.

Are we talking about a half dozen? A dozen? Or are we talking about dozens?

I think my overall description would be in the dozens. There are a few very large ones, some of which have been leaked to the press. But in a year or two the very large ones will be the small ones because the growth rate is such that we keep building even larger ones, and that's where a lot of the capital spending in the company is going.

Why do you have to control your own fiber to connect the data centers?

One of the neat things about the bubble is that people built all of this fiber that is now essentially free. What's funny about our fiber leasing and purchasing is that people are always assuming that we have some master plan involving telecommunications when, in fact, if you think about it as just solving the supercomputer problem, we just want the thing to be faster.

Then why are you being so aggressive trying to get muni Wi-Fi projects going around the country?

Because we benefit from broadband. Remember, one of the critical things in our model is that having inexpensive or, ideally, free access to broadband is a good thing. Especially if it's somebody else who's going to subsidize that using their economics, we think it's great. And the more broadband we can get globally, the better. It's better for the world; it's better for our advertisers; it's better for Google.

So you want to increase broadband penetration in San Francisco, for example, rather than replace the broadband penetration (from cable and telephone companies) that already exists?

Yes. That's a better way of saying it, thank you. So if you have 10 percent market share and we can get it to 50 percent, we know that produces a happier citizen. And we know that those searchers will use our services more. They're much more likely to become a calendar user or a G-mail user or a news user, or whatever, because they like the performance. And, by the way, we're indifferent as to which broadband it is. I mean, it could be Wi-Fi, it could be fiber, it could be coax, it can be any of them.

But you're not doing last-mile fiber?

No, no, no. There are plenty of great companies doing that, and we're perfectly happy to go right on top of all that fiber they're putting in place.

You just recently joined the board of Apple and have talked about potential partnerships between Google and Apple. Explain.

Google's architectural model around broadband and services and so forth plays very well to the powerful devices and services Apple is doing. We're a perfect back end to the problems that they're trying to solve. And they have very good judgment on user interface and people. They don't have this supercomputer I'm talking about, which is the data centers. What they have is a manufacturing business that's doing quite well. And the obvious example is the iPhone, which they announced has in it Google Maps.

Let's talk about the fun stuff. Why do you think Viacom sued you guys?

It's a business negotiation. And it's well established that we've been negotiating with them, and I'm sure at some point we'll negotiate with them some more.

Their argument is that you're not working hard enough to keep infringing stuff off.

Well, if they would look at the law, they would understand that under the Digital Millennium Copyright Act, there's a shared responsibility. The law basically says that the copyright owner monitors, and then we expeditiously remove, and we've done that. And it's well documented because Viacom told everybody that they gave us 100,000 video take-downs, which we did very, very quickly. And what was interesting was that our traffic to YouTube has grown very strongly since then. So one of the arguments that they made was that somehow YouTube was built on stolen content, which is clearly false.

What's so hard about finding all the infringing stuff users put up on YouTube? Surely, you at Google have the ability to write better content filters.

We are working on the appropriate filters, but it's a hard problem. You could do audio sampling. You could do video sampling. But the audio and video sampling technology is not nearly as good as the technology that allows you to spot infringing text. But it's important that we do it because we don't want to be in the position of having to be given these constant takedown notices.

News Corp. and NBC Universal just announced they are joining forces to create their own video-distribution channels online. Does that represent serious competition?

No. They are taking content that they created and using an exclusive licensing agent to license that content for anybody. That's, clearly, a good thing because it means that that content will be available to everybody.

I got the impression that they were gearing up to compete with you?

No. In fact, Peter (Chernin, News Corp. COO) and I have had a long conversation about this. We spoke yesterday (March 22) before the announcement, and Peter explicitly said that this was not a competitor to YouTube. In fact, their site doesn't even exist yet. They're still designing it. So, to me, it's responsive to finding a way to get the content out.

More broadly, how could copyright law in the digital age be clearer?

The balance that has been struck in the DMCA has worked pretty well, overall. And I think that it may be better for all of us to work within it for a while as we develop these new businesses. It's the unintended consequences of laws that always get you.

You thought there was a good chance of litigation when you bought YouTube. The deal sets aside a $200 million reserve. Why do the deal if you anticipated so much hassle?

Because we think it's fantastic. I mean, we really do think that the YouTube phenomenon is a sustainable phenomenon for many, many years. And the argument is very simple: People are using video everywhere. People are building communities of people who use video. They're sharing them. YouTube's traffic continues to grow very quickly. Video is something that we think is going to be embedded everywhere. And it makes sense, from Google's perspective, to be the operator of the largest site that contains all that video.

Obviously, we would like it to include licensed, copyrighted content, legally, and make money on it. But YouTube itself can pay back - and this is where the critics get it wrong - YouTube itself can pay back in simple searches. Because, remember, when you go to YouTube, you do a search. When you go to Google, you do a search. As we get the search integrated between YouTube and Google, which we're working on, it will drive a lot of traffic into both places. So the trick, overall, is generating more searches, more uses of Google...

Which generates more pageviews, which generates more advertising revenue.

You got it.

The other interesting thing about pageviews is that we make our money by improving the quality, not the quantity, of ads showed on a page. This is very confusing to people. In a normal media business, you make money by showing more ads.

What does it take to improve the quality of ads on Google?

More computers, basically, and better algorithms. And more information about you. The more personal information you're willing to give us - and you have to choose to give it to us - the more we can target. The standard example is: When you say "hot dog," are you referring to the food, or is your dog hot? So the more personalized the information, the better the targeting. We also have done extensive engineering work with Google Analytics to understand why people click on ads. That way we can actually look at the purchase and go back and see what buyers did to get there. That is the holy grail in advertising, because advertisers don't advertise just to advertise, they actually advertise to sell something.

How big is the market for all these Google ads?

Today, the vast majority of our revenue is in text ads correlated with searches. In the last couple of years, we have developed what are called display ad products, including banner ads, video ads, click-to-call ads, and things like that. And I've also said that we are pursuing the possibility of television advertising. By that I mean traditional television advertising. And we bought dMarc Broadcasting to do radio ads.

So let's rank the probability of them being affected by targeted ads. There's search: That's 100 percent affected. What about radio? Is it possible to get a targeted ad right to your car right now? Not yet because we can't target the individual receiver in your car. If two cars are next to each other, the same radio station cannot have two different ads. However, if it's at a regional level we can do it to the zip code level. So let's call that partial targeting.

Now, let's look at television. Every one of the next generation of cable set-top boxes is going to get upgraded to an IP-addressable set-top box. So all of a sudden, that set-top box is a computer that we can talk to. We can't tell whether it's the daughter or the son or the husband or the wife in a household. All we know is we're just talking to the television. But that's pretty targetable because family buying patterns are pretty predictable, and you can see what programs they're watching. And if you're watching a YouTube video, we know you're watching that video.

My point of going through this little treatise is to say, if the total available market is ($600 billion to $800 billion, we won't be able to target all $800 billion. It will not be a 100 percent perfectly targetable, straight into your brain, but we should be able to offer a material improvement (in response rates) to many businesses.

Do ad agencies want that kind of automation?

Sure they do, because the advertisers do. You'd be amazed at how sophisticated the ad agencies are now.

That wasn't true a couple of years ago.

No, but it's changed. This is actually an important piece of data. When we started talking with them, the ad agencies were not sure what their role was going to be. And, in some cases, we were at odds with them. That is all gone as best I can tell. The ad agencies now see us as a major new revenue stream because all of the advertising models that I've described require the services of an ad agency. Somebody still has to produce the targeted ad, somebody still has to figure out what the demographic is. Somebody still has to figure all that out.

Don't you guys do that?

Well, we certainly don't make the ads, and we're certainly not the creative people. All we are is a targeting mechanism. We're just a distribution channel. So we need these ad agencies. And I'll tell you - and I spend a lot of time with these global ad agencies - I can tell you that it is very impressive how quickly they have changed.

Are advertisers going to start actually producing video ads to run on YouTube?

Absolutely. These ad systems tend to produce a lot of video they don't use. So for a 30-second ad, they actually will shoot hours of video. With that they can do the five-second teaser and the 10-second teaser and the single-shot still, and the low resolution one and the high resolution one - and they have terminology for each of these and ad formats for each of these. So the Internet, for them, represents a new creative medium. So we will see the emergence of new categories of ads and ways of making money.

When you and I talked a while ago, you talked about Dell and Sony and Intel as being sort of models for how you manage. Still true?

Very much so.

Despite the fact that Dell and Sony have had issues.

I don't think that those are management issues. I think those are just changes in their ecosystem.

Google's revenue and employee head count have tripled in the last two years. How do you keep from becoming too bureaucratic or too chaotic?

It's a constant problem. We analyze this every day, and our conclusion is that the best model remains small teams running as fast as they can and tolerating a certain lack of cohesion. The attempt to provide order drives out the creativity. And so it's a balance.

How do you keep different groups of engineers from unintentionally duplicating one another’s work and wasting resources?

Well, there is some duplication, but most of it is avoided through communications. The information systems that are within the company are quite good. But we've reined in certain things. For example, we don't tolerate the kind of "Hey I want to have my own database and have a good time" behavior that was very effective for us five years ago because the cost of this from a manageability, maintenance, and scaling perspective is a problem. So virtually all of the product groups are now told, "Build on top of this common set of services. Now, we internally use exactly the same code running on the same servers - like Gmail and Calendar and Google apps - as our customers do.

You mean you eat your own dog food?

Yes.

Google is a global corporation. What do you do to make employees in other countries all feel like they are working here in Mountain View?

It's a great unsolved problem. We do videoconferencing; we do a lot of visits where people are invited to one of the main campuses for a month or two. So they feel a part of a bigger entity when they go back. And that model does work. Of course, we do all the normal meetings - the sales meetings and training meetings, and all that. More and more of our time is being spent on that.

What about "20 percent time" - the time everyone is supposed to allocate in their week to personal projects?

It's still essential. Virtually all of the innovation at the company is still coming out of 20 percent time.

How do you and Larry and Sergey divide your duties?

Pretty much the same as we always have. Today, I can tell you that Larry and Sergey spent all day in the boardroom doing product reviews. They haul the engineering teams and product managers into a room, and they go through each product in great detail. At the same time, I'm interviewing some (prospective) executives, talking to a couple of partners about potential deals, and then in a few minutes the three of us are coming together on stage to answer employee questions. That's a very typical day. Today is Friday - Friday is essentially an all product-review day on their part. I do the same meeting on Tuesday, which I run and they attend.

I think it's fair to say that the skill sets (of the three of us) are just as complementary as they were five years ago. They have brilliance and technical understanding, and they're quicker in some things than I am. They're very clear thinkers. And there is my background of knowing how to scale things (grow a corporation). I think the combination has worked very well, and it's not going to change. We're going to do this for a long time. We enjoy what we do. We like to work with each other. And we're all the best of friends.

Microsoft CEO Steve Ballmer called Google a one-trick pony. Thoughts?

I don't think it makes sense for me to comment on words and actions by Steve Ballmer. You can phrase the question in a way that does not involve him, if you want.

Google gets its revenue from online advertising. One could make the argument that it is not diversified enough. Is that something that you think about? If so, what are some of the things you are doing about it?

The criticism is correct. We do get the vast majority of our revenue from advertising, and it's a business that a lot of other people would like to be in. So the first thing is, let's understand that we're in a great business. Also, there are some emergent models for revenue that are very interesting. The one that is probably most interesting is Google Apps. We're now beginning to get some significant enterprise deals. Basically, companies are tired of dealing with the complexity of the old model, and our products are now strong enough that they really can reliably serve a corporation.

Implicit in any kind of conversation about Google Apps is the fact that if it is successful, it will take market share away from Microsoft Office. Do you agree?

It may very well be that what you said is true, or it may very well be that consumers will drive us to solve completely new problems.

Why do you place such a premium on hiring the smartest people and developing and releasing software so quickly?

Fast learners win. We're in new, uncharted space. So the traditional assumptions that you and I might have about the future might actually just be wrong. There might be a new answer. And the only way to discover that is to put out your idea and then test it. And we track the results of that very, very, very rigorously, and this is not something we talk a lot about, but it's critical for us. How are these new ideas doing? What's their growth rate? What are the issues around them? And we push. What can we do to accelerate the development of this feature? What's the new problem? What's the new opportunity?

Google gets criticized a lot in the following way: It's a great search engine, but all the other products it has invented and released haven't done well at all. Thoughts?

The person who's saying that does not understand the economic leverage of the company. We know that Google Earth and Google Maps have had a tremendous impact on Google traffic, users, brand, adoption, and advertisers. We also know Google News, for example, which we don't monetize, has had a tremendous impact on searches and on query quality. We know those people search more. Because we've measured it.

Another example is Google Base, which we have been derided about for years. Google Base is how we get structured data. The quality of our index is better because of Google Base. The computers are smarter because of Google Base. Now, would you say Google Base is a mistake? Under the initial formulation, it's a waste of money because we don't monetize it. We could, but what we really do is we use it to improve search.

So one way of thinking about it is it all gets back to search. If you think about YouTube, YouTube is a "searching the world's videos" problem, right? They all have to be there, but how do you find them? What I guess I'm trying to say is that search is still the killer app.

As Google Challenges Viacom and Microsoft, Its CEO Feels Lucky