Amazon's Jeff Bezos Dishes The Dirt On IT 'Muck'

At the Web 2.0 Summit in San Francisco on Wednesday, publisher and conference co-founder Tim O'Reilly introduced Amazon CEO Jeff Bezos as "the Energizer Bunny of Web 2.0." That's a compliment. After more than a decade with Amazon, Bezos keeps going and going and going, something that can't be said of many Web companies that followed in his footsteps. Recently, he's been banging the drum for Amazon's newest business: selling computing infrastructure by the sip. He calls it "muck." He sat down with InformationWeek at the Palace Hotel in San Francisco to discuss Amazon Web Services.

InformationWeek: What would you say to convince CIOs that they could really move their infrastructure to Amazon?

Jeff Bezos: I would say we have demonstrated the ability to be good at it. We eat our own cooking. So if these services aren't highly available, highly reliable, low-latency, then our own business suffers. Then I'd also say, over time the comfort level of people using these services will increase. People who are using them today are fantastic early adopters. They have a very strong business needs to use these services. They're saving a lot of money in many cases in using these services. And as the experience base of people using these services continues to increase, other companies will look at it and say, "Well it's been working for companies A, B, C, and D, and seems to be working well. So I'm going to give it a try as well." There's a sort of natural life cycle for these things where people develop comfort levels at different rates.

InformationWeek: Is Amazon Web Services better suited for a smaller company or could you just as well move YouTube's or Visa's internal operations over tomorrow?

Bezos: You could move large pieces of infrastructure and we do have large clients using it in different ways. Xerox Global Services is an example of a very large customer with very spiky utilization, so it's an ideal service for them, because of the spikiness. It's the kind of company where they will have a customer that says, "I have five acres of filing cabinets and I need to digitize it all, OCR it all, and then make it all searchable." So those kinds of things are very compute intensive, but they're projects. They have a beginning and an end. And if you're Xerox Global Services, you don't necessarily want to have to carry that compute capacity to be able to do those things 24 by 7, if you're only doing it in these spikes. With the right amount of due diligence and the right amount of care and thoughtfulness, very large services could move onto this kind of infrastructure.

InformationWeek: It seems that Amazon has scaled back on its efforts to develop its own search technology. Is this a missing piece that's going to hinder Amazon's ability to compete in the software-as-a-service platform war that seems to be shaping up with Google and Microsoft? Or is that is that in some ways a fiction invented by the media that just wants to write about blood sport?

Bezos: It is true that we humans love stories about conflict, so if things can ever be viewed through that prism they get instantly more interesting. But I actually believe that ... industries generally succeed or fail. And generally there are a lot of winners when an industry succeeds. So it's not exactly like a sports metaphor where there are winners and losers and the conflict is very clear. I believe that Web services, and the kind of infrastructure services that we're building, is going to be an important industry. I think it's going to be a meaningful third business for Amazon.com. But I think there will be multiple winners.

InformationWeek: It seems that with Amazon Web Services you're removing a lot of the barriers to entry that traditionally would have insulated businesses from competition. Companies like barriers to entry because they provide a little bit of breathing room and little bit of profit margin. What are the new barriers to entry going to be in a world where you've got all these companies operating on your undifferentiated infrastructure?

Bezos: The game should move to a higher level where the unique parts of what you do become the source of your differentiation, instead of the kind of heavy lifting, price of entry thing. What we're doing is leveling the playing field so that small companies can have access to the same low-cost structure as big companies for very reliable backend infrastructure, and to do that in a pay-by-the-drink way -- so that you don't have these big fixed-cost steps that you have to subject yourself to. But there will still be whatever it is that the individual companies is doing ... is it a photo sharing Web site? So SmugMug is a user of our services. They are a very good photo-sharing Web site. And they will create differentiation on the part of their business that really matters to their customers. Their customers don't care at the end of the day what piece of infrastructure is holding their photos, as long as it's reliable. And what SmugMug really cares about is how are they going to make their photo-sharing service better than everyone else's. It's already an excellent service and they can keep adding to it and put their energy into those things that truly differentiate it for their customers instead of just paying for the price of entry sort of things. Nobody really cares whether you're using these servers or those servers, but you have to put a lot of thought and work and energy into those things. And a lot of that energy doesn't help you compete.

InformationWeek: Is there any risk that you'll end up creating more competition for yourself?

Bezos: I will be disappointed if we don't. (Laughs) Because we're opening up things like our fulfillment infrastructure and we're saying you can come, and for 45 cents per cubic foot per month, you can store things in our 10 million square feet of fulfillment center space, and use Web services to inform us -- it's called Fulfillment By Amazon -- you can inform us that we're about to receive those things. And we'll receive them and stow them. And then you can make more Web services calls and tell us to ... pack those things and ship them to a particular address. Does that enable competitors? Yes. These things are designed to be empowering. But our retail business will do great in the face of competition because that competition is going to exist whether we empower folks with these kinds of services or not. So we're very excited about leveling the playing field and moving the game to a higher level, where you have to compete above the infrastructure level. InformationWeek: Is there any chance we might see Amazon partner with a company like Yahoo to fill out its online ad platform? Or do you see yourselves developing your own ad platform?

Bezos: For Amazon.com's Web site, we've just launched a Web platform called Clickriver ... in the last couple of weeks. And it basically allows people to put ads on the Amazon Web site using that platform. But we're very open to partnerships that make sense for both parties. We've done a number of partnerships over the years and I suspect that we'll do more.

InformationWeek: Amazon has made a decent-sized bet on open source telephony as a central piece of its internal communications strategy.

Bezos: We use VoIP phones internally, if that's what you mean.

InformationWeek: Exactly. Does it bring Amazon anything other than lower purchase costs?

Bezos: Lower purchase costs, but also lower cost of administration. You can put the phones anywhere you have IP connectvity, so you have one phone cable instead of two sets of cables. It's simpler to administer. People move offices, they change their phone numbers. There are a bunch of advantages in terms of total cost of ownership. It's not a dramatic improvement in the quality of voice service or anything like that. The old phone system, in terms of connecting you with a clear conversation with somebody else, worked extraordinarily well. I don't see it as being a huge customer experience improver, except in terms of the total cost of ownership, which is a big deal.

InformationWeek: In your discussion at the Web 2.0 Summit with Tim O'Reilly, you mentioned that the biggest cost for infrastructure was not power but lack of utilization. Doesn't Amazon have to take on that cost? If you have a bunch of customers, you still have to have more capacity than you're typically using to account for spikes in demand.

Bezos: Totally true. You'll never get to 100% utilization. But what you can do, if you have a mix of different kinds of customers ... what happens is, if you're one kind customer and you have your own dedicated data center for your capacity, then you don't get to offset ... compare a situation where one company does rendering for computer graphics. Render Rocket is an example, actually, that does use our Web services using EC2. Their peak utilization times are going to be completely different from a company that does, say, e-commerce. So by pooling together dozens of different kinds of companies that have different kinds of utilization peaks and troughs, they tend to cancel each other out. And you'll still have peaks and troughs, even for the average pool, but they won't be as steep. And as a result, you'll get better utilization.

InformationWeek: I understand you're also involved in a space startup, Blue Origin. Is there any reason beyond that fact that space exploration is just a great dream to have?

Bezos: I'm just now trying to think about how we could open up Blue Origin as a Web service. It's not immediately clear to me. (Laughs) But if we can figure it out, we will.

InformationWeek: What's in space for you?

Bezos: This is a childhood passion of mine. It's a very talent team of people working on building a vertical takeoff, vertical landing, suborbital vehicle. It's a separate company and they're doing a fantastic job and I'm very proud of them.