Web 2.0 Podcast: A Conversation with Jeff Bezos

by Daniel H. Steinberg

What business is in? Founder Jeff Bezos tells Web 2.0 Summit program chair Tim O'Reilly that there are three components. He says that everyone knows about the consumer-facing and seller-facing businesses. Bezos spent most of his time talking about their developer-facing business, with elements such as S3 (Simple Storage Service) and EC2 (Elastic Compute Cloud). Amazon's goal is to take the pieces of the heavy-lifting infrastructure that don't differentiate your business from any other and provide them "by the drink" so that they change from a fixed cost to a variable cost.

You can download the audio as an mp3 or download the video as an mp4. Here are all of the Web 2.0 Summit Podcasts released so far. You can also subscribe to the audio podcast or to the video podcast.

Intel Software Network Intel Software Partner Program

This episode is sponsored by the Intel Software Network.

The following transcript was created by Casting Words

ANNCR: Day two of the Web 2.0 Summit 2006 began with a conversation with founder Jeff Bezos. Now a conversation with Amazon's Jeff Bezos.
Tim: Our next speaker is somebody who I know very well. He's now my largest customer in my book business. He is a friend. He is sometimes a competitor. He is a tireless innovator. I tend to think of him as the Energizer bunny of Web 2.0 because he keeps reinventing himself and never stops. Jeff Bezos is going to start with a brief presentation so that we have a common ground to talk from. With that, let me invite Jeff Bezos, the founder and CEO of

Jeff Bezos: Thank you Tim. I'm also an inveterate attendee of Tim's excellent foo camp.
Tim: Which I recommend to you all except there's no room.

Tim: It's even worse than this.
Jeff: It's kind of just a tease rather than just a recommendation I'm afraid. Talk to you a little bit today about what Amazon is doing with web scale computing. We'll have a conversation with Tim. This is the Amazon that people know about: sixty-one million active customer accounts, we'll do over 10 billion in sales this year, 35 different retail categories, we're in seven different countries. Many of you in this room are probably Amazon Prime members. Bless you, thank you. We have 1.1 million active seller accounts and 10 million square feet of fulfillment center space.

The business that people know a little less about is our developer centric business. So now we have 200, 000 active developers registered using that service. We have ten different web service offerings. I'm going to talk briefly about Simple Storage Service, S3, and the Elastic Compute Cloud, EC2.

S3 is an aptly named service. It's very, very simple. All you can do with this service is put, get, and delete data. Storing data in the Cloud is 15 cents per gigabyte per month. A gigabyte transferred for bandwidth is 20 cents. So it's very inexpensive, designed to be highly available, extremely easy to use.

EC2 is, just as S3 is storage in the Cloud, EC2 is compute in the Cloud. It's 10 cents per compute hour. If you work out how many hours there are in a month, it's about $70 per month to have a server that's equivalent to a 1.7 gigahertz x86 box. But the really cool thing about it is that it's elastic, so you can either have one server for a month for $70 or you can have 700 servers for an hour for $70. It really changes how you can think about what kind of applications you can build, what you can do with that kind of compute capacity in the Cloud.

We've gotten a lot of these things are new we've gotten a lot of very strong levels of excitement in the blogosphere and from customers. Just recently, Linden Labs Second Life, when they released a new client, used the service to distribute their clients. What's really cool about this is that they did it right in the middle of the distribution and they did it incredibly quickly without even talking to us. It's completely self-service and very inexpensive.

This is what we're trying to do with these kinds of fundamental web services is to empower developers. We're a small company with big ideas. We now have the capability to build killer application. Infrastructure is no longer an issue. Just wanted to say thanks.

We're a little bit surprised by how excited people seem to be by these services. There's certainly nothing new about compute and there's certainly nothing new about storage. People have been remembering things and persisting things for a long time. So why are people excited about this?

I have a theory about this which is that people are excited because they see a future where they will be able to go more quickly from their idea to a successful product. I think they are seeing that in a lot of different ways. I think Amazon web services and things like S3 and EC2 fit into that framework for people. So they see it as a piece of this mosaic that is very empowering.

The reality, of course, today is that if you come up with a great idea you don't get to go quickly to a successful product. There's a lot of undifferentiated heavy lifting that stands between your idea and that success. The kinds of things that I'm talking about when I say undifferentiated heavy lifting are things like these: figuring out which servers to buy, how many of them to buy, what time line to buy them.

Eventually you end up with heterogeneous hardware and you have to match that. You have to think about backup scenarios if you lose your data center or lose connectivity to a data center. Eventually you have to move facilities. There's negotiations to be done. It's a very complex set of activities that really is a big driver of ultimate success.

But they are undifferentiated from, it's not the heart of, your idea. We call this muck. And it gets worse because what really happens is you don't have to do this one time. You have to drive this loop. After you get your first version of your idea out into the marketplace, you've done all that undifferentiated heavy lifting, you find out that you have to cycle back. Change your idea. The winners are the ones that can cycle this loop the fastest.

On every cycle of this loop you have this undifferentiated heavy lifting, or muck, that you have to contend with. I believe that for most companies, and it's certainly true at Amazon, that 70% of your time, energy, and dollars go into the undifferentiated heavy lifting and only 30% of your energy, time, and dollars gets to go into the core kernel of your idea.

I think what people are excited about is that they're going to get a chance they see a future where they may be able to invert those two. Where they may be able to spend 70% of their time, energy and dollars on the differentiated part of what they're doing.

For us, web scale computing means these things: it has to be elastics, so you can scale it up and down quickly, low latency, reliable and always available, simple, inexpensive and it has to be paid by the drink so that people can use it in very small, variable quantities if they want to.

Here's some of the companies that are using these services today ranging from Smug Mug, a professional photo sharing web site that is doing very well to Microsoft. We now see a couple of new customers today. One is Power Set It's a startup company that many of you probably already know about. It has a lot of buzz. Very brainy team of folks working on natural language search and they have decided to use EC2 for their back end infrastructure.

Xerox Global Services, if you don't know what they do, this is the group that if you have five acres of documents in file cabinets and you need to get them digitized, OCR'd and indexed they do that for you. It's a very compute-intensive task and it's very spiky. They're actually using not only S3 and EC2 but also our simple queuing service, which I didn't talk about today. If you want to learn more about this, you can go to Just remember, we make muck so you don't have to. Thank you.

Tim O'Reilly: That was great. You might want to sit a bit further...
Jeff: Where?
Tim: Oh just sit down.
Jeff: But further from you?
Tim: No just a little forward. Just sit down a little forward. You were disappearing into the couch.
Jeff: That is a cushy couch. Very comfortable.
Tim: So anyways, I guess the big question is I was blown away. Microsoft is a customer. You would think Microsoft would be in this business not a customer of this business. How does that happen?
Jeff: I think that Microsoft has a lot of different business units and I'm sure that they think about infrastructure services as well. But they also have other units that want to get work done today.

Tim: That's sort of a damning comment.
Jeff: It wasn't intended to be.
Tim: I guess the big question that must be on everybody's mind, this is a fabulous set of services. A lot of people have been nibbling around the edges talking about it. Eric was talking yesterday about how this is just a fundamentally better development model, developing for the data center rather than for the software artifact. But at the end of the day, it's not Google or Microsoft that is offering these services. It's a retailer. Clearly you're thinking about more than what other people are imagining what Amazon is. Maybe you can tell us, what do you see as the future of Amazon?
Jeff: I know that you've talked about us in a very positive way as a dark horse in this area. I have a different opinion about that. I don't know what the opposite of a dark horse is -
Tim: Well you're the leader right now.
Jeff: I think we're a light horse, a lightly colored horse or something. But this is what we've been doing for 11 years, operating a web scale application, a very complex one, that is based on a lot of transactions with high availability, reliability in a high volume, low margin business where we need to be very conscious of our cost structure.
Tim: So actually you're saying that your low margin is actually a competitive advantage.
Jeff: I'm weird that way. I always think high volume, low margin businesses are more defensible because they don't provide a deep underbelly that other people can go after. What we're always looking at is absolute dollar level of profit and free cash flow rather than percentage margins.

But, yes, we have to be very efficient at in order to offer the kinds of pricing, free shipping, and things we do. We've always operated our infrastructure with that mentality. What we've done is take the things that we were really good at internally and figure out how to expose those and charge for them.
Tim: Right, but why? The street is obviously conflicted about this because your stock is doing quite well. But the analysts are all going, what?
Jeff: I think why is a really good question. I'll tell you the direct answer and then I'll tell you more generally what I think about why. The answer is we've been doing this for 11 years. We didn't just start doing it. For several years now, we've been working on these services.

We sat there two years ago at Web 2.0 and you interviewed me. You asked me, "What is Web 2.0 to you?" And I said I think people have come to a different conclusion, it's kind of a semantic argument really. To me Web 2.0 was computers talking to computers. Web 1.0 was humans talking to servers and Web 2.0 is a lot of these things that you see with web services.
Tim: I'm also remembering, after you heard me give a talk about some of these things back in 2003, I came up and gave a talk at Amazon for your developer group. You complained because I didn't use one of the lines from the talk I had given at our E-tech conference, which was a platform beats an application every time. I remember you jumping up in the back saying "You didn't make that point!".
Jeff: You also neglected to make the great point that you always make. I invited him to talk, and he has many great points and he never makes them. But the other one that I always love that you say is follow the hobbyists, which I think is good advice.
Tim: You're certainly enabling a lot of -
Jeff: Yeah. The point is, why are we doing this? We're doing this because we're good at it, we know how to do it and we think it can be a meaningful, financially attractive business one day. We think this is a great business. I would also say, more generally -
Tim: Will it replace your other business or will you...?
Jeff: No. We have three businesses today. They're all customer facing businesses. We have a consumer facing business, a seller facing business, and a developer facing business. The consumer facing business and the seller facing business are the businesses that drive the economics of Amazon today because they're so large. Both of those businesses do.

The seller one is newer than the consumer facing one, but it's already big enough to have a meaningful impact on the financial results. I believe that one day our developer facing business will have a meaningful impact, too. But let me just say when somebody asks you why, why do this. I think sometimes that's a very good question and it's worthy of an answer. I think sometimes the right response when you're thinking about things the more powerful question is why not.
Tim: Right, OK. Well, because you could. [laughs]
Jeff: Because we think it's an attractive business opportunity.
Tim: So Jonathan's going to hate me for asking this question, but at Sun they started talking a while ago about grid computing. They put together this big grid. They didn't get any takers as far as I can tell, or certainly not much uptake. You've done pretty well right out of the box with EC2. Do you have any thoughts on why the difference?
Jeff: We have a long standing practice, which I will adhere to, of not talking about other companies. But what I can say about EC2 is it's designed to be simple to use. It's designed to be self service. And it's very inexpensive. I think all of those things are important. I think that that is -
Tim: I think there's a real Web 2.0 lesson here. A lot of people focus on the social aspects of this but there really is just that business practice self service, easy to use, and then you really are servicing the long tale.
Jeff: Right. If you have to sign a contract with every one of your customers, and negotiate rates, and do all of that, you're never going to get the small companies, and some of those small companies will grow into big companies.
Tim: So where does Alexa, and A9, you know Search Crawling, which is another set of services, where do they fit into this whole picture?
Jeff: It's the same thing. The basic idea is that there are certain pieces of heavy lifting infrastructure, that you can take all that fixed cost, and resell it to people by the drink as a variable cost.

So, Alexa has had for many years one of the best web crawls of the whole web. Now crawling the whole web is a heavy lifting exercise. But if you can buy web crawl by the drink, and convert that from a fixed cost into a variable cost, we think over the long haul that's a very attractive service to offer to people.
Tim: Right. So what other pieces are missing? You know like if you were sort of to imagine that you're gradually building out this suite of muck. What's missing from the muck mix?
Jeff: Well there are several pieces missing, and some of them we've been working on for more than two years, and we still haven't released them - and you know I'm not going to tell you what they are.

Tim: Let me ask you about one specifically - how about OneClick as a web service? I'd use it in an instant if it were priced right.
Jeff: I think that's a very good idea. I mean you know it's something...
Tim: So when can we have it?
Jeff: When can we have that?
Tim: [laughs] Self service.
Jeff: What times is it? You know I...
Tim: Yeah, shopping carts suck.
Jeff: Shopping carts?
Tim: Yeah, you guys have the best one out there.
Jeff: Well we also have, I mean, I think another interesting service that may be instructive to people who are curious about this, that I think really catches people by surprise, is the one we've just released in beta called, "Fulfillment by Amazon." Fulfillment by Amazon is a very simple web service, where again we've made it completely self service, and you can, using web service API's, you can make a call and notify us that you are about to ship something to our fulfillment centers.

We have this global network of fulfillment centers, they all run on the same software. You can ship something to us and we will receive it, using web service you've given us advanced notification that we should be expecting it. We get it, we stow it, and then once we've stowed it, you can send us more messages, you can make more calls to us, and when you do, we will pick those things that you shipped to us, and send them to any address that you specify.
Tim: So in a way, one of the things that is also a competitive advantage besides your low margin, is that you actually can deal with stuff in a way that nobody else can? I mean everybody else can do bits, you're the only ones who can do stuff?
Jeff: Sure, absolutely, and here what we're doing is we're saying, "We've got this big, you know fulfillment center network that does 10 billion dollars a year in sales, with very low selling priced items, " so it's you know, hundreds of millions of transactions. The fixed cost to open a fulfillment center is so high, and so what we're doing is saying, "At pay by the drink, variable cost, you can use that fulfillment center, that world class infrastructure."
Tim: Well at what point do people using this excess capacity, cause you obviously you have enough capacity to service other people besides yourself, at what point does that start to improve your margins? I mean, how far are you from seeing measurable impact?
Jeff: We charge prices so that we can make money. With these things like the fulfillment center you make money on the very first day, because our own business is covering the fixed cost of that infrastructure. And as we you know, add new nodes to our fulfillment center network, the pricing that we put out, the pricing performed by Amazon is incredibly inexpensive, it's like 45 cents per cubic foot of storage space per month.

Something like that, I might have it wrong. But this is really cool, because it allows developers who are you know, it empowers them to, just by writing software, to treat this 10 million square foot network of fulfillment centers as a huge peripheral device. They can treat it just like a printer.
Tim: Right. So that's for, right - the solid simple storage service.
Jeff: [laughs] That's a better name.
Tim: Simple storage service for solids, maybe.

Tim: So let me ask you about power. Because one of the things that's sort of been emerging as a little back story in the data center wars that are looming.

Is - who gets power the cheapest, you know Google, Microsoft, Yahoo, they're all building data centers right near damns. What's your cost of power? Is that something where you feel like you're going to be able to have a pricing...
Jeff: The biggest cost is not power, and it's not the servers, and it's not the people that maintain the data centers. The biggest cost is - is lack of utilization. It dominates all other costs. So yes, I mean because we're a high volume, low margin business, we are focused on things like power, and we're well positioned to be extremely competitive with these kinds of services, like EC2, and S3, and so on. But the dominant costs is, that everybody in the world who operates a data center, if you add up how often you use your servers, 24/7, you're going to find out that you use them like 17% of the time.
Tim: Right.
Jeff: So that's the big cost. You've got all this infrastructure, and air-conditioning, and etc., etc., and you're hardly using your assets - you bought a Boeing 747 and left it parked on the runway for 83% of the time.
Tim: So, all right. So this really is, in some sense, it is profoundly compatible with Amazon's core business in you using your assets up.
Jeff: Absolutely.
Tim: Yeah, so I guess...
Jeff: By the way we're not, that's how we got into the business. I think it's very important that our developer customers know, that we are in this business to serve "them." So it's not just - here are the things we needed to do for ourselves, and if you can make use of these that's fine.

What we do is, we're in the mode now, now that these things are launched, of listening and trying to figure out how we can make these services better, what kinds of things customers really need. So we'll be developing these services specifically for that developer customer set, and treating them as customers.
Tim: Right. Hey, one last question. We have a little less then 37 signals, I mean 37 seconds left, I want to ask you about...

Tim: ... you're relationship with 37 signals.
Jeff: Literally true, we have 18 seconds left.
Tim: It was 37 when I thought of it.
Jeff: These guys are incredibly humble in my opinion, and, but the fact is, I'm going to learn a tremendous amount from working with those guys. I think they're some of the smartest people out there who are, they have a spare approach that I think is very attractive, and I'm looking forward to being able to learn from them.
Tim: All right. Well listen, thanks a lot Jeff.
Jeff: It's my pleasure. Thank you.

Tim: Jeff Whitehorse.
Jeff: [laughs] Thank you.
ANNCR:'s, Jeff Bezos at the Web 2.0 Summit.

Daniel H. Steinberg is the editor for the new series of Mac Developer titles for the Pragmatic Programmers. He writes feature articles for Apple's ADC web site and is a regular contributor to Mac Devcenter. He has presented at Apple's Worldwide Developer Conference, MacWorld, MacHack and other Mac developer conferences.