Amazon.com made $10 billion in fiscal 2006 selling items in 35 product categories to some 59 million registered users around the world. To do that, dozens of data centers must process millions of transactions and terabytes of data each day.
“We work as hard as we can to make the guts of Amazon invisible to consumers and sellers,” says Amazon chief executive officer Jeff Bezos, who delivered a keynote speech at Technology Review’s Emerging Technologies Conference on September 27.
Now, however, the company is bringing its invisible back-end operations front and center. In a series of initiatives launched over the last year, Amazon has begun to rent out parts of its IT infrastructure as Web services.
Startups in the social-networking or media-sharing arenas, for example, can use Amazon’s S3 database service to store their users’ photos and videos, rather than buying their own servers. In another service, the Elastic Compute Cloud, or EC2, business customers can offload computing-intensive tasks to temporary “virtual servers” at Amazon’s data centers. And last week, Amazon announced two additional services, Webstore by Amazon and Amazon Fulfillment, that allow outside companies to build their own commerce sites around Amazon’s software, then ship products using Amazon’s own warehousing and shipping network.
In essence, Amazon is finding ways to monetize the software and computing resources it has developed over the last 11 years to handle its own business. “There’s a big need at many companies for back-end infrastructure,” Bezos remarked. “The problem is server hosting, bandwidth management, and the like really have nothing to do with your actual business.” Amazon wants to use the Web services it’s developed to “help with some of that heavy lifting,” Bezos said.
Technology Review contributing editor Wade Roush followed up with Bezos after his speech.
Technology Review: In your keynote talk you spoke about the benefits of S3, EC2, and other Amazon Web services for small businesses that don’t want to invest in building data centers. But isn’t there a benefit in it for Amazon as well? You get to put the vast computing and storage capacity you’ve built around the world, which would otherwise go unused, to gainful employment. I imagine there are times when you’re using less than half of your capacity.
Jeff Bezos: There are times when we’re using less than 10 percent of capacity. And that is one of the advantages of doing things this way–it promises higher rates of hardware utilization. That’s a system-wide efficiency that should make everybody happy.
But our thought process for getting here was less about underutilization and more about our own developers–building tools to lower the “cost of coordination.” Web-scale applications are hard to run reliably at a small scale, and it doesn’t get easier as you go to larger scales. At Amazon’s scale, one of the issues is coordination between the application development team and the network infrastructure team inside the data center.
Say you’re a team leader at Amazon responsible for the personalization service, and you want to make some improvements. You have to plan a set of server deployments and coordinate that with the infrastructure team several weeks or months, or even as long as a year, in advance. The cost of that coordination can be very high.
The first Web services we did, about three years ago, exposed transactions and e-commerce data, things like our catalog of products. But, as we went on, it became clear that the more primitive services that we had really built for ourselves would be very appropriate for use by developers and other companies. All we had to do was add some billing functionality.
TR: You’re talking mainly about infrastructure services today, but as you just mentioned, Amazon’s first foray into Web services was creating interfaces so that outside developers or webmasters could access your databases, for example, to grab product data that they could then display on their own pages. What’s happening in that area?
JB: There’s a significant team of people who keep improving and extending those services, and that’s a very significant source of customers for Amazon.com. It ties into our associates program, so the people using these Web services earn a commission when they send us a customer who buys product. That continues apace. One of the latest things we announced, back on September 19, is WebStore by Amazon. [The program allows businesses, such as Seattle’s Gifts, to run their own branded e-commerce sites using Amazon’s back-end infrastructure. –TR] That continues to develop.
TR: In your talk you mentioned Fulfillment by Amazon, which was also announced on September 19. Explain how that works. It sounds a little like FedEx, except that an item spends some time in your warehouse before it reaches a customer.
JB: What you do is, you send us a case or a pallette of a product you want us to distribute. You might be a bricks-and-mortar retailer and want to do e-commerce as well. So you send us a portion of your products, and we store them in one of our fulfillment centers. We charge 45 cents per month per cubic foot of shelf space, or about $5 per cubic foot per year. We designed the program so that a used-book seller, for example, can very profitably do business with that model. You pay for the shelf space, and then when you send us a request–which is essentially a Web services call–we ship the product.
TR: Amazon, Google, eBay, and even Microsoft are all rolling out Web-based infrastructure services of different varieties, building on what they’ve learned about how to run their own businesses. We may all eventually do most of our computing within the cloud of computing and storage resources these companies are making available. Can you imagine Amazon evolving into a company that’s as famous for being a kind of Web utility as it is for being an e-retailer?
JB: No. This is a completely separate business that will grow up in its own way. We really have three businesses at Amazon: our consumer-facing business, our seller-facing business, and our developer-facing business. Of course they interact–the seller-facing business supports the consumer-facing business, for example, when used-book sellers interact with their customers on the Amazon website. But the way we’re organized, Web services is a separate subsidiary–Amazon Web Services LLC–and that group of people is focused on our external developer customers.
So the goal is to have the retail business continue to grow in its own right and to have the developer business grow in its own right. They aid and abet each other. Today, Amazon itself is the biggest user of things like EC2 and S3. We’ve been our own beta customers. But one day my hope is (to answer your question about the future) that Amazon will be just one of many big providers of infrastructure services, not the big provider.
TR: Over the long term, though, meaning 50 years or more, companies can go through pretty large transformations that no one would have predicted. IBM, for example, was once known mainly for making office equipment, then mainframes and PCs, and now it’s largely a consulting firm. You’re experimenting with technology that could evolve in equally unpredictable directions.
JB: I don’t expect that kind of transition, because all of our businesses can thrive separately. The retail business is global and thriving, and so is the seller business. Maybe it’s semantics, but I see Amazon Web Services as an extension of what we do, not the beginning of a transition.
TR: I guess it’s not really fair to ask you to speculate about the 50-year horizon.
JB: By then, we’ll all have downloaded our brains into microchips.