In most of corporate America, you write the press release when your creation is finished. But at Amazon, you write it before you've even begun.
"If you were pitching something to Jeff Bezos or other senior managers below Jeff, the first thing you did was write a press release for it -- as if it were a product that you were putting out into the world," says Chris Brown, who spent more than three years at the company and remembers joining at least two pitch meetings with Bezos, the company's founder and CEO.
These Bezosian press releases are designed to focus pitches squarely on the needs of the company's customers, and they illustrate a much broader force that drives the Amazon machine. "That's one of the things that impressed me," Brown remembers. "If someone came up with an interesting idea -- if they said: 'Wow, I would find this useful' -- the next follow-on question was: 'Are there customers who would find this useful?'"
This is how Brown explains why Amazon -- of all companies -- created the Elastic Compute Cloud, an internet service that has completely changed the face of computing since it debuted a little over six years ago, providing instant access not to an online store or a search engine or an e-mail account, but to a virtually unlimited collection of computing power. Brown was among the many who helped gestate the idea, and he was part of the small team of engineers that built the service at an Amazon satellite office in Cape Town, South Africa.
Yes, Amazon is the world's largest online retailer. It made its name selling books and DVDs and so many other physical goods. But somewhere along the way, as the company worked to build new technologies that would make it easier to run its vast retail operation, Bezos and the rest of the braintrust realized that if Amazon and its partners needed new technology, so did the rest of the world. The result was the Elastic Compute Cloud and various other Amazon Web Services that would make it easier for anyone to run their own operations -- whatever those operatons might be.
With EC2, you can use all that computing power to run just about any software application you like, including a website such as Instagram or Pinterest, or a banking application that simulates credit risk, or a research tool that analyzes the human genome. Rather than loading your software on physical computer servers you've set up in a closet or a data center, you can load it onto virtual servers you've set up in your web browser. And whenever you need more virtual servers, you can have them.
The service debuted in August 2006, just after a complementary offering called S3, which let you store vast amounts of computer data without setting up your own hardware. Six years later, these and other Amazon Web Services run as much as 1 percent of the internet.
But more than that, these services have changed the way we think about computing. In the years since the launch of EC2, the likes of Google and Microsoft and HP and Rackspace have launched similar services, and there are countless other outfits offering to help you build your own EC2.
As the company holds its first conference dedicated to Amazon Web Services, this still seems like an odd success story for an online retailer. "I was shocked that they did it, and not someone else," says David Patterson, a University of California, Berkeley, computer science professor who started using EC2 for various research projects in 2006 or 2007. And, unfortunately, the evolution of these services remains rather murky -- even after conversions with Brown, several other ex-Amazon employees, and Andy Jassy, who wrote the business plan for Amazon Web Services and continues to serve as its "CEO." The story of EC2 is like Rashomon. Each player saw a different part of the story -- and some may have reason to omit parts they did see.
But you can see Amazon's corporate culture reflected in these seminal creations. And though some say Amazon will struggle to fend off competition from Google and Microsoft, the company is remarkably well suited to playing the cloud game.
Part of the genius of EC2 is that it gave software developers virtual machines that behaved a lot like the physical machines they were familiar with. They could run the same sort of software they had always used. Amazon didn't try to tell the customer what he wanted.
Google and Microsoft released beta versions of similar cloud services in 2008 -- Google App Engine and Windows Azure -- but these big-name competitors failed to completely grasp what made EC2 so successful. App Engine and Azure tried to make it easier to run software in the cloud, but in doing so they restricted what developers were able to do. The learning curve was steeper, and the public never really embraced them in the same way.
After leaving Amazon in 2007, Brown moved on to Microsoft, where he saw Azure develop first hand, and even then, he felt the company was missing the point. "I ranted at some of the architects at Microsoft that they were starting at the wrong end, that they were constraining the sorts of things you could do," he says.
"I actually wrote them an e-mail that said: 'These are the five ways you will be compared to EC2 the day that you launched,' and they were all about having control over things so that you could build and deploy stuff that you already knew, to get jobs done that you already knew."
This year, in a kind of tacit admission that Amazon had gotten it right, both Google and Microsoft unveiled services that look a lot more like EC2.
David Patterson argues that Amazon didn't know what would catch on any more than Google or Microsoft. "They were all running a great experiment," he says. And to a certain extent, Brown bears this out. "We had no idea if it was even going to work," he remembers. "As an infrastructure geek, I found it to be an interesting experiment, and I thought that other infrastructure geeks would too, and suddenly, it becomes this giant thing that everybody knows.... I realized that it had changed the world and I thought: 'Wow, this is not what I set out to do."
But however the company got here, Amazon is now well positioned to fend off the competition. Jassy says that Amazon didn't necessarily plan it this way, but EC2 and the other Amazon Web Services are businesses of low margins and high volume -- the kind of businesses that Amazon knows better than anyone else trying to play the cloud game.
"Amazon is very good at operating in a low-margin environment, and Jeff is very proud of this," says Chris Pinkham, who oversaw the development of EC2 before leaving the company around the time it launched. "He feels that low margins promote customer loyalty and -- frankly -- inhibit competition. I don't know what Amazon's margins are right now, but there are some significant forces on its side."
