CEO Andy Jassy: How Amazon Web Services won the cloud – and what’s coming next


Not too many years ago, Peter Burris, head of research for the analyst group Wikibon, recalls being in a room full of chief information officers. They were all giggling about Inc.’s attempt to persuade them to use its fledgling cloud computing services in place of the computer hardware and software on which they were collectively spending $10 billion a year.

Nobody’s giggling anymore. Scarcely 10 years after starting to offer computer storage services, Amazon Web Services is now producing revenues at a $13 billion annual run rate, growing 55 percent a year and accounting for most of Amazon’s profits. In the process, AWS has defined the new world of technology, one in which many or even most information technology products can be bought as online services.

AWS Chief Executive Andy Jassy, a longtime lieutenant of Amazon CEO Jeff Bezos, shepherded the unit’s surprising rise on the strength of an army of startup customers — including some now large upstarts such as Netflix Inc. that went all-in on AWS — and departments at larger companies. Already, he has outmaneuvered the world’s biggest and most venerable technology companies, such as Microsoft Corp., Oracle Corp. and IBM Corp. that have depended on selling computer hardware and software to the tune of tens of billions of dollars a year.

But Jassy isn’t stopping there. He’s now aiming much higher, hoping the capture the business of every large enterprise. Some, such as Unilever PLC, Capital One Financial Corp. and NTT Docomo, have already started using AWS services in a big way. And more are joining in. On Monday, shipping giant Matson Inc. closed four data centers to move computing operations to the AWS cloud.

Jassy faces plenty of challenges — now more than ever. Rivals are all targeting cloud services as well. And they often have more longstanding relationships with large enterprise customers that may give them an edge.

But so far, AWS is retaining a firm hold on the cloud market thanks to starting early and releasing a constant flurry of new features–likely more than 1,000 this year alone. “AWS certainly faces strong competition from both Google and Microsoft, but so far it’s been able to hold its own, in part because it has the first-mover advantage here for the kind of workloads it specializes in,” said Jan Dawson, chief analyst at Jackdaw Research.

On a recent rare clear November evening at his home in Seattle, Jassy spoke at length about how Amazon Web Services got this far, how Amazon’s management style keeps it churning out about three new features every day, and what he views as the next frontiers.

This is the first of a three-part series on that conversation that will run through Thursday. It was edited for clarity.

Part 1:  Cloud Landscape and the emergence of the Enterprise Cloud

Q: How do you see Amazon Web Services in terms of product and market maturity?

A: Even though the business is a $13 billion revenue run-rate business, it’s still growing at a fast clip: 55 percent year over year. And we’re at what we consider the beginning stages of the meat of enterprise and public-sector adoption in the U.S. Outside the U.S., they’re 12-36 months behind, depending on which country you’re talking about.

It’s so different today than it was even three years ago [when] startups were using AWS to build their entire businesses. For enterprises, up until a few years ago we used to have “if” conversations: “If we decided to use the cloud in a meaningful way, here’s what we would think about it.” These conversations now with the enterprise and the public sector are much more about “when” and in what order and how fast.


Enterprises move to the cloud

Q: How fast are enterprises moving to the cloud and how are they thinking about the process?

A: We continue to believe in the fullness of time — and I don’t know if that will be five years from now or 10 years from now or 20 years from now — very few companies will own their own data centers. And those that do will have much smaller footprints than they have today.

Especially if you’re building any kind of new application, there’s really no rational reason for not building out on the cloud. So there’s a huge amount of enterprise net-new application development on top of the cloud.

Photo courtesy of

Q: Since enterprises can’t usually move all their workloads at once, how are they taking the journey to the cloud?

A: It usually is three or four different tranches of sequencing to the cloud. Almost always enterprises gets started with dev and test [initial development and testing of cloud applications]. You see that all over the place: the Tokyo Stock Exchange, Lionsgate, Galata Chemicals.

The second tranche tends to be three different workloads. One is digitization. Almost all of these big enterprises now are going through IT modernization where they’re basically digitizing all their assets. Unilever has about 2,000 websites they built on top of us. The Navy is saving 50 percent of their operating costs running their websites on top of us.

Second is analytics. It’s just never been easier and less expensive to collect, store, analyze, and share data than it is today with the cloud. And again, whether it’s NTT Docomo running their large four-petabyte data warehouse on top of us, or the Financial Times doing analysis 95 percent faster at a quarter of their costs, it’s just so much easier and so much more cost-effective to run analytics in the cloud. That is often a huge early enterprise workload that they move over.

Third is mobile, because most companies have employees and customers that now have three screens between laptop or desktop and tablet and phone, and most of the applications weren’t built with those other interfaces in mind. So we have lots of companies that are either building all of their new applications as mobile-first or taking all their existing applications and making them mobile-usable.

Q: And the third tranche?

A: That tends to be companies that have a data center that’s coming up for renewal or a tech refresh, and they’ll work with us 12 to 24 months in advance to be able to close down that data center and move all of those workloads to AWS. Examples include Condé Nast that had about four months to move all of their datacenter to the cloud because they had a data-center renewal coming up. They moved hundreds of applications, including their most mission-critical — legal and HR and financial systems and several databases. They’re saving 40 percent versus what they were doing on-premises.

News Corp. took their data center footprint from roughly 75 down to about seven or eight, and we’re doing a lot of that data center migration with them. Capital One is moving from about eight data centers down to three, and a lot of that migration and a lot of that workload real estate is moving to AWS.

Q: What’s the last group?

A: Customers who decide they’re going to get all the benefits of the cloud as fast as possible, who go all in. Netflix was really the first of any renown that made that decision. When they made that decision seven or eight years ago, it felt like a very bold decision by Reid Hastings and Netflix. Today it’s a lot more commonplace.

You see more and more companies, whether it’s Suncor or Intuit that are moving all their applications to AWS. FINRA is all in and moving everything to AWS. You continue to see more and more companies betting their business and establishing business models on the cloud.


Drivers for the enterprise cloud

Q: What are the drivers to the cloud?

A: Cost is almost always the conversation starter. For most companies, if you don’t have to lay out all that capital for the data centers and the servers up front. You get to pay as you consume it. Turning the capital expense into a variable expense is attractive.

And it’s a lower variable expense than what companies can do on their own because we have very large scale that we pass on to our customers in the form of lower prices. You don’t have to guess what your capacity needs are as you need to scale up.

Q: What else?

A: Probably the single biggest reason that they move to the cloud is agility and the ability to move much more quickly. For most enterprises, if you ask them how long it takes to get a server to try some kind of experiment, you usually get answers that are between 10 and 18 weeks, which is really demoralizing for product development folks. They’re not really going to spend much of their free time thinking about new customer innovations because why bother? It’s going to be so hard to get a server.

Once you get the go ahead to experiment, then you’ve got to build all of the surrounding infrastructure services around it: compute and storage and database and content distribution and messaging and a data warehouse and analytics. Most folks would prefer if they didn’t have to build all of those infrastructure services and instead get to take advantage of the 70-plus services that AWS has available to them at their fingertips.

Q: So AWS is moving up the software stack pretty quickly. How are developers looking at those efforts?

A: They get to spend their time and their cycles on what’s different about their idea and their business rather than the infrastructure pieces. If you think about it, if you work at a company that primarily uses an on-premises infrastructure, you have much less functionality than you’re going to have in AWS and the cloud, and it moves much more slowly.

In 2013, we launched what we consider 289 significant services and features. In 2014, that number was 516. Last year, it was 722. This year I think we’ll end up close to a thousand. If you are a developer and you have an idea for a service, on average every day you get three new capabilities available to you, that show up on your doorstep for free, that you can either use or not use. And you only pay for what you use.

That completely changes what’s possible and what you’re going to build. Many more employees spend free cycles coming up with new ideas. They don’t have to recreate the wheel.

Q: AWS helped pioneer the use of application programming interfaces to build your cloud business. But now you’re offering more services directly. How are developers looking at using AWS APIs and services today?

A: APIs are always the fundamental building block for what we build. Early on — and still today — we got asked this question a lot: “Are you guys focused on IaaS [Infrastructure as a Service, the most basic level of cloud computing] or PaaS [Platform as a Service, the set of services needed to run apps]?”

You’ll never hear those terms used inside of AWS just because we don’t think of the world that way. A large number of those developers want to have access to the APIs … to stitch together whatever applications in whatever way they see fit. And we will always provide our services at those atomic levels for them to have maximum flexibility and creativity in how they build.

Q: But not all developers want that?

A: There are also a large set of developers who are willing to give up the control of having access to the APIs in exchange for ease of use and abstractions that are at higher levels. And that’s why we’ve built, over the years, so many things and, I guess, what people would consider PaaS layers — everything from drop-in containers with Elastic Beanstalk to DevOps tools like OpsWorks to confirmation with your templates to abstractions like our console.

Q: It’s not entirely clear to developers where AWS stands in the cloud software stack. Are you trying to move up that stack toward applications developers?

A: No. One of the reasons that we built in this decoupled API fashion was the lesson that we learned in the first eight or 10 years of building Amazon the retail business, which is: When you build at a much higher level of abstraction and you make a lot of those decisions for people on how things should fit together, you either get it wrong or you got it right — but then as the world evolved, it turned out you wanted to change it.  And having to change it when you don’t actually have the independent primitive building blocks, which are often the APIs, is much harder than if you actually built on these much more atomic units because you can then kind of refactor whenever you see fit.


Competition intensifies

Q: Now you have a lot of serious competition for the first time, including Google, Oracle, IBM and Microsoft. Are there new rules of engagement given that many enterprise customers are already buying other products and services from these vendors, which are now telling customers they should buy cloud services from them too?

A: The cloud is that it really eradicates the old-guard model of all the hand-waving and all the bombast and all the wild claims that vendors would make. Customers didn’t really have a chance to understand what was real before they had to make a purchasing decision. Vendors could make up what the product was going to be, or they could manipulate benchmarks.

Now customers don’t have to make these blind decisions about how the product is going to perform and what’s actually there before they buy. The cloud really is the great playing-field leveler for customers, and I think that’s a very different way of engaging with your infrastructure [party] than has been the case for the last 30 years.

Q: What was your strategy? How did you manage to work under the radar for so long?

The biggest surprise of the many that we’ve had in the 13 years we’ve been working on it and the 10.5 years we’ve been in the market is just how long it took. I don’t think in our wildest dreams, especially here in Seattle, that we thought we’d have a six-year head start.

In the beginning, we had to hire most of the initial team that built all the initial AWS services without telling any of those people in the interview process what they were going to work on because we were so paranoid about not letting anyone know we were building a technology infrastructure platform.

Q: But now, with pressure from competitors, you have to be more open to communicate what’s coming.

A: Two things help you with that. One is if you’re fast, if you’re agile and you iterate quickly like we do, customers don’t have to wait too long for a lot of the capabilities. The other thing for us: We are incredibly lucky because we have so many customer-feedback loops. We have very well trafficked forums. Our customers like to tweet and blog. We have a large sales team. We have a large support team. We have a large professional-services team.

And then we have data. And so a lot of times people have complimented us by saying, “Wow, it’s really cool that you released that because that’s just what we were thinking,” as if we were somehow prescient. But it’s really much more that we’re getting our information all the time. We’re just listening to what customers tell us matters.  

Q: So are you really data-driven?

A: Oh yeah. But there are also things that have nothing to do with what we’ve done with the past, where you can’t look at that data. You have to have an institutional capability to hear what customers are telling you and then have a way of figuring out which are the things that really are being broadly requested by a lot of customers. Then have teams that are autonomous enough and plugged in enough to their customers to hear that feedback, reprioritize their roadmap based on what they’re hearing matters most, and then iterate quickly on behalf of customers.

Tomorrow: How Andy Jassy manages an organization operating at massive cloud scale.

Photos courtesy of Amazon