Amazon.com, Inc. (AMZN) Earnings Call Transcript & Summary

Great. Thank you. We'll go ahead and get started. My name is Heath Terry. I lead TMT research for Goldman Sachs. As a Goldman Sachs employee, it's been a special point of pride over the last 3 years as, first, Roy Joseph and then David Solomon took center stage at the AWS re:Invent opening keynote in front of over 50,000 people. Having been in the audience for every re:Invent conference, it's now a particular point of pride for me to introduce Andy Jassy, Chief Executive Officer of Amazon Web Services. Andy joined Amazon in 1997 straight out of business school. And in 2003, he led that team in founding AWS, creating the world's undisputed leader in cloud computing, a company inside Amazon worth, according to our math, well over $0.5 trillion. Andy is joined in conversation by Goldman Sachs CEO David Solomon. Andy, David, welcome.

Good afternoon, everyone, and Andy, thank you for being here. You've got it a lot easier as we start this. When I was at re:Invent, you made me actually work before you put me on stage.

David DJ-ed before my keynote. It was quite good, by the way. People are still talking about your set.

Well, I mean, at 7:45 in the morning, it is not easy to DJ for 10,000 people.

You had good energy.

It was 7:45, but you try your best. So let's dive right into it.

You've been at Amazon for 22 years and you founded AWS with a little over 50 people in 2003. And you're going to generate over $45 billion of revenue in the business this year and more than half of the company's profits. And so when you put the financials aside but you just start to think about the scale of what you built, kind of quantify for us a little bit the scale of AWS and how fast it continues to grow.

Well, it's been kind of an amazing adventure, and the business now has a $40 billion revenue run rate growing 34% year-over-year. And we have millions of active customers, which we define as non-Amazon entities that use the platform in the last 30 days, but it's really a very broad and diverse group. It's -- in the early days of AWS, it was mostly start-ups building their businesses from scratch on top of AWS, and these were companies like Slack and Instagram and Pinterest and Stripe and Robinhood and companies like that. But what's happened over the last 6 years is that the enterprise and public sector have very rapidly adopted AWS in the cloud. And you see every imaginable vertical business segment now using the cloud: if you -- in financial services, we work with Goldman and Capital One and Intuit and the Commonwealth Bank of Australia; in health care, Johnson & Johnson and Merck and Pfizer and Bristol-Myers Squibb; in manufacturing, GE and Schneider Electric and Philips. It seems every imaginable vertical business segment now is using the cloud in a meaningful way. And in the public sector, we have 6,500 government agencies worldwide, 11,000 academic institutions and 29,000 nonprofits. So it's a very, very broad and fast-growing customer base. And there -- other than looking at revenue and the trillions of requests that we get in the various services all the time, most of the -- it's a little hard to be able to track everybody's relative market segment share just because we're the only ones that really publish the full metrics. But we have a pretty significant market segment leadership share. And where you -- where we really kind of see the breadth is, if you just look at your everyday life, if you look at ridesharing and if you use Lyft or Uber, Ola in India or Grab in Southeast Asia or mytaxi in Europe, all running on AWS in the cloud. Or if you use media services, Netflix runs completely in AWS and, obviously, Amazon Prime and Disney+. And if you watch the Super Bowl, you watch it streaming on FOX live on top of AWS. And if you use Peloton or if you use -- my son is a very big orderer of food, so Postmates and Grubhub and DoorDash. It's just very pervasive impact every day in your life. And yet, we always say, and we really strongly believe it, that it's still very early days in people moving to the cloud. And even though the business is growing fast and it's a reasonable size, I think it's going to be very different 10 years from now.

Yes. So one thing that's fascinating to me, a student of big businesses, is this was a significant innovation in a large organization. I mean, at the time Amazon -- much bigger now, but it was a large organization. It was a big innovation. What advice -- and I think about this, actually. What advice would you give -- Goldman Sachs is trying to innovate in its business. What advice would you give any larger company around innovating or building new business lines inside a strong organization?

Yes. Well, there are a few things -- a few things that have been helpful for us that I think -- it doesn't necessarily mean it would be helpful for everybody, but I think they're generally useful principles. The first is, you have to hire people who want to build. We disproportionately index in hiring builders. And we think of builders as people who like to invent, people who look at different customer experiences and are honest about what's not working and try to reinvent those customer experiences. And people who really understand that launch is the starting line and not the finish line. There are so many great people in technology who love to get to launch and lose interest. And you really rarely catch lightning in a bottle. You have to keep listening to customers and iterating. So you got to hire the right types of builders. And then I think you've got to organize them in teams that allow them to move quickly. I think one of the most important things we did in AWS, in the early days, was the natural place to build AWS would have been in our infrastructure team because we leverage our data centers and network. And instead, we pulled it completely out of that group of people that had to run the consumer business and that scale every day because the new businesses will always lose out to the existing surer bets. So making sure you organize them to separable or autonomous teams as you can, where they own their own destiny and aren't burdened by having to run the core business you're running every day. And then you got to give those teams the right building blocks to be able to move really quickly. So I think one of the interesting things about Amazon in the last 10 years is just how unbelievably quickly our consumer businesses have moved. And they would tell you a big piece of that is being able to use those AWS building blocks that all our external customers use. And then you need a culture that encourages innovation. And that sounds fairly obvious, like everybody wants to encourage new things. But I think it -- a lot of big companies, especially as you get bigger, the senior people walking into meetings where you're hearing a new idea are looking for ways to say no, not because they're ill-willed. It's just you get more conservative. It seems harder to imagine being successful on something new. It's hard to think about having one more team and effort that you have to manage. And I would say that the opposite is really true at Amazon. All the senior leaders will tell you that their favorite meetings are the ones where we listen to new ideas. And we don't say yes to everything, but we say yes to a lot more than most, and we spend most of our energy in those meetings trying to problem solve with the team to get to yes. And I think the last thing is, if you're going to invent, especially if you're in a significant existing business that's had some success, you have to be willing to tolerate failure. And if you hire great people and achievement-oriented people, it's sometimes a weird dichotomy. If I think about Amazon, there are a lot of Type A people who hate to fail. And yet if you're pushing the envelope, you're not going to get it right all the time. And I always found our launch of the phone very culturally reaffirming. Those of you who may not have known, we launched a phone and it didn't exactly work. And yet because we like the inputs -- the output was it didn't work, but the inputs were: the team we hired, the plan we built, the customer conversations we had, the technology that was built, the timing. And because we liked all the inputs, even though it didn't work, we took the learnings and the technology we could reuse and reapplied it and then gave people a good landing spot. Because if you don't, all the great people will fear working on the new things because they'll worry they won't have a spot if it doesn't work out.

Yes, that's a really good point. You talked about -- you said we were early. There's a lot more adoption to go. First inning? Second inning? I mean, how early are we?

I would say that maybe 2 outs, bottom of the first, runners onboard. Runs in on both sides, though. So I think that it's -- if you look at the total global IT spend, only 3% of it is in the cloud, but it's moving fast, and it's growing fast. And so I think that it's -- we have this very strong opinion, which we have really seen play out faster than we imagined. In the fullness of time, and I don't know if that 10 years from now or 15 years from now or 20 years from now, but very few companies will own their own data centers. And those that do will have much smaller footprints, only for workloads that just aren't worth moving because they're going to leave a mainframe to die or something that just has to be close to something, like a factory, latency-wise. And all of that is moving to the cloud. So it's very early days. And I think that people will make moves, and when they make those moves, they will choose that time to modernize a lot of the older software. Look, we have lots and lots of customers who are moving away from mainframes. And we have lots of customers who are moving away from the older guard commercial-grade databases to things like Aurora and -- lots of folks that are moving quickly from -- modernizing from Windows to Linux. And so you see all those types of moves. But you also see this huge explosion of SaaS providers who are building their products really organically right from the get-go on top of the cloud. So it's a very different world but still early days.

So you talked about some things that are obviously changing as we move from the bottom of the first with 2 men on and move towards the ninth. What else changes? What -- how does it look when we get closer to the ninth inning? Besides some of the changes you just mentioned, what else do we see?

Yes. I think, for whatever it's worth, I'm not sure that we'll get to the ninth inning in my lifetime working. I think this is -- we are in a period of unbelievably fast growth. And if you think about what's changed in the 14 years since AWS launched, and the pace of change has been really rapid, I think the next 14 years is going to be even more rapid. And so I think what it will look like is I think that you'll have -- most companies won't have data centers. I think that you'll see more and more compute that's done in a serverless model. I think a generation of developers will choose that as their programming models so they don't have to think about clusters or servers. I think you'll see a very different play with regard to what people use for databases, not just on the relational side where people are moving away from a lot of the more expensive and proprietary and punitive old guard providers. But also increasingly now, companies don't use one relational database to satisfy all their database needs. They use purpose-built databases. It's why you see people using things like key-value stores and in-cache database and graph databases and time series databases. So the database will be different. Almost every application will have machine learning and AI infused in it. There'll be billions, maybe trillions of devices that will live everywhere, in our offices, in our homes, in factories, in cars, in planes, in ships, in oilfields, in agricultural fields. And all those devices are going to be collecting data and sending them to the cloud and doing analytics on them and then sending smarter actions back to those devices so they can be more productive. Lots of tasks that humans do today, kind of mundane tasks will be done by robotics applications so humans focus more on higher-level value-added activities. I mean all kinds of ways it will be different. And we're also optimistic in the future that quantum computing will play a role as well.

So let's shift. I want to talk about artificial intelligence a little bit. And I know that it's a huge area of innovation that you're spending a lot of time on. And I think you have 10,000 people working on Alexa alone. Talk a little bit about kind of your vision for the direction that AI will play going forward? Not just in the context of something like Alexa and in the context of what we've been talking about but just more broadly, let's talk a little bit about the impact of AI.

Yes, I think that -- we've been working with AI and machine learning for over 20 years in Amazon, and you can kind of see it everywhere in our business. The recommendations you see on our retail site are fueled by machine learning. And if you look at the pick paths the people in our fulfillment centers use to go get products and box it is used by machine learning algorithms. If you look at Prime Air, our drones effort; if you look at Amazon Go, those are computer vision algorithms that drive it. If you look at Alexa, obviously. So we've been doing it for a long time, and it is deeply embedded in all of our businesses and all of our capabilities, and yet, we still have tons more ideas. Now today, there just aren't that many machine learning and AI expert practitioners out there, and a lot of them that exist live at kind of the big technology companies. And so we're trying to -- I think, collectively, more and more are getting educated, but it's part of why, in AWS, a lot of the services and the capabilities we're building there are to enable more companies to take advantage of that promise. And there is -- we see it as kind of being 3 macro layers that are stacked. At this bottom layer is for expert machine learning practitioners, of which there aren't that many in the world. But these are people who are comfortable building their own models and picking the algorithms and training the models and tuning the models and deploying them, and you have to keep running those models over and over again or they become stale. And we provide lots of ways to use all the popular frameworks, and we build chips to make it faster to do training and to do inference and predictions, things like that. But because there aren't that many expert machine learning practitioners, the most important thing that we feel like we have done to make it much more accessible to every enterprise is to build this middle layer of the stack, which we built a service called SageMaker there, which enables everyday developers and data scientists to much more easily build, train, tune and deploy machine learning algorithms in fractions of time than before. I mean, if you look at -- Intuit now trains their fraud models in less than a week when it used to take them 6 months. These are big differences in what you can do. And then that top layer of the stack are for people that don't want to have to build the models at all themselves. They actually just want to plug into a model that we've built and get answers back through application programming interface calls or API calls. And so these are things like, here is some text, turn it to speech. Here's some audio, transcribe it to text. Take this text and translate it into lots of languages. Take this transcribed translated text and tell me what's in it so I don't have to read it. Let me do natural language processing. Help me do internal search. If you have an intranet and you tried to do search on your intranet, it's pretty lousy at most companies. Can you use natural language understanding and natural language processing to change that? Help us -- give us your personalization algorithms. Give us your forecasting algorithms. All those are those services at the top layer that companies are really hungry to just plug into their models from companies that are doing it at scale and have that data. So I think most companies will operate at all 3 layers of that stack. I would argue that, that middle layer may be the single most important just because most enterprises have so much data that they want to use predictive algorithms to get value out of.

So the transition we're talking requires significant investment. And one of the things -- and I see it's obvious you go to build something new, we go build the Apple Card, we build it in the cloud. But when you're looking at legacy and you're thinking about upgrading and moving forward, there's a massive investment that's necessary. How do you talk about that investment to Boards of big companies, the founders of businesses, to people running big businesses? How do you talk about the investment and do it in a way that motivates people to make this investment quicker and move it along faster?

It's a really good question, and it's really relevant right now, just the stage of adoption we're in, in the enterprise and public sector, where I feel like we're at the early stages, the meet of that enterprise and public sector adoption. And I think that almost always the conversation starter in these discussions is cost. For most companies, if you can turn capital expense that you lay out for servers and data centers into a variable expense that you pay as you consume it, that's attractive. And then you get to turn that variable expenses into something lower than you can do on your own because we have such large scale that we pass on in the form of lower prices. We've lowered our prices on about 75 different occasions in the last 8 years largely in the competitive -- in the absence of competitive pressure just because the DNA inside of Amazon is to relentlessly work to take cost out and give it back to customers in the form of lower prices. So you turn CapEx to variable expense, so lower variable expense than what you could do on your own, and then you get real elasticity instead of having to decide how much infrastructure to provision and either provision too little and have an outage if you exceed it, which nobody does. So what everyone does is provision for the peak, but there's a reason they call it the peak and you sit on all those wasted capital on the cloud. You just provision what you need. If it turns out you need more, you pay -- you scale up seamlessly, and then when you don't need anymore, you give it back and stop paying for it. So cost is very different in the cloud than what's existed before, and that's compelling to virtually every company. But the #1 reason that enterprises and public sector organizations are moving to the cloud is agility and speed and the ability to innovate at a much more rapid clip. And unless you operate in a business segment that has no competition, which is very few people, you have to move quickly. And people are inventing at a rapid rate, not just start-ups but also your competitors who are adopting the cloud. And so those are the 2 most compelling reasons that boards and leadership teams want to move. And then I think there's -- if you think about how to really just get off the dime and get going, a lot of the challenge, it turns out, isn't technical. You spend a lot of time talking about the technical pieces, but it's pretty trackable. A lot of the challenges, frankly, are leadership. And so the first thing that really has to happen is the senior leadership team has to decide that they really want to move. They have to have that conviction because inertia is a really powerful thing inside organizations, and there are often people in the middle of organizations that aren't as excited about moving as some of the others because they maybe don't know what it means for their job or their scope or they were proud of what they built. And so you have to have senior leadership teams that are truly convicted about moving, and they set an aggressive top-down goal that forced the organization to move faster than they otherwise would. And my favorite example of this maybe is -- GE was making this decision that they really wanted to move to the cloud. And their CIO at the time, Jamie Miller, decided that she was going to move 30 -- 50 applications to AWS in 30 days. And she got her technology leaders together and said this is what we're doing. And she said, for 45 minutes, they all told her how dopey that was and how it wouldn't work and why it didn't make sense. And she said I just listened to them and then said, "I heard you, but we are going to do it. So let's go." And they got to about 42 applications in 30 days. But in the process, they learned their compliance model, their governance model, their security model, and they had some successes and the ideas started rolling in, such that she could set a second really aggressive goal top-down, which was to move about 7,000 applications to AWS in a few years. So you have to have that senior-level conviction and an aggressive top-down goal to force the org to move and that you mean business, or it's just really easy to dip your toe in the water for several years and wake up and realize your competitors are doing it and you're far away from it.

Well, good point. Talk about the competitive landscape. Talk about the competitive landscape in cloud computing.