 Live from Las Vegas, it's theCUBE, covering AWS re-invent 2018. Brought to you by Amazon Web Services, Intel, and their ecosystem partners. Okay, welcome back everyone. Live here in Las Vegas, day three, last interview of the day, three days of wall-to-wall coverage, two sets here at AWS re-invent 2018. Our sixth year, we've been at every re-invent, except for the first one, and it's been great to watch the rise. I'm John Furrier with Dave Vellante. We're here with Andy Jassy, the CEO of AWS. Started this as a working backwards document years ago, 12 years ago, 12 and a half years. Seeing years ago was when the document was written and we launched it 12 and a half years ago. Great to see you, thanks for spending the time. I know you're super busy. Congratulations, we met last week. You couldn't really talk about it, but boy, it was so much more payload in the announcements than ever before. Are you happy with the results? Certainly, three-hour keynote was taxing. What's your impression? I was happy when the keynote was over, but no, we're thrilled with it, and most importantly, the reason we're thrilled is because our customers are thrilled. I think they just couldn't believe how much we delivered this week, well over 100 capabilities, and they were super excited about the storage announcements. The thing is, when you have millions of customers, any announcement you make is going to be popular with thousands of customers. So some people walked up to me and said, I know it's not sexy, but I love the storage announcements. I needed the file systems. I wanted that Glacier Deep Archive. Some customers love the database releases. We have lots of customers who are excited about the machine learning piece, and another unsexy one were the enterprise abstractions, just to make it so much easier for that type of builder who wants more prescriptive guidance to be able to get started quicker, and then people are pretty excited about Outpost too. So I'd ask you a question. I'll talk Amazon Speak now. What areas of the show do you feel you raised the bar this year on? What would you point to? Bar raising, moments, announcements. Well, I think each year, one of the things I like about re-invent and that we work hard on is we like to have, we don't really want it to be a corporate event. We want it to be quirky, and we want it to be authentic, and we want our community to have fun here while they also learn. So midnight madness is, for instance, something we do every, you know, we've done the last couple of years and we try radically different things. And so I thought the Tatanka eating contest raised the bar again this year. It was the second year in a row that we said again, it's Book of World Records. And, you know, I thought, I really liked Peter DeSantis' and Werner Vogel's keynotes on Monday and today respectively. I thought they both were fantastic and keep raising the bar every year. And, you know, I hope the party tonight too will be something that people feel like raised the bar year over year. I thought the house band's synchronicity was quite good too, you know? Yeah, I agree. That band is terrific. And, you know, and I think that, again, all those things I mentioned are part of what we, you know, think makes the event fun and quirky and different. But the most important thing by far is the learning of the education. And then our customer's excited about not just the platform, but we launch so many things. Do they feel like it helps them do their job better? Well, while we're on the raising bar, we got a prop here. This is the deep racer. Deep, deep, the deep racer machine learning. It's a toy for testing. And the question comes up, how old do you have to be to use this? And I said, hey, if your kid can code machine learning, good for them. But talk about this, because this is kind of interesting and because it's fun, but where'd this come from? You know, it came from last year when we released SageMaker and we were making machine learning so much easier for everyday developers and scientists. We said, what can we do to give people hands-on experience? Because you'll learn things better if you actually try it. And so we tried to help developers get more experience to computer vision by having a deep lens, you know, video camera and that was wildly popular. And so as we were thinking about this year, making reinforcement learning available as easily as we are in SageMaker, which we think is a huge potential game changer reinforcement learning, the team kept thinking about, it's great, but nobody knows about reinforcement learning and nobody has experience with it. How can we give them experience? What are ways they can get hands-on experience? And that's how the DeepRacer car came up, which is really making it simple where they can just give us a reward function with a line of Python script and then SageMaker will automatically train an RL algorithm and then they could deploy it to the car and then race against one another. When we watched how competitive it was getting inside our own house on these RL infused cars racing each other, we figured other people might find it compelling as well. And I couldn't believe how many people participated yesterday. And then I don't know if you saw it during right before Werner's keynote, the finals were really exciting too. Like the fact that there were some imperfections were actually made it more compelling to watch. And so it was- So you had a racer cup coming up too, I'm at 2019, competitive, that's going to happen. Yeah, today was the accelerated version of the first ever DeepRacer League championship cup, but next year will be a full season in our 28 WS summits. The top winners in the DeepRacer League races at each of those summits, plus the top 10 vote getters in points from those summits will come here and compete for the championship cup next year. You and I talked about a new persona last week when we met, but now the announcement's pretty clear. Now why? This points to a whole new persona developer. You got eSports on the Twitch side booming. eSports is changing the game in the whole digital sports category. Robotics space, you got a satellite announcement. This is a genre change in digital culture. And you see the AI stuff and machine learning. How does a web services stack play in this new world where AI is now a service? It's a whole nother paradigm shift. What's your thoughts on all that? Well, you know, I mean, all those areas that we're continuing to expand into are areas that our customers are asking us to help them with and where there are huge opportunities for customers, but where it's hard. I mean, if you look at space as an example, if you have to interact with a satellite, it's expensive to have to have all those satellites set up and those ground antennas set up and then you have to program them and then you actually have to pay this fixed price instead of on-demand. Customers said, why can't you give us access to those satellites the way we consume AWS? And then if you can have the ground antennas where when the data comes down from the satellite, it's basically on the same premises as your AWS region so we can store the data and process the data, analyze it and take action, that is very compelling. So that just felt like a natural fit. And the same thing with robotics. I think that robotics is one of the most underrated areas in technology. I think robots will do all kinds of things for us at work and in the home. And the tools out there to make it easy to build robotic applications and to do the simulation and to deploy them and then have them work with the rest of your applications and infrastructure have been pretty primitive. And so RoboMaker is a big- I agree with you. I think you look at the younger generation too, even at the high school elementary level, people are gravitating towards robotics. Robotics clubs are booming, that maker culture goes to a whole other level with robotics, congratulations. You know, it's funny, we had the youngest person to ever pass the AWS certification exam is a kid named Karthik, nine years old, he was here this week, actually. And I got a chance to meet with him today and I said, well, after the certification, what are you doing? He said, well, I'm building a robot. And I'm building a robot. And he said, now with your launch of DeepRacer, I want to try and find a way to have the DeepRacer car be the eyes and the camera and the reinforcement learning for my robot, nine years old. So it's going to be a different generation with what they build. John and I were talking this morning, Andy, at our Open, about you're making it harder for the critics. You used to be self-service only, criticize your open source contributions, the hybrid strategy, you're kind of ticking the boxes on all of those. Outpost was, I think, surprised a lot of people. It didn't so much surprise us that you were moving in that direction, but I wonder if you could sort of talk about some of those key initiatives. I know it's customer-driven, but wow, the TAM expansion, the customer value that you're bringing, it's like a whole new era that you're entering. Yeah, you know, everything we build, as you guys know, we talk about all the time, it's just, it's driven by what customers want. And so we just started over the last six months, you know, and really by virtue of having this partnership with VMware, where we have a lot of enterprise engagement as they're moving to the cloud, using VMware cloud and AWS, we had a bunch of customers say, it's really great. I'm moving most of my applications to the cloud, but there's some that aren't moving for a while because they got to be close to selling on-premises, and I want to use AWS for this. I don't want a different environment. Can you just find a way to put some services, like compute and storage, on-premises and hardware, but I want to actually use the same control plan I'm going to use for the rest of AWS, and I want it to easily connect with the rest of my applications in AWS. And we didn't like, as you and I talked about a week or two ago, we just have not liked the model that's been out there so far to do this because it's, you know, the control plan is different, the APIs are different, the tools are different, the hardware is different, the functionality is different, and customers don't like it, it's why it's not getting much traction, and we didn't want to pursue it if we didn't think it was going to be useful, but we had this concept we were working on with a couple of customers where they wanted compute and storage on-premises, but they wanted to have that connect with all the other applications in the AWS cloud, and so we had this idea that maybe this local set of compute and storage would be like a far zone from an availability zone they were using, and we started thinking about that and we thought there was a much more generalized idea which became outposts, and so the thing that I think people are going to love about that is for the applications that can't move easily because they need to be coastline on-premises, you get AWS, like real AWS compute, real AWS storage, analytics, database, SageMaker will be in there as well, but it's the same APIs, the same control plan, the same tools, the same hardware we use in our data centers, and it'll easily connect through the same control plane to the rest of AWS, the rest of the services and the rest of their applications there, so. And it provides a platform for a whole host of new services down the road. I mean, every customer meeting I've had in the last, you know, since we made the announcement, people are excited about it. I wanted to ask you guys, you're talking about all the innovation and new areas, and we're seeing an expansion of the AWS distinct brand in things like TV advertising, StatCast, I wonder what's behind that, can you address that? Yeah, it's a good question. I mean, there's kind of two different types of, I'll call it TV advertising of sorts we're doing. One is straight up advertising, one is less so, which is, you know, the one that's less so is that a number of the sports leagues are really interested and actually pretty sophisticated in using cloud computing and analytics and machine learning. If you look at Major League Baseball and now NFL and Formula One, and they want to make the user experience and the viewer experience so much better, and so they're building on top of AWS, and then we like the ability of helping them showcase the capabilities that they're, you know, both the customer experience and the ML and AI capabilities. Then there's just the straight up advertising that we've been trying. We tried a little bit of it last Q4, and, you know, it's always very difficult to quantitatively measure TV, but we have a lot of ways that we try to triangulate that, and we were really surprised at what looked like the positive numbers we saw for both TV as well as the outdoor media and things like in the airports and things like that, and so we decided we would try it again this Q4, and, you know, I think, I would call us right now still experimenting, and it's very much kind of what Amazon does, which is we try different things and see what resonates and see what the field says. So far, so good, and we expect to keep experimenting. I think that's a good call because the brand lift is probably there. Obviously impressions get reached vehicle, but you guys are in a rising tide market. We're hearing co-creation, VMware co-creating, deep, meaningful partnerships. You always talk about that. So it's kind of this success model of innovation to reimagine the satellite, Lockheed Martin, a partnership. This seems to be a new way to do business in this rising tide. How are you guys getting the word out, education? People want to know more. This is a big kind of movement. Yeah, well, you know, I think that if you looked at the first several years of AWS, I was always surprised when I would go see enterprises and they would have no idea that Amazon was doing anything in the cloud, even though we had the only cloud offering at the time. So I think if you compare where we were a few years ago to today, there's gigantic awareness relatively speaking, but I still think that there are so many, the vast majority of workloads still live on premises. I mean, we have a $27 billion revenue run right business. It's growing 46% year over year. And yet we're still at the early stages of the meat of enterprise and public sector adoption in the US, you go outside the US where they're 12 to 36 months behind, depending on the country and the industry. And sometimes it feels like, you know, like Groundhog's Day. But you guys are doing regions out there, Italy was announced, you're expanding very fast globally. Can you talk about that real quick? Yeah, it's a, you know, we've had customers from 192 countries using AWS for many, many years, but they've been using AWS in regions outside of their country usually, because there are a lot of workloads that could stand that latency and where the data doesn't have to be on natural soil. But increasingly, if you want to help customers get done what they want to and serve the broader array of their applications, you have to have regions in their country, both so that they have lower latency to their end users and because of the data sovereignty laws, which are getting really more rigid rather than more flexible. Let me ask you a question about competition. You said, I can't remember some of the cube we're in person, there's no compression algorithm for experience and time. Classic economies of scale. When you have copycat, people trying to copy Amazon, how do you talk about some of those things that are the diseconomies of scale? What are the points that customers should look at when they say, okay, I got someone else was talking cloud, Amazon's got years of experience ahead of the competition, more services. What do you talk about? What do you point to? It's not about slimming the competition, but what is the diseconomy of scale to try to match the trajectory of Amazon? Yeah, it's a bunch of things. First of all, it's operational performance. A lot of the hardest lessons you learn in operating a scale only happen when you get to that level of scale. And there's some events that we see sometimes elsewhere where we look at that and then we read the post-mortem and we say, oh yeah, 2011. We remember that. We went through that. And I don't wish it on anybody, but when you have a business that's several times larger than the next four providers combined, you've just set a different level of scale and you've learned lessons earlier. I also think that the reason that we continue to have both so much more functionality and innovate at a faster clip and seem to get capabilities that customers want is because we have so many more customers than anybody else. You know, a lot of times, and this has happened all week too, where customers will say to me, I can't believe that you knew that I wanted that. And I always say it's because you told us. It's not like we're noses or namas, you've told us that. And so when you have so many more customers and when they feel free to give you feedback and when you've built good mechanisms like we have to get that feedback from the field to the product builders, it means there's this real flywheel of getting, you know, getting more customers leads to more feedback, leads to more features, leads to better functionality, where there's a network effect from being on the platform with all those other customers and all those industries. I wonder if you could add some color to a premise that we've put forth on your edge strategy. So as you know, we do a lot of these shows and a lot of the IoT and edge strategies that we've seen from traditional IT players, what you call the old guard, have fallen flat in our opinion because it's a top-down approach. It reminds us of the Windows phone. It just didn't work and it's not going to work because they're operations technologies people. We see what you've announced here as a bottom's up approach. You're developing an application platform to build, secure, and manage apps for those folks at the edge. I wonder if you could add some color to that and some thoughts on your edge strategy. Yeah, I mean, again, for us, we don't have some top-down strategy that I think is grandiose. It's just what customers want. And so we have so many customers who have all these devices at the edge and all these assets at the edge and they say to us, well, the first problem I want to get this data into the cloud and then I want to do analytics on it. We say, okay, well, how can we help? They say, well, the first thing is, I don't even know how to translate this data from the device protocol to just being able to operate in the cloud. So that's the first problem we go solve. And then people say, okay, now I can get it in, but I actually, I need security. Like, if you look at the amount of security options for these edge devices, it's a new field. That Dine attack that took a lot of the internet down a couple of years ago came from a device on the edge. And so that's why we built a security capability and people say, well, okay, now you've made it so I can run devices, but if I'm going to run thousands of devices, I need a way to manage all those devices at scale. And we built something to manage the devices. And people say, well, okay, it's great that I can do it in devices big enough that have a CPU, but what about when they don't have a CPU? They have just a microcontroller and that's why we built the RTOS piece. And the list kind of keeps going. People say, this is great. Now that I get all this data in the cloud, I can take all these analytics actions, but on my device sometimes I don't want to make the round trip to the cloud. So can you give me a way to use the same programming model and pick which triggers I want to take action in the cloud versus those that want to take on the device itself, which was what Greengrass was. So all of those pieces is not some kind of top down master plan as much as we know that customers have all these devices at the edge that they want to use that data, analyze that data, take action on that data and send it back in multiple ways. And you have the cloud platform to give them the services to make the tools, the right tools for the right job. That's the main theme. So I got to ask you about one of the big controversies that we don't think that's that controversial, but the chips that you announced, new Amazon web services, microprocessors, the chips. The two of them, talk about them, and Intel's also a partner. A lot of people were talking about this in the press. Intel, Amazon, chips. Well, that Annapurna acquisition is starting to bear fruit. It was 2015, I think, really. It really, the Annapurna team is fantastic, and they've added a huge amount of value to AWS and Amazon as a whole. You know, the first thing I would say is that Intel is a very deep partner of AWS and will be for a long time. I mean, that's not changing. And we've long thought that there were going to be lots of different processors out there and different ones that did different things with different price points. And so, like a lot of other companies, we've been interested in ARM for a long time and for a while it wasn't mature enough and the technology's matured. And we found a way in building our own ARM chip with Graviton, where we think we can allow customers to run a lot of their scale out generalized workloads, but up to 45% less expensively. And so, when you find a value proposition that compelling for customers, you need to do it. And you know, as I mentioned in the keynote yesterday when we were talking about inference, we feel like a lot of the world has been solving for training and not solving as much for inference yet. And we've made training so much easier with the things that we've built in AWS over the last couple of years, but inference is where most of the cost is going to be. And so, elastic inference, we think, you know, will allow people to be much more efficient in how they use GPUs and how they spend money. But when you've got the type of workloads at scale and productions that use whole GPUs or that need that low latency where you need it on the hardware, a chip that's optimized for inference that is faster, that's more cost effective, that's high throughput, where you can get hundreds of tops on it and thousands of you band them together is going to totally change the game for inference. And so, that was something that wasn't easy for us to find elsewhere. And when we have a team, fortunately, they could build it. And it's the combination of the elastic service of inference with the chip that makes the difference. It's both, yeah. And so, it's a specialism there. So, it's not like a... I mean, you can use each on their own and we expect there'll be a bunch of customers who will use each on their own, but there will be an opportunity to use those in combination that will be very powerful. But it comes down to really deeply understanding the customer problem again. The training versus inference, and everybody talks about the training. But the technical challenge, you guys realize there's the inference. Intel's going to make a lot of money and the time's expanding, market's expanding. So, they'll get their share. You know that the chips get taped out. They're kind of a couple year, two, three year life cycles and everything starts anew every time somebody's building a new chip. So, I think it's actually great for customers of all sorts that there's multiple processors that are possible, but we will have a deep relationship with Intel forever, I think. So, I want to talk about one of the cool demos you did on stage. No, you did. Customer did. F1. That was super cool. I love that imagery because it's an analogy of high performance, competitive, racing. That can be applied to business, play, sports, anything. And the level of accuracy that they need in the real time time series kind of encapsulates a lot of the cloud value. Talk about the F1 analytic thing. Are you guys going to sponsor these events? Is there a relationship there? Give us the picture of what's going on there. We have a deep relationship with Formula One where they're using our platform to do all their digital properties as well as their analytics and machine learning. And it was super cool to see Ross demo the way that they're changing the user experience for viewers. It's an amazing sport. It's not watched as much maybe in the US, but outside the US, that is the motor sport. And the way that they're changing the experience, the way that they're able to assess what's happening with drivers and with cars and then predict what's actually happening and make the viewer feel like they're actually either in the cockpit or actually in the pit itself with the crew is, it's really exciting and it's an honor to be a partner. So, you have to do some events there. We've got to get the cube there. These big time events. There's a tech angle now on everything. That's a plug for you to be at the F1 event. Cloud democratizing, you're hitting now new industries. I mean, this is the thing, right? I mean, it's disrupting every industry. I mean, what aren't you disrupting? I mean, what areas do you see that yet aren't coming online to the cloud? I don't see industry segments at this point that aren't moving to the cloud. I would have told you 18 to 24 months ago that I felt like financial services was moving a lot more slowly than I thought they should. You know, probably healthcare also was a little bit slower, but both of those industry segments are moving very aggressively to the cloud. Well, it's taking longer. They're high-risk industries and the digital transformation hasn't occurred fast enough, but it's coming. And there's regulatory pieces that they legitimately have to sort through. And, you know, we have just, if you look at financial services as an example, we have a pretty significant team that does nothing but work with our partners to help them with the regulatory bodies. Because what we find is, when we go with a customer to a regulator and show them a real use case and then how it will be done in AWS, the regulator says, oh, well that's more secure than you do on premises. And so it's just an education process. And, you know, I think that's been helpful and helpful. Okay, final question for you. What have you observed here at ReInvent? Obviously, a lot of people talk until you get a lot of feedback. Actually, two-part question because I always ask the final, final question. So I'll just get it out front. What are people missing? Of all the announcements you've had, a lot of signal in there, a lot of announcements, what is something that you've observed that you think should be amplified that people might have not overlooked but you feel like it's more important to shine the light on? We'll start with that one. Well, you know, it's a little hard for me to tell this moment just because there have been so many it's a short amount of time. And if we just look a little bit at the coverage, it seems. And if I take just as inputs, the comments and the questions from customers, it's been pretty broadly understood and people are pretty excited. And as I said, different segments have kind of their favorite areas. But I feel like people are pretty excited by the breadth of capabilities. You know, I think that if I picked two in particular, I would say that people are still in the machine learning space. People are blown away by how much we provided all three layers of the stack. I think people are still getting their heads around which layer of the stack am I going to participate at? You know, I mean, the one that probably has the most potential for most companies is that middle layer because most companies now have gobs of data and there are jewels in that data. And if you can enable their developers, their everyday developers to be able to build models and get at the predictive value and add value, that has huge impact for companies moving forward. But most modern companies with technology functions will use all three layers of the stack. So just getting their arms around which layers of the stack they should take advantage of first and having the personnel to be able to do it. And we're making that much easier with things like SageMaker. And then, you know, I think if you look at the blockchain space, I think that that is just one of those spaces that has a huge amount of buzz. People talk a lot about, they're not exactly sure sometimes what they're going to do. But I also think that a lot of people said to us that breaking those into those two real customer jobs to be done and then having a great solution that does each of those jobs really well is not only something that AWS does all the time that makes it easier for them, but it also made it easier for a lot of them to understand, like a lot of customers said to us, you know, that QLDB, that ledger database with a single trust of central authority for my supply chain, that's what I need for my supply chain. I don't need all the complexity of a blockchain framework. And then there were a lot of other people who said, oh yeah, that is what I wanted. I wanted the decentralized trust between peers, but I just needed a way easier way to manage hyper ledger fabric and Ethereum. And so I think those are two that people like are so interested in still figuring out how to use as expansively as I think they hope they will. Andy, thanks so much for your time. And I want to just say watching you guys for the past six years has been a fun journey together, but watching the execution, you guys have done an amazing job of keeping your eye on the ball and being humble, but being proud and loud at the same time. So congratulations and guns blaring in 2019, what's your top priority? Besides listening to the customers, what's your top priority in 2019? We know you're listening to customers. Oh my gosh, we have so many things that we're doing in 2019, but you know, we have a lot of delivery in front of us. I mean, as much as we launched 140 unique things over the last six to eight business days, and yet I tell you to stay tuned the rest of 2018, we have more coming. And then in 19, you should expect to see more compute capabilities, more database capabilities, more machine learning capabilities, more analytics capabilities. Look at a lot, I could spend all night, John. We're going to need to post reinvent, post traumatic announcement syndrome because just to digest it all is a lot of work. I'm also looking forward to seeing how enterprises continue to kind of manage their hybrid approach as they're making this transition from on-premises to the cloud. How many continue to jump on to VMware Cloud and AWS? How many jump on to Outpost? I think that that transition and helping customers do that easily is something I'm looking forward to. It'll be a fun year. Of course, we'll be commentating and pontificating on that for the next year. Thanks for your time. I appreciate you having me. And I appreciate that you guys come every year too. Thanks Andy. It's a really part of that. Hey, our pleasure. Okay, winding down, that's the last interview here. The wall-to-wall covers two sets, 110 interviews in the books. We'll have 500 video assets total. Blog post on SiliconANGLE.com. That's reinvent closing down 2018. Thanks for watching.