>>from around the globe. It's the Cube with digital coverage of AWS reinvent 2020 sponsored by Intel and AWS. Yeah. Welcome to the cubes. Virtual coverage of 80 was reinvent 2020. I'm John for your host. We are the Cube virtual not there in person, but we're doing remote, as is a W s. Although there there on stage live. And we're here with Dave Brown, Vice President of the Sea, to compute. Great to see you again. Great keynote last night, kicking off everything for the opening night. Great stuff. >>Yeah, well, John, it's always good to be on the Cuban. Thanks for having me back. >>You know, you're in the hot seat these days in the sense of there's so much going on. I mean, Andy, that could do a three week announcement. Keynote. It was like in three hours of nonstop you take a break to go The bathroom. You missed two announcements, right? So, so much going on. You opened up reinvent 2020 with your announcement ec2 of mac instances. And there was a ton of compute. And the theme was really you know, reinventing and reimagining compute both. I want to get into that. But let's start with the hard news. Tell me about the Mac instances. Um, you had a great use case there, That kind of illustrated in your talk. But where is this coming from? It's obviously Mac developers are big, but is this market something that you guys saw from customers or was a necessity? Take us through the thinking around the Mac instance. Easy to for Mac instances, um are going for >>absolutely absolutely So I mean me personally Matthews, a longtime Matthews that we've often thought about. Could we ever bring Mac OS to AWS? Right This thing we've spoken about on and off for many, many years and, you know, it was about a year and a half about two years ago. You know, we're always hearing new use cases from customers, and that's kind of what we're doing. So we're saying what a customer is trying to do that we don't support today, and how would we support them in that? And we started a year from customers that they have been able to successfully migrate all of the AWS workloads to AWS. So most of the server workloads to AWS and then they've got this Mac bold workload that they just weren't able to bring to us. We just didn't support Max into. It was a great example who I had on stage with me last night where you know, they over the last couple of years have been moving Ah, lot of their workloads to AWS. And and then they had these Mac money sitting around that they had to manage themselves. And so we said, could we actually do this? And so that was the one thing the customer ask. And the other thing that we realized was with the nitro system in the work that we've been doing there over the last, you know, six years, seven years since 2012, Really? And just where we are from the From nitro system point of view, we were able to wrap a Mac money without making any changes to it with nitro cards plugging a FireWire to the thunderbolt port and and and actually control that device. And so it means that you get the best of Apple hardware, which is what Apple's all about is the hard way that they make and the way that their software works with it. together with the nitric system and the cards around that inte integrating with the rest of AWS. So we're giving you, you know, high speed secure networking. We're giving you great access to elastic block store Was just integrates natively into the magma Nias? Well, a So we realized that the technology was there, the customer asked, was there and then obviously went to Apple and worked with them very closely to make it happen. And so that's kind of how it all came together. And I was incredibly excited to announce it last night. And the feedback today has just been amazing. A lot of excitement. >>Yeah, take me through the use case because, you know, obviously there's two trends going on. There's custom chips and server list kind of thing happening where you guys, I mean, really doing a good job of the eye as layer, innovating there and then platform as a service. All that software on top. I totally get that. You could see that happening. Chips custom ships to Intel, A, M, D. And others. Now you got Mac hardware. Where's the innovation use case because one would start would say, Hey, why don't you care about whether it's Mac hardware or not. Because I'm server lists. I should be programming the infrastructure actually be getting compute generically. Where does the Mac tying come in? Because that's the first question I was thinking of was, I'm a Mac user. I love Mac, but I'm also got some windows actually going on now. And ultimately, do I really care if it's compute? What's your reaction to that? Yeah, >>absolutely. I mean, if you look at Apple's ecosystem today, right, they have millions of applications in the APP store. They have 28 million developers worldwide, actually building those applications just incredible. And many of those applications, all these millions in the In the APP store itself, there's many more applications that are both by enterprises and companies, right? We have an application that we use internally at Amazon is available on my phone. That's not in the APP store, and you know, many companies are doing that and to build applications for the ecosystem, they have to be built on Mac hardware. And that's just how Apple works, right? So if you wanna build for iPad or iPhone or even Apple TV and Apple watch, you have to build those applications on a Mac. And so what we see companies doing is, you know, the old develop a meme off. Well, it works on my computer, right when you build something, you don't wanna be bullied on your local laptop for production. So they typically have a fleet of machines that they either under somebody's desk or in a data center somewhere that they use for for building these Mac applications. And so it's not possible to build a Mac application on anything other than a Mac itself. And we when we looked at it, we really didn't feel that virtualization made sense, right? Apple? I mean, they have some some virtualization that they're able to do within Mac OS itself. But if you think about how do we solve the customer use case, it's really bringing apple hardware too easy to to solve the problem and giving customers that exactly same exact same experience that they have on prep. And if you look into it like that, models just worked right. We gave them better access. Uh, you know, they've been using that data which you normally say, Hey, don't don't run production workloads on a beta. But you know, I found out if I interview with the BPS at Intuit critique that they've actually moved 80% of their production pulled wear clothes too easy to already to run on the Mac instances. And so that, and that's in the space of two months. And so, just as seamless ability to move because it's the same hardware is kind of what we were going >>after. Great, thanks for sharing that and say, one thing I wanna point out is Mac does have their own chips as well. They're going custom chips. Amazon's going custom chips. And I think I think you nailed what I was trying to understand, which is this developer community for Mac. And there's some things that are purpose built for Mac devices. So on Mac ecosystem, get the marketplace as well as you know, that that was the hardware PCs and devices, and they're only doing more and more. So this brings me to the i o t. Um, piece of it, because Apple does make devices that people wear and I watch is, um, iPhones. I mean, they're not computers anymore. They're everything. So this kind of brings up the edge conversation. So whether it's an iPhone or a five G in a Metro or I'm a stadium watching a football game and there's some sensor camera vision industrial thing there, this is the new normal. This is where you guys are kind of eating, eating up the software side that that business, because there's new capabilities here. Can you explain how compute he's, particularly C two gets to the edges because no one wants to move data around. They wanna move, compute, not data, because data is expensive and it's and it's fat. So we we talked about that we keep on years ago, but you gotta move. Compute. So how does that work Take us through your vision? >>Absolutely. And this is This is a massively growing area for us. I mean, you mentioned Apple's new M one silicon Apple silicon that they just launched a swell, and we're super excited about Apple's been doing there. We've been doing the same thing with our grab. It's on two processor and really saving customers. An incredible amount on price performance. Tried customers moving and getting 40% improvement and price performance just by moving to grab it on too. It's just incredible. Um, in terms of the edge, you know, we started this journey. We started this journey quite some time ago and bringing, you know, Lambda functions to cloudwatch and things like that. How do we bring compute to the edge? We took a look at five G, which I think it's gonna feel a lot of this right if if we look at our cell phones today was actually just talking to the Apple team yesterday with the iPhone, only came out, you know, 13 years ago. It's kind of amazing to think just how much progress we've had and what four g did for the device that's in our pocket in terms of, you know, just how much we rely on that today and what we get. Well, five g is just a step function in both in terms of latency, but also in terms of throughput. And so, you know, one of the projects we announced last year with Verizon and we now Andy announced this morning we're also gonna be rolling out with Katy D I and SK Telecom and Vodafone next year. Um is a project always like that brings aws compute to the edge of the telco network. And so with Verizon, we now have eight locations around the U. S. Where we have AWS compute capacity. And what I mean by that is literally C five instances uh, G four GPU instances for customers that want to do influence and graphics processing on the edge. And that's embedded into the five G network on DSO customers. You know, we've got a number of customers that are doing a lot of interesting things with five G in the sports area, where they have five G cameras that are, you know, submitted directly to wavelength. We no longer need to drive a truck to a stadium to record a game. You just have five G cameras, um, to, you know, automated factories where they doing robotics in factories and yet really low latency. And they don't want the computer, the factory they wanted in five G and so just exciting area for us. That's growing really, really quickly. Thea Other thing we did is obviously with local zones. We launched our first local zones in L a X last year, Los Angeles on that's being used by the movie industry, so you know right now is a lot of exciting up and running off the covert and shut down for a period of time and filming the next release of all of our favorite episodes and across all of these various streaming platforms. And a lot of that work is actually the post production is being done on on AWS on G four instances within the Los Angeles region. So, you know, very low agency for colorization animation, special effects, all that sort of things happening there. What we heard from a lot of customers was they loved outposts as well, which is our offering to put a server into a data center. And you heard from riot games in Andy's Keynote, where they actually bought a number of outposts and put them all over the U. S. And also other places of the world to really lower the Leighton see for their latest game. And so what Andy also just announced is the availability off three additional local zones. So Atlanta, Miami and Houston Sorry, Boston Miami in Houston available today, and then additional 12 available local zones next year, and what that does is that sort of spreads AWS capacity compute capacity at the edge in all of our major metropolitan hubs all of their capacities on the AWS backbone as well, but brings customers that low latency connectivity that they're looking for. Gaming developers were, you know, every every millisecond counts in terms of gameplay on so super excited to be going after that use case, which I think, you know, it's difficult to tell what the next 10 years is gonna be like. But I think Layton's he's gonna have a big part to play in the types of applications we see on our phones going forward. >>Great stuff, final question for you as we wrap up, obviously with virtualization with virtualization. But you know, the cove it is. And he pointed out, People are gonna change, is gonna be winners and losers. He kind of clearly pointed out, But the people who do lean into the cloud who have been on the cloud or taking advantage of the tail winds of cove in because of the capabilities there are two bills air higher, and you should be happy for that. But they're also gonna have more demand for you to say, Hey, I need more services. So How do you speak to those people who are leaning in who are leveraging, more, compute? What should they be looking at? What kinds of services should be connecting into compute? How should they be thinking about the future of compute so that they can take advantage of those capabilities? The lower costs, higher performance? What things are complementary for these customers as they come in, not toe dip in the water kind of things against really driving. And what do they need? >>Yeah, absolutely. And this has been a big focus on us. You know, things has bean, as I cover in my keynote, which leadership session that I'm doing tomorrow Wednesday. You know, a lot of this year has been helping customers through covert and what covert is meant for their business. Whether that is cost savings for many of them or whether it's just demand, you know that they've never experienced are expected before. I mean, we've been incredibly hard at work in servicing those customers, right? I actually catch up with Scott Sikora. In my keynote. He leads our capacity team. We talked through what it meant and how we actually provided the capacity that our customers needed during Colbert Times. But for a customer moving to us, the first thing is obviously we wanna find ways to make them very successful in the cloud, but more importantly, lower price performance for them. So what we wanted to do is give them the best possible performance that's available at the lowest possible cost. And if you look at a number of the announcements that Andy made today, you know whether it's our latest graviton processor where you can, you know, when you move to arm. I think customers often overestimate how much work it will be to move to arm. And when I talked to them after they have moved, that's ahead. Wasn't actually that much work. We actually got it up and running relatively quickly. So what's simpler than people expect? But that's an opportunity to save 40% on price performance. You know these new newer workloads like our graphics. We just launched a new G four a D, which is an AMG based GPU solution, the first time we have had an AMG GPU on the EEC too. And that's also looking to say, if you know upwards of 40% price performance of other GPU offering so just incredibly exciting for graphics, work, clothes and then in the machine learning space. Like I think, if you know, machine learning is just become the new normal, like everybody is doing it. And you know, just three years ago, everybody was thinking about whether they should do it. How would how they would use it Now that it's a lot of companies are doing it. It's really How do you How do I use it more? And that comes down to again saving costs. And so what we know with without Inferential Chip and then the new Habbaniya chip we just announced it with with the work with Intel that we're doing and then a new trainee, um, ship for training, training. We're really working to lower the cost of machine learning. And so, like we've seen many customers like Alexa was a great use case the other day. Being able to lower the cost of inference for Alexa by 35% again just helps customers, you know, move to the cloud. But I mean, just generally, you know, we're trying to support customers everywhere where there were, you know, if there are many customers are in their own data centers looking to move to AWS. You know, we have great models that can support them with our existing compute. A new savings plan offering we announced last year just great for saving costs on getting the price down So a lot. You can look at it. You know, I could go on forever. Really. It >>Certainly it's certainly is MAWR. We'll we'll do a deeper dive follow up after reinvent, but it is a wake up call. As I wrote in my post, um, for a cloud on Finally, I've been saying this for years. Horizontal scalability is a disruption on the infrastructure side, but you've got vertical specialization with data to create great modern apse of machine learning. And I actually playing out in full display here is Andy said, um, net right now. So all this benefits and all these opportunities to disrupt horizontally and then leverage the data all tied together, all coming together. You're clear. Leading the team. Great Brown, vice president of E C. Two in charge of the team that's driving the future. Compute. Thanks for coming on The Cube Cube Live coverage. Thanks. >>Thanks for having me. >>Okay. I'm John for the Q back for more live coverage after this short break