 So I'm John Furrier with theCUBE here at Palo Alto on a remote interview for a special video interview at the EC2 15th birthday party celebration event, Raj Pai, who's the vice president of EC2 product management, AWS is here with me. Congratulations on Amazon web services, EC2, it's the compute, what a journey, 15 years old. Soon we had the keys to the car and a couple more years. So Raj, great to see you. You guys have been doing great work, congratulations. Thank you, it's great being here. It's super exciting for me too. I can't believe it's 15 years and we're still at the very beginning, as you know, that we often say. The building blocks that have been there from the beginning really set the table and it's just been fun to watch the innovation on behalf of customers that you guys have done at AWS and more. And for entrepreneurs and for developers, it just continues to be great and the edge is right on the corner, wavelength, all the great stuff. But let's talk about the specific topic here that I really want to drill into is that as you look at the 15th year and birthday for EC2, you're looking at the future as well. You're looking at the past, present and future and one of the things that's most compelling about recent re-invent was the Graviton performance numbers are amazing. You guys have been building custom silicon for a while. You also work with Intel and AMD. What is it about? What's the huge investment for you guys? Why do you start to see some returns? Are you seeing returns? And then why did AWS decide to build its own processors? Yeah, no, it's a really good question. And I mean, like with everything else we do in AWS, it's all about innovating on behalf of our customers. And one of the things our customers are telling us and they continue to tell us is they want to see, better performance at lower prices. And we've been able to deliver that with our hardware partners for the last 15 years. But as we've understood the workloads that run on EC2 and AWS, we saw an opportunity. What if we were going to go and design our own processor? That was really optimized for the sort of workload that customers run on the cloud. And make design decisions when designing the CPU and the system on a chip around the CPU that does things like bring a lot more core local cash and speeds up the parts of the operations that really benefit real world workloads. So, you know, this isn't about benchmarks, it's about like how do real world workloads perform and how do we build systems that optimize that performance? And with Graviton, we were able to hit the nail on the head. We were also very pleasantly surprised when we got our first chips off the line. And we're seeing that a customer like about 40% performance improvement at a significantly lower cost. And that's super exciting. And that's one of the reasons we're getting so much interest from our customers. I got to say as a geek and a tech nerd, I love the silicon development and there's benefits there and also the performance is there. The thing that also is pretty obvious that's happening and the world seeing it is the shift towards ARM based computing. What kinds of customers and use cases are you seeing adopt the Graviton? And what kind of workloads are they running on? Were the things that surprised you guys that you didn't surprise me? Andy Jackson was talking about the uptick and how everyone's leveraging it. What are some of the examples? Take us through some of the customers use cases, workloads. What's surprising you and what's going on with Graviton? Yeah, so I think the biggest surprise for us is how broadly applicable it's been. So one of the things we did, we launched with re-invent is as you know, like we have different form factors of compute, we have memory optimized instances that are good for databases and in memory caches. We have compute optimized for HPC and workloads that really take advantage of the performance of the chip. And then we have dental purpose workloads. And we introduced Graviton variants of all those instance families. And we're actually seeing the same sort of performance benefits across workloads. So, and it's one of the reasons why companies like Nexroll and Snap and Smugmug, they move one workload over, they see the performance benefit. And before you know it, they're starting to move workloads and mass over across kind of that spectrum. So I think that's one of the biggest surprises is that Graviton seems to do well across a wide range. And we're going to keep on introducing it more and more of our instance families because we've seen this uptake too well. You're seeing a lot of people move to the Graviton. You mentioned a few of those early adopters who are pushing the envelope. And they're always kind of trotted out there as examples at re-invent, which is always fun to see what they're working on next. What is the, and now is that people see the Graviton two instances, okay? The architecture difference, higher performance. How much effort do our customers typically need to move to Graviton two instances? And what are some of the benefits they're seeing on performance and price performance? Can you talk about that transition because that's natural evolution for them? Yeah, it's actually a lot less than they originally think. So some of the hardest effort is just getting them over the line to try it. So, one of the things that we tell our customers who are considering Graviton is, you just take one or two of your developers, take one workload and go off for a couple of weeks and just try porting to Graviton. And more often than not, they come back to us in like four or five days. They're like, it works. Now we have to do some testing and verification but we were able to port it because the operating system support was there. The ISP support was there and the tools that they use. And they're using, in most cases, modern programming languages, like Python or Go or Java or PHP, where the interpreter languages just run. And so it was very, very little lift in comparison to what people think it's going to be. And that's one of the reasons that, one of the big announcements we made in the last few weeks is what we're calling the Graviton Challenge, right? So it's a set of blueprints for customers to essentially have best practices on how to, in four days, take a piece of code, a piece of workload and execute and run it and migrate it to the Graviton. And we're seeing a lot of interest in that as people in the community realize how easy it actually is. What are some of the cool price performance things that are emerging? I mean, obviously it makes sense if you don't really need it, don't pay for it. But you have that option, a lot of people are going there. Is there a way of you see coming that Graviton 2 is going to be really set up for, that you kind of see some early signals coming in, Raj? Because, I mean, I can see the 64 bit, I can see where Graviton fits today. Obviously performance is key. Is it certain things that are emerging? What's the main problems that it solves? Well, I think anything that's a multi-threaded architecture is going to do really well on Graviton because of the, we have really densely packed 64 cores. And so you're going to see things like microservices and containers and workloads that are able to take advantage of that parallel execution do really, really well. And so, we say 40% performance improvement, but like when our customers have gone and tried this, they've seen upwards of 50% depending on the workload. So it's going to be more your multi-threaded there's some applications that may not be a fit. Like you give a legacy, for example, there's some software that hasn't yet been moved over and we're going to continue to invest super heavily in our whole ecosystem of hardware for the long term. So I think that if it's because there's a great option and we just encourage them to try it and they'll learn from their experience what works and what doesn't. Well, 15th birthday still growing up and it's starting to get more mature. You're the VP of product management. You have the keys to the kingdom. So you have wide-ranging responsibilities. Share with us if you can. I know that you really can't say much, but try to give a little bit of teaser. You got wavelength. I can see the dots connecting at the edge. You got outpost. So I see all that emerging. I can almost imagine that this is going to get stronger. What should people think about? Where's the headroom for EC2 with Graviton too? Yeah, I think you are connecting to the dots yourself, but our goal is to have AWS everywhere our customers are. And that means the full power of AWS. So I think you're going to see more and more of us having EC2 and compute capacity wherever customers need it. That could be in an outpost. That could be on their 5G network. That could be in a city right next to them. And you're going to see us continue to offer the variety, the selection of instances and platforms in all those locations as well. So I think the key for us is to be ubiquitous and have compute power everywhere our customers need it in the form factors they need it. That's awesome. Congratulations. I love the power. You can't go wrong with sending computers where the data is, where the customers are. AWS, Amazon Web Services building their own custom silicon with Graviton II processors. This is EC2 continuing to grow up. Raj Pai, Vice President of EC2, product manager. Thank you for coming on and sharing the update and congratulations on the 15th birthday to EC2. Yeah, thanks for having me. It's been great. Have a great Friday. All right, great. I'm Jeff Furrier with theCUBE. You're watching theCUBE coverage of EC2's 15th birthday event. Thanks for watching.