 So we're here with Ampere here at Mobile World Congress and hi, so who are you? Hey, I'm Matt Taylor from Ampere Computing. I manage our sales and business development team here at Ampere. And this is the chip right here? Yeah, this is our first generation EMAG processor. So EMAG is our family name for the Ampere server roadmap. This is actually our product that we launched late last year. It's a 32 core ARM V8 processor. It's our first generation as Ampere. We have multiple products on our roadmap now. And we'll be starting to sample our next generation product later this year. And hi, so who are you? My name is Scott Hara. I'm from Ampere. And what we're demoing here is an edge compute solution where we combine NGIC, the networking stack along with big core problems like in this particular case we're running a Cassandra database to run data analytics. This is one example of many. We also have deployments where we're showing off gaming in the cloud, machine learning applications in the cloud as well as other data analytics frameworks. So what are we standing next to right here? Yes, this is actually a vapor IO chamber which is actually a partner of ours where they're actually building these, think of these as edge data centers that they're deploying in multiple locations around the world. What we're showing here is actually one of the racks of servers that they have in one of these chambers with the Ampere server. This is actually a Lenovo server that's actually production today. So this is a Lenovo HR350A. So this is a Lenovo production server that you can buy from Lenovo. High volume production, shipping to multiple customers now and all sorts of CEOs around the world. Lenovo is one of the biggest manufacturers of this kind of stuff? Yeah, they're top five OEM in the world. They have a very large presence in basically every geography in the world. We've actually launched two servers with them, both a 1U and a 2U version of this. We launched those back in November of 2018. And how's the performance? Yeah, from a performance perspective, really our value prop is around performance per watt and performance per dollar. So performance-wise we're about equivalent to a Xeon 6130-ish. So basically high end of the gold family of the Xeon roadmap. We're similar performance, but we're a fraction of the price. We're somewhere between half and a third of the price for the same performance that you get with the Xeon gold. Half and a third of the price. Correct, yeah. And this is Mag1. Yes, correct. But it's building a history of applied micros designs over the years, right? Yeah, so the history of actually Ampere is we bought the assets of applied micros Xeon product. So they had done actually two generations of ARM servers, 64-bit ARM servers actually technically. We actually bought the company. We actually took the design they had in-flight which was the Xeon 3 design and brought that to market, made a few enhancements to the design and took that from basically what was a database to production in under a year. That's our first generation product. That's the Emag product that you're seeing here. Database? Yeah, so target workloads for us on first generation product are really web scale. Think of things like WordPress, Drupal, IngenX, things like that. Database, one of the unique things about Emag, our first generation product, is it actually supports eight channels of memory. So each socket has eight channels of memory. You can do eight or 16 dims per platform. That's a lot of RAM. It's a lot of RAM. So for things like in-memory database, for example, our performance is phenomenal. And especially for the price performance that you're getting out of that, it's leadership actually in the industry. So if you look at things like Cassandra, which is what we're showing off here, if you look at things like Redis, we have great performance and amazing performance per dollar on those sort of workloads. So how much of the cloud needs this? Well, if you look at cloud workloads, for the most part, they're very much in line with what we're targeting. Web scale, scale out sort of applications, database applications, storage. And so we think that we can address a large, large portion of Web scale workloads today and increasingly going forward as our next generation products come out and have even higher levels of performance, higher levels of features that allow us to address even more and more workloads in the cloud surface profile. So you did talk about NextGen also? All we're saying right now is we'll start sampling our NextGen processor later this year. We're not talking about performance or anything like that. We have said that it will be actually both one socket and two socket capable. So we'll have a multitude of designs and a multitude of OEM and ODM partners that we'll actually launch systems with when we bring this to market. The vision is to part a whole cloud, right? Yeah, there's a lot of opportunity right here. Yeah, our focus is really around two markets. So it's really around cloud and edge. These are our two main markets that we're building products for. And one of the things that's unique about Ampere is that we're not trying to build a product for everything and everybody. We are very focused on billowing solutions that address scale out workloads for the cloud and edge. And that's really what you're going to see us focus on from a customer and also a design perspective. But cloud and edge is everything. Yeah, I mean if you look forward, today cloud's probably half of the server sales that are done. If you fast forward three, four years, we think that cloud and edge will be probably 70 to 75% of the total servers in the market. So we're focused on really where the growth is and also where people need high throughput, highly scalable solutions that deliver high performance at very efficient performance per watt and performance per dollar designs. So how's the traction right now for this chip? How's it going with the industry looking at it and thinking about using it? So we have lots of POCs and deployments in flight. I think momentum is building. We're in a time where we now have available performant ARM based servers and I think the software ecosystem is slowly catching up. So it's going really well. And it is also exciting to see what's called what Amazon is doing, right? Because it could be also an example of what you're just about to be maybe doing too. Yeah, I mean I think the... Just providing instances people can just use? Yeah, I think the Amazon announcements made it actually great for us. I think from an ecosystem perspective it's done two things. One, it's made ARM server real and it's something that now all the cloud providers are looking at and saying how do I do something better even than what Amazon is doing? Which is great for the ARM server ecosystem. The other thing that it's done though is really started to accelerate software ecosystem. So one of the things you've seen is there's a lot of software that's available on ARM today but people haven't optimized as much as they should and what you're seeing now is a lot of optimizations happening. A lot of people coming and saying how do I make my workload run as well as it does on x86 on ARM. And that's really helping everybody. It's helping lift all boats. And it's creating frankly a lot of interest in alternative architectures and making it easier for people to go and deploy. So I'm guessing in your office in the Silicon Valley, right? The headquarters. There's a lot of very talented chip designers over there. Like you have a very strong team. Yeah, one of the things that's unique about Ampere is we like to say we brought together the best of multiple worlds from the silicon space. So we bought the xGene team, the applied micro team. Several years of experience. It was roughly about 200, 250 people when we bought it. Since then we've actually almost doubled the size of the company. We've built a large team in Portland. We've actually hired a lot of people from Intel there. So our chief architect was actually chief architect for Xeon at Intel before joining us. We actually hired a number of Intel fellows now on both the silicon design side as well as the software side. We've hired a number of chip architects from Intel. And the CEO is from Intel? Yeah, Renee James, our CEO. She was former president of Intel. Before that she ran the software and services group at Intel. So understand what it takes to build and develop a software ecosystem around silicon. And that's really part of our strategy too is it's not just about building silicon. It's about building platforms and an ecosystem to make this easy for customers to deploy. And you also have some expertise from other places, right? Yeah, that's the other thing I was going to say. We've also been lucky enough to hire probably about 75, 80 people now from Qualcomm. So the Qualcomm data center team, unfortunately, they've divested of their data center assets. We've been lucky enough to pick up a number of people from that team. And now we have probably close to 100 people in our Raleigh location as well. So that means basically you're picking up where they left off kind of? Yeah, for the most part. We've followed a similar strategy as where Qualcomm was. Qualcomm is very focused on cloud and edge and also on the telco carrier sort of space. We're very much following the same sort of strategy which is building scale out compute solutions that are highly efficient and deliver great performance per watt and performance per doll. So how's the ecosystem in the ARM server industry? Is it a great place to customize, optimize, do something different? How do you differentiate, let's say, from the AWS, whatever they do? And then there's all these other companies that want to do stuff like that, right? They want to have ARM instances potentially. They could be Google, Microsoft. Maybe you want to take it a different way. So at this point, ARM has become a first-class citizen in most of the Linux Foundation projects. So from an OS distribution standpoint, it just works. Linux Foundation projects, most of them just work. So from a software ecosystem perspective, we're getting to the point where this is becoming very boring. So I think software is a key to all of this. This is something that Renee James understood as a leader of SSG in her previous life. She became president of Intel. She understands the software ecosystem really well. We've hired a great number of very experienced people in the open source community. So you'll see a lot of contributions in the software ecosystem from AMP here going forward. Yeah, our strategy from a software perspective is to upstream everything. We're not trying to build proprietary software stacks. Everything we do from compiler on up into open source communities is all open source and available to the public. And that's really our strategy, and it's about basically advancing the overall ecosystem. And to make this work great, you need to have a very strong software team. Not just some cool chip design. Yeah, part of the philosophy of the company is one thing that Renee has enforced is part of our ethos is it's not just about building silicon. Silicon is important but insufficient to actually be successful in this market. We're really focused on building platforms and building a software ecosystem to make this easy for customers or what. So one of the things also is that there's eight billion people. Everybody wants to have these cloud services more and more. There's more and more demand for this. I'm guessing the industry is building so fast that things can really ramp up very quickly with this. Yeah, it all comes down to a couple of different things. One, having highly available platforms and having highly available partners that can actually go and deploy them. So we have Lenovo at this point. We have a couple other partners that are building systems. But basically having availability of systems for customers to deploy that are truly data center class servers. This is a server from Lenovo that they sell at high volume around the world. That's number one. Number two is having the software ecosystem there that people can just go deploy. As Scott said, software is actually becoming kind of boring now. And that you can just actually go and run all of this open source software on ARM. And that's really part of our goal is not to actually make software hard to just make it easy and boring. So you can just go and deploy it and an ARM server just looks like an x86 server. It looks and acts just the same. That's really what we're trying to get into. Do you work with the Red Hat? Yeah, Red Hat's a partner of us. We've shown Red Hat running on our platform. We've worked with them not just on OS, but also on OpenStack and a number of other projects as well. Do you have any kind of Windows server stuff yet? Yeah, so Windows, we've shown Windows running publicly on our server. They've talked about actually their plans, actually run Windows server on ARM platforms in their data centers. Right now, Windows server is only available for Microsoft's internal consumption. That might change in the future. But as of today, it's a Microsoft internal consumption for their own cloud services only availability today. Can we tape a discussion about the software team? Yeah. Do we have 100 people or do we have 200 people? It's about 100 people. Right, right. So I think the important, so software is really important. About a quarter of our company is dedicated to contributions to the open source community. So our strategy is to enable optimized software in the open source. Compiler work, library work, framework optimization. We'll all be pushed upstream. And you were saying that this is already a bunch of percent better than some Intel products. So why is it not knowing that it's shipping? Yeah. So at which point does it reach some kind of benchmark or something? Where it's just a no-brainer for everybody to just adopt it? Yeah, I mean I think two things. One is we just launched our product in November of last year. So we've only been shipping now for three or four months. We are seeing tremendous ramp. We're seeing a ton of interest, a ton of customers coming and wanting to go and deploy. The reality is it takes a while for customers to onboard new technologies. And what we're seeing right now is large scale POCs with a lot of these very large cloud providers where they're actually working, getting all of their software not just running but optimized. And that's really the stage that we're in is go and get all of your hardware deployed, get your software running, and now start to optimize your software so that you can actually get performance that's at or above which you can get on your Intel platforms today. Do you talk about how it compares with the Marvel solution? We haven't really talked publicly around that but you can look at, this is actually public data at this point, you can look, we're similar performance as to where the Marvel Thunder X2 product is at a significantly lower price and a significantly lower power. Oh really? Yeah. So actually that's a performance level people are very happy with, right? Yeah. And so they're performing well, we actually are very much encouraging our other ARM partners to bring different solutions to market. One of the things you see is, Marvel for example is very focused on the HPC market which we think is great. It actually helps the overall ARM ecosystem develop and get meaningful deployment in different segments of the market. And so while they're focused on HPC, that's great. We're going to be very focused on cloud and edge. Nice. That's really awesome. Is this like a working demo that works? Like it's running? Yeah, yeah. Talk about the demo actually a little bit more, actually talk about what's running. The demo is actually an example of how software is just working, right? This is NGIC demonstration. So this combines DPDK, Kubernetes, OpenStack on a standard Linux distribution and it just worked. We got this demo up and running within weeks. It was reasonably simple. So it was just running right now? Or it could be just click and run? Right. Oh. Yeah. It was running. Yeah. There it is. Cool. So really looking forward to what you can come up with. An emperor? Is it absolutely electricity? Yeah. It's the inspiration for the name of the company. You know, it's the history, but there's not really a large story behind it. All right. Cool. Looking forward to that. Yeah, sounds good. Thank you very much.