Welcome to today's webinar, celebrating the first anniversary of our high-performance EPYC processors for data center and server. This is Laura Graves, Corporate Vice President of Investor Relations, and we're excited to have you with us today. Today's webinar will include a presentation and a Q&A session. We ask that you submit your questions for our speaker today through the private chat window on the webinar platform. This webinar will be available via replay on the investor relations website at www.amd.com or ir.amd.com promptly after the webinar ends. During today's webinar, we may make forward-looking statements about AMD technology and our data center business. These statements involve risks and uncertainties and are based on current beliefs and expectations. Actual results could differ materially.
Investors are urged to review the risks and uncertainties detailed in AMD's filings with the SEC, including our most recently filed Form 10-Q for the quarter ended March 30, 2018. I am pleased to welcome Forrest Norrod, Senior Vice President and General Manager of AMD's Data Center and Embedded Solutions group. Forrest is a veteran of the data center industry with more than 25 years of technology industry experience across a number of engineering and business management roles, most recently as General Manager of Dell's server business prior to joining AMD in 2015. Since that time, Forrest has been responsible for data center technology development and strategic go-to-market planning at AMD. Welcome, Forrest, and congratulations on the one-year anniversary of EPYC.
Thank you, Laura. It's great to be here today. In June of last year, a year ago, AMD took a bold step to re-enter the data center server market with the launch of EPYC. We're driving innovation by creating a modern server architecture that captures the imaginations of customers and reestablishes AMD as a thought leader, a role we played well a decade ago. When we first introduced 64-bit instructions, we were the first to introduce integrated memory controllers, and we were the first to introduce scalable CPU interconnect fabrics, all the foundations of today's server architecture. Our goal is to drive continued innovation and deliver with EPYC a substantial jump in performance, capability, and improved TCO, total cost of ownership. I think we did just that when we launched EPYC.
I'm very proud of what the team has done, very proud of the engagement with our customers and the ecosystem to date, and I'm super excited to be here today telling you about our progress to date. At AMD, we see the data center as a $21 billion-plus opportunity for us with a combination of CPUs and GPUs, and where on the CPU side, and really overall, x86 is the dominant architecture. We, of course, are one of the only two vendors that have access to the x86 technology that's most relevant to the server market today. Today on this webinar, I'm going to focus on the CPU portion of that market, and we estimate just the CPU server silicon portion to be well over $15 billion. Today, we have roughly 1% market share, a little more, but a vast opportunity ahead of us.
Today's data centers really require high-performance computing and, more and more, heterogeneous computing to be successful. Whether in the enterprise, public cloud, or the private cloud, customers are going to need to optimize their data centers for a wide variety of demanding and emerging workloads, including artificial intelligence and machine learning. Over time, that means making CPUs, GPUs, and other accelerators work well together in the data center and really make the results sing. Look, to participate at all, to be relevant, we have to have relevant CPU performance. That's the ante for playing this game. AMD's approach to data center processor development began in 2012 with the initiation of our Zen architecture, which was a clean-sheet design to enable today's workloads and applications. Zen is a very power-efficient core coupled or implemented, sorry, in a very power-efficient process that, frankly, was fundamentally developed from mobile processors.
That's the 14-nanometer process we use currently at GlobalFoundries. In power-constrained environments, which is typically what you have in high-performance server applications with many cores, many threads, a high demand, that combination means we operate very efficiently and very effectively. Therefore, we knew we would have that power advantage, that efficiency and effectiveness advantage with a large number of cores. That is certainly true for customers operating at web scale in a highly virtualized environment or in many applications of high-performance computing. This Zen architecture is the foundation of AMD's EPYC data center processor. Each first-generation EPYC processor has 32 Zen cores coupled to 8 separate memory channels, which provide, in total, a tremendous amount of performance. Harnessing that performance means first we had to wait to interconnect those cores and connect them to memory controllers and I/O controllers.
An important differentiator, therefore, an enabler in our EPYC architecture development is AMD's Infinity Fabric. This is the DNA, if you will, for our chip design across AMD's product portfolio, starting in the data center, but you'll see it in other places as well. It's really built on the heritage of that scalable interconnect fabric that I mentioned that we brought first to market with HyperTransport over a decade ago. It enables fast data transfer and the ability to scale efficiently within a die, across cores, memory controllers, and I/O controllers, across multiple die, and across multiple sockets. It does so in a very efficient way. Using Infinity Fabric and coupling the cores together, we're able to have design scalability on the first generation of EPYC all the way through 32 cores and 64 threads per socket, per EPYC server chip.
The other thing that I mentioned Infinity Fabric allows is it allows us to take a different approach to providing server silicon to the market. EPYC itself is a multi-chip 4-die module, and where each one of those modules, each one of those die, is an 8-core complex itself knit together with Infinity Fabric within the silicon and then bound together with Infinity Fabric across the multi-chip module. That allows us to have 4 pieces of silicon delivering the performance and operation of 1 integrated unit. AMD's multi-chip module approach is, we believe, absolutely the future of the market, especially given the slowing of Moore's law and the constraints that we're running into in physics.
EPYC's multi-chip approach delivers tremendous performance and capability without the limitations of high costs, low yields, high variations that you would get out of one monolithic device, which is, quite frankly, what we're competing with today. Bottom line, our multi-chip approach allows us to get more silicon and more features and performance into each CPU socket. Frankly, it gives us an amazing part with leading features. EPYC today is available in both dual and single-socket configurations. Frankly, it's the same chip. Our 32-core design is the largest core capacity in the industry, outstripping the 28 cores available from our x86 competitor. We added to that 128 lanes of PCIe to allow us the I/O connectivity to support the increasingly accelerator-intensive environments of today's and tomorrow's data centers.
This also allowed us to be able to attach Radeon GPUs, our accelerators, as well as FPGAs or other accelerators, and also flash storage, which increasingly is connected to a PCIe very efficiently and effectively to our CPU without introducing PCI bridges or other components that would slow down performance of the system. Beyond the I/O are eight memory channels per socket that support up to two terabytes of RAM capacity, deliver a new standard of memory capability for the data center, and far outstripping both the capacity and bandwidth available from our principal competitor. Beyond those performance features, all of this is wrapped up and delivered with the security and resiliency features that the modern data center needs. On EPYC, we make those features standard on every processor we sell.
A customer does not have to give up on resiliency, security, or other important features as they move to different SKUs. We don't bottleneck memory performance either at the lower-end SKUs, and we don't constrain feature sets, and we meet the market where the majority of workloads are today with a real server part. Each two-socket EPYC platform, putting two of those processors together, therefore delivers 64 physical cores, support for up to four terabytes of DRAM on 16 different memory channels, and support for 128 lanes of PCIe I/O. What does that mean? It means more cores than the highest-end part that our competitor offers to get today. Far more memory bandwidth, far more I/O, and that combination delivers for targeted workloads a huge performance advantage. Why did we design for such capacity?
Because high-performance computing, both in the way we traditionally think about it in modeling scientific applications and product development, but also increasingly in web scale applications, is working its way into every application, every business workload today. A simple example is to think about the intense modeling work that's present in product development applications. A car, for example, used to go through many physical iterations before it would move into production. Now, most of those iterations happen digitally in a computer where performance, safety, and other features, efficiency can be modeled long before the first piece of metal is bent and the first piece of plastic is extruded. Customers tell us that their need for compute will only become greater as the amount of data and devices continue to grow across their data center deployments.
The other example of high performance is social media, is a great example nowadays of high-performance computing, where the computational load to correlate friends, to correlate tweets, maybe sometimes unfortunately, is an ever-increasing demand.
Yep.
If you look at those high-performance demands, in the server market today, over 80% of the server market is dual-socket systems, because that's historically been the only way to get high performance and to get true server features. Well, with our approach on EPYC, we can demonstrate in two-socket server configurations, absolute leadership on many of the industry's most common benchmarks and applications. This is a great slide from my point of view. At the high end of the market, when comparing performance of the highest-end processors available from us or our principal competitors, on the industry standard SPEC floating-point and SPEC Int benchmarks, we compare our $4,200 EPYC AMD 7601 against the competitor's $13,000 Intel Xeon Platinum 8180M, and on both SPEC FP Rate 2017 and SPEC Int Rate 2017, we beat them. Our highest-end part has actually better or higher performance on these benchmarks than Intel's.
In the dual-socket market, we positioned our products not so much at the very high end. Again, $4,200 is a high-end processor, but a fraction of what Intel tries to charge for their highest end. We positioned our products to be substantially higher performance in the part of the market where customer actually buys processors. Remember, very few people actually buy $10,000, $12,000, $13,000 processors. AMD offers much better throughput performance at the heart of the market, and I'll show you in a moment. EPYC was specifically designed to deliver better performance in key workloads where the market is headed and where customers are buying data center processors today and out into the future. If you take a look at the price ranges where the volume actually resides, EPYC delivers more performance at every competitive price point on the industry most common benchmarks and the most common workloads.
Performance and perception of performance is everything. We believe we've delivered, we had to have to come back into the market, a leadership two-socket system that could compete with the best that the competition had to offer, but we could differentiate by bringing so much more to the heart of the market, which we believe we have. We have this with EPYC, and the opportunity to really disrupt the volume portion of the two-socket market is ours to take. Beyond that, when we began designing EPYC, it became clear to us that we could do more. Apart from just addressing two-socket, we could develop a processor with a single-socket configuration that would frankly outperform the vast majority of two-socket systems from our competitor. We could do this while driving higher performance, much better power efficiency, and much lower total cost of ownership.
Without a significant two-socket business, quite frankly, to worry about cannibalizing, we were free to act, and we have. We have a compelling single-socket approach. We are offering capabilities that the vast majority of customers that are currently buying two-socket server processors need at a substantially better TCO, total cost of ownership, through what we call our no compromise one socket. Our single-socket solution allows customers to buy the right size and the right system for their workload without artificial constraints determined by a particular SKU or a particular model. The opportunity for single-socket solutions that really deliver the right level of performance and features exists across mega data centers, cloud providers, and certainly within the enterprise. HPE, Dell, and Supermicro all took leadership positions by putting enterprise-class single-socket servers into the market and positioning them against the competition's dual-socket servers. Likewise, in the cloud, Baidu, Dropbox, Yahoo!
JAPAN, and many others all saw the promise and have adopted single-socket EPYC servers to drive substantially better TCO while delivering the performance that their customers need. TCO is for somebody whose business is their data center, it is everything. The CPU is only a fraction of the total cost. With EPYC single-socket CPUs, we can drive lower power, better cost of software in many cases, and really everything else in the network where the price is often determined by the number of sockets. EPYC delivers higher performance at really every price point for the processor, be it two-socket or single socket. An AMD single-socket solution, even when priced similarly to a competitor two-socket solution, brings lower TCO advantages for the entire project.
If you step back and look at the overall market and where the performance lands and look at the cost of buying two of our competitors' processors versus one of ours, you will see the opportunity. The majority of our competitors' two-socket systems, quite frankly, sell for an aggregate of maybe $4,500 for the pair of processors. One EPYC part, our 7551P, operating in a single-socket configuration, outperforms the vast majority of two-socket systems actually sold today. That means real performance for a much lower cost on the acquisition. When you add in power and other factors, the overall TCO of a comparable EPYC system provides, in many cases, 20%-30% cost advantages over the life cycle to similar performance or even lower performance two-socket competitive systems. All that performance is in service of doing what customers actually want. It is something I often tell my team.
Nobody wants to buy a server. Nobody needs to buy a server. They need to run their workload. They need to have a database servicing their business needs. They need to run that model of that car. They need performance for their workloads. EPYC is designed to assert leadership performance in today's key workloads from mega data center, cloud, HPC, and enterprise customers. These are workloads like high-performance computing and product modeling that I have mentioned before, virtualization for the general enterprise and for cloud virtual machines on demand, machine learning, big data analytics, and even software-defined storage, which is increasingly prevalent on the on-premise market. EPYC excels in these workloads by packing increased core performance, memory bandwidth, and IO into smaller footprints to better enable heterogeneous computing solutions with overall total cost of ownership.
Today, EPYC addresses, I would say, approximately 60%-80% of today's workloads well, with a significant competitive advantage in probably 40% or so of those. Based on our multi-year roadmap, I believe the next generation EPYC will have a growing advantage across a larger number of workloads. This should disrupt and redefine large portions of the markets as our customers and end customers find the best solution for running their workloads is an AMD-based solution. A big part of that market, of course, is the cloud. The cloud has reshaped the IT industry, whether on-premise with private cloud or increasingly with a third party. This offers, obviously, increased agility for all classes of customers, from high-performance computing through classic enterprise. For those cloud deployments, it's all about the cost per VM, the efficiency per VM.
Our combination of high core density, memory throughput, and memory capacity deliver the best cost per virtual machine in the industry, and that's been shown by multiple of our cloud customers. Think about, beyond that, the current and growing needs for machine learning to process and extract more intelligence from all of that data that's up in the cloud. AMD has a distinct advantage with a combination of EPYC CPUs and Radeon GPUs, an advantage that is again made possible by our approach with Infinity Fabric and our innovations around interconnect, as I mentioned earlier. Today, we have solid engagements across a broad ecosystem, the ecosystems that's necessary to supply the overall server industry, from mega data center through mom-and-pop shop. In terms of ODMs and OEMs, AMD has over 50 systems currently available in the market from a wide range of customers.
Obviously, we're very proud to have names like HP Enterprise, Dell, Supermicro, Sugon, and others as our flagship partners. On the ODM side and OEM, Supermicro and Wistron, ASUS, and Inventec, and others are building systems that are based on the unique specs of their mega data center customers. In terms of those mega data center customers or the hyperscale players, I'm pleased that five of the current Super-7 mega data center customers in the world are already working on AMD with EPYC for their platforms. MDCs are the largest customers on the planet and already represent approximately 40% of the total server market. They can buy in large volume when they buy. They can move the market. Those customers have been perhaps some of the most vocal in welcoming AMD back into the market, welcoming innovation and competition back into the market.
They're always looking for multiple vendors. They're always looking for innovation, they have embraced EPYC in a very meaningful way. We continue to work with them closely to optimize their platforms and software for EPYC and the AMD architecture and are making great progress. That software work is pretty tractable because we're x86, just as our principal competitor, and unlike some of the other alternatives. Data centers have been optimizing for one particular vendor for many years. There is some work to do there. We're proud at the wins that we already have and the customers that have already proven to be public testimonials for us. Some of the most recent we were just announcing at Computex last week.
We announced that Tencent, for example, one of the leaders in the cloud in China, will offer EPYC services available immediately as part of their public cloud platform, where they see and their leading customers see, beyond the general advantage of lower cost per virtual machine, specific advantages in 3D rendering and genetic modeling based on our superior and usable floating point performance, as I showed you in the SPECfp scores, but also in e-commerce, where the combination of our cores and bandwidth provide great advantages as well. Microsoft Azure is another phenomenal partner, and they've been working with us, as they announced, really at our introduction a year ago, to deliver public cloud capabilities since the launch of EPYC. Baidu, another one of the Super Seven, is also working with AMD, not just on their server platform, but in their artificial intelligence platforms as well.
The recent announcement of our joint innovation lab with them is another proof point. We expect to make additional Super Seven announcements over time as our footprint and work with them grows and deepens. On the OEM side, I said before, we're super proud to be in with many of the leading vendors in the world. We've seen Dell EMC and HPE offer EPYC-based platforms, and these are now shipping in volume. Earlier this month, HPE broadened their product portfolio, and Cisco announced a new system that I'm extremely excited about. I'm extremely excited about both of them. HPE announced their ProLiant DL325 Gen10 single-socket server, capable of providing significantly lower cost per virtual machine than the leading two-socket competitor for on-premise virtualization. It's a great addition to their already broad portfolio of EPYC-based servers.
Cisco is adopting EPYC to build and deliver the highest density UCS server solution they have ever brought to market. With 128% more cores, 50% more servers, and 20% more storage per rack available in the latest Cisco EPYC-based UCS solution, it's a compelling solution that I think will be very impactful for them. Beyond the OEMs and ODMs, the software and IHV support from the other partners in the ecosystem is growing as well. Given that ecosystem support, given the wins that we have with the cloud guys and end customers through our OEMs, we believe we have clear line of sight to achieve the mid-single-digit unit market share exiting Q4 that we've been articulating to the market for some time.
Part of that success is not just the work that we've already done on EPYC, but it's the recognition that for customers to embrace AMD solutions in the data center, they need to know that we have a long-term commitment, that their investment of incorporating AMD into their data center will pay off for the long term, that we're not going to be one and done, a flash in the pan. We have that long-term roadmap that they can believe in, and more importantly, perhaps, we've put the investment, the engineers, and the process technology behind delivering it. Investors may remember, I'm going to shift to code names here for a moment.
You may remember that Naples was the code name for the first EPYC processor that we delivered based on the Zen core architecture, and we delivered that in the middle of last year, implemented in a 14-nanometer process. We're continuing with our tour of Italy, one of my favorite countries, with the next several products as well. Rome is the code name for our 7-nanometer server chip, which is based on the Zen 2 core. That part, I'm very proud to say, we'll be sampling later this year and introducing in 2019. Even beyond Rome, our design teams have been working on the server part, code-named Milan, based on the Zen 3 core, for the past few years, and advanced teams are already well underway on the Zen 4 and Zen 5 processor cores as well.
With each generation, we're not only improving core performance, but we're also going to be delivering continuous innovation in the rest of the chip, the rest of the system, and the features that we incorporate. Again, our mission is to bring real customer-relevant innovation back to the data center and to stay innovative for the long term. We're investing in that long term so that customers can be confident to build their own roadmap around AMD future technology generations. Importantly, our multi-generational server roadmap is designed to steadily increase AMD's serviceable market with each successive generation as we ramp those deployments.
You may recall I mentioned the number of workloads that we can address with our current Naples generation, and we think that the number of workloads, the percentage of the overall workloads that are out there that we can address and competitively win, goes up substantially with each successive generation of AMD server parts. Our first part, our first year with Naples was all about reintroducing AMD to the market and really familiarizing customers, ecosystem partners, and software developers with AMD technology once more. Quite frankly, we had not introduced a new high-performance server processor for quite some time, since the early part of the decade, and we had a lot of work to do here. People remember Opteron. They remember the incredible excitement and innovation that we brought to the market. They remember that we defined the architecture of the modern server nearly 15 years ago.
We've got the opportunity with the first generation of EPYC to reintroduce ourselves to those customers that desperately want innovation and choice again. With Naples, look, we're taking market share. We've got that line of sight, the mid-single-digit unit share exiting 2018, exiting the fourth quarter of 2018. Beyond that first generation, the roadmap that I've showed you will press on with more cores, more bandwidth, more capability, and more differentiated features that we believe are highly relevant to our target customers and to the workloads that are increasingly important in the data center going forward. A highlight, therefore, of our second year, now that we're entering it, is going to be not just continuing to ramp Naples, but it's going to be the sampling of our 7-nanometer design. The 7-nanometer Rome processor incorporating the Zen 2 cores is complete.
It's in the lab. The silicon looks very good. In fact, Lisa held up a 7 nm EPYC Rome CPU.
At Computex last week in Taiwan, which I think caught quite a few folks by surprise. We're super excited by this part. We're going to be sampling it shortly in the second half of 2018. We're going to be launching it in 2019, and we are very excited about what this part, what 7 nm can do for us and for our customers. Stay tuned. In the meantime, though, we continue to drive momentum and ecosystem support for EPYC at 14 nm with what we've got today. It's always great to have customers planning EPYC 7 nm for their future deployments and expansion. I think they all realize as well that the investment they make today will pay off when the socket-compatible Rome part allows them to easily adopt it into their deployments.
As you know, commercial market moves at a deliberate pace, and we continue to invest resources to ramp their deployments. The competitive landscape is changing, and competitor 10 nm delays provide an opportunity for additional market share with customers. We're on track with 7 nm. We're delivering per our plan. Look, in summary, I could not be more excited. It does sound like I'm not excited. I could not be more excited as we build traction and momentum for EPYC in the current generation with Naples and beyond. We're gaining mind share. We're gaining market share. We've got the excitement and support of customers, the support of an ever-expanding ecosystem, and we've got a strength of our multi-generational roadmap that is well in hand. To our investors that are listening out there, thank you for your enthusiasm and support today.
For our customers that might be out there, thank you for your enthusiasm, support, and trust that you've put in AMD. We delivered for you, I think, with this first generation. We'll continue to deliver for you many generations into the future. Thank you very much.
Awesome. Well, thank you, Forrest. Appreciate that. In the few minutes that we have left, I want to go ahead and take a few questions, if we can, from the online portal. We have several questions on the line, as they've been coming in, I've been trying to group them by topic to get through as many as we can, many as possible here today. The first thing is the following, it goes to the mid-single-digit share goal that we have exiting Q4 2018. As sort of leader of the business, do you expect, based on customer engagements, that we would gain share faster in the cloud and data center, or would it be OEM? How do they buy and in quantities, what are your thoughts there?
Yeah. The mega data center customers tend to buy in larger quantities when they buy, and they tend to, quite frankly, move a little bit faster because they're collecting for a smaller number of workloads. The more traditional enterprise end customer has a rich set of applications that have built up in their data center over a very long period of time. They're running everything from their Oracle database to their web servers to, Lord only knows what, doghouse controller software. They take a little bit longer. They're reached through the OEM market. The OEMs have done a great job embracing us and making EPYC available. I do think, when I think about just this year, the data center customers will move a little bit faster this year, or are moving a little bit faster this year than the end customers.
Okay. In thinking about, again, staying on market segmentation, I've got a couple questions here. Do you have sort of share aspirations? You've talked about mega data center, then I know there's enterprise on the other side, and there's a type of data center in between. Do you have share aspirations within each of those?
Yeah. We'd probably classify them as the mega data center is about maybe 40%-45% of the overall server CPU market. Frankly, of those, the top three probably are close to 10% each.
Wow.
Microsoft, Google, and Amazon.
Got it.
There's probably about 35% of the market is the traditional server OEM servicing more the traditional enterprise through SMB customers. The balance is probably HPC and the next wave cloud companies. I would lump them together because they've got very similar characteristics for making a decision. It's always a very technical sale. It's always a very tight performance criteria or TCO criteria.
Just like Dropbox.
Yeah, like Dropbox, Rackspace.
Got you.
Then some of the HPC, high-performance computing customers as well. A lot of those, they're really being by going by the moniker of the next wave, the next wave of cloud companies. That's the partitioning. When I look at what EPYC can do, where its workload performance really excels, what it's really good at, quite frankly, we see pretty strong applicability into all three of those segments. Over time, I would love to have a pretty balanced share across those three segments. As I mentioned a moment ago, the mega data centers probably are moving a little bit faster, this year, the share is probably a little more predominantly skewed towards them then the next wave, then traditional enterprise SMB in that order.
Okay, when you're ramping, this is a question I've gotten many times too in IR, when you're ramping a new product like EPYC to a data center customer, can you talk about what that ramp looks like? What is sort of the process that you-
For the mega data center?
For the mega data center to ramp.
Yeah. First off, it's a long process. We began, I think this is indicative, we began engagements with the mega data center customers early last year, very early last year, maybe late 2016, when we really had EPYC processors that were representative of the full performance that we were going to be bringing to market. Part of that was, quite frankly, we had credibility to regain. Some customers were, "Show me.
Yeah.
They're from Missouri. "Show me that you really do have what you say you have, a high-performance processor." We began working with them. We would ship small number of reference designs. They'd kick the tires. Generally, they all have the same reaction, which was, "Holy cow. You're real. You've been telling me this for the last couple of years. You actually delivered it." That would get them interested. They would typically go to their favorite ODM or contract manufacturer. They would specify a system. They would build hundreds of units of a pilot build. They would do initial trials, again, in their configuration for their workloads, is this performance really being delivered? They would do, for the ones, they were all, I think, pretty pleased. They would do pilot runs of thousands of systems.
To make sure that nothing happens when they put those pilot systems into clusters into the data center, do their operations get disrupted? Is the performance available at scale? Do their sysadmins have any learning curve to learn the new EPYC-based systems? They would go through that, and then finally, they would start moving into production-
Yeah
tens or multiple tens of thousands of units. Most of the cloud guys now are in that initial production or maybe the pilot thousands of units, but I'd say most of them have made it through. That whole process is nine months to a year.
Okay. That's where. All right, again, looking at the portfolios, we have a number of questions like, can you talk about where you will be-- There's a number of foundries. Where will you be manufacturing EPYC?
EPYC is a 7-nanometer processor. We're very pleased with-
Yeah, 7 nm EPYC is what I'm talking about.
Yeah, we're very pleased with that. Yeah, Rome is a 7 nm processor. I assume that's what your question was.
Yes, it was.
Yeah, the current product is 14 nm being built by GlobalFoundries.
which is a strong partner of ours.
Yep.
Our other major partner is TSMC. We really use both quite a bit. We're going to use both for 7-nanometer. We haven't commented beyond that, where we're specifically building the EPYC 7-nanometer part.
Okay. What would you say to analysts whose channel checks say that they may not be seeing much in the way of EPYC momentum at this point?
Well, it's still early. We made very nice progress in the last few quarters. Obviously, in Q1, we talked about double-digit server revenue growth, doubling the EPYC unit volume. We grew quite substantially since we introduced. We really are getting traction across all areas, with obviously the MDCs in front, then channel and the traditional enterprise OEMs. Again, I said it earlier, we have line of sight to getting to our mid-single-digit market share by the end of the year. I'm very comfortable with that goal, and I see exactly how we're going to achieve it based on the engagements that we have and the way that those are ramping. Look, a year ago, we were essentially 0% market share.
Yeah
when we started.
0.2.
0.2% market share. Yes, thank you. We ramped beyond 1% in Q1. We're on track, but we're still a tiny portion of the market in terms of what we're doing today. That just, again, means incredible opportunity ahead of us.
Okay. Then offering something similar here, so I'm going to try and do this justice, but sort of Given recent comments in the marketplace today by our competitor, does a delay at 10 nanometer technology for them mean an increased opportunity for AMD?
Well, there's sort of lies, damn lies, statistics, and process node names.
What our competitor refers to as their 10 nm part or 10 nm process, and the industry standard 7 nm process that we have access to from multiple foundries are in many ways roughly equivalent, with maybe a slight nod to the 7 nm in terms of SRAM density and a few other parameters. I think, for the first time in a very long time, maybe ever, we have access to technology that is at parity. The foundry industry is going to be at parity with our principal competitor, and that's a new paradigm.
Yeah.
In manufacturing process technology, sort of a law of nature for the last 40 years has been Intel has got a process lead. That law of nature is broken with 10 nm versus 7 nm. We've made a big bet in 7 nm. We're very comfortable and confident with the 7 nm process, and we're going to be sampling our 7 nm parts very shortly. That 7 nm process gives us a lot of capabilities. It gives us a lot of options, increased density, much better power efficiency. That translates into more capability, more performance, more features. We're adding extra innovation on top of that. As I mentioned earlier, Laura, it's not just the process node and dropping the same core into the next generation. We're doing new stuff. We're making architectural innovations as well. I think, bottom line, we've got a pretty aggressive roadmap.
I think we've got architectural innovations on top of a solid process that our foundry partners are delivering in production, in advance of when we're going to be bringing EPYC 7 nm into production. Beyond that, there's some uncertainty around when our competitor's 10 nm will actually be in production. I'm very comfortable with our 7 nm versus their 10 nm, and anything beyond that would be upside.
Terrific. Thank you, Forrest, and congratulations again on the first anniversary of EPYC. It's exciting. It seems like there's more good things to come. Ladies and gentlemen, thank you very much for joining us on the webinar today. Please contact a member of AMD's investor relations team with any follow-up questions from today's call. We thank you for your participation and your continued support of our company. This concludes our call today. Thank you.