Welcome to everybody on the webcast as well. We have several hundred people here with us today. Our 2015 Financial Analyst Day, we have our executive management team here today who are going to give a series of presentations. Lisa Su, our CEO and President, will kick off, followed by Mark Papermaster, who is also going to give us an in-depth review of technology. Then we will go back to Lisa as the interim lead of our Computing and Graphics segment. We will take a break for about 15 minutes, then we will go on to Devinder Kumar, who will come after Forrest Norrod. Forrest is a new face to us here at AMD. He is the lead of our Enterprise, Embedded and Semi-Custom Business. We have a great demo area outside.
I hope everybody has had the opportunity to take a look at that. You will also have more time at the break to go through it in further detail, as you need be. There is a big queue for the Oculus virtual reality demo. If you have not tried that out, that certainly will be a lot of fun. I have some cautionary statement guidance update I would like to give everybody. Before we begin today, let me remind you that today's discussion contains forward-looking statements on the environment as we see them. Those statements are based on current beliefs, assumptions, and expectations, speak only as of the current date, as such, involve risks and uncertainties that could cause actual results to differ materially from our current expectations.
For a detailed discussion about our risk factors, please review our SEC filings, our quarterly AMD reports on Form 10-Q for the quarter ended March 28th, 2015. Additionally, please note that non-GAAP financial measures during this meeting are reconciled to the most directly comparable GAAP financial measure in the presentation that will be posted on our website, amd.com. Just a couple of small housekeeping items. We would appreciate if you would put your phone on silent, please. If you could refrain to keep your questions and answers to the end of the presentation today after the break, we will have a very robust question and answer session for everybody. Last but not least, I would like to thank all the AMDers who really put this event together today to make it happen. If you could give us a round of applause, I would like to introduce AMD's CEO and President, Lisa Su.
Hey, good afternoon. How is everyone doing?
Very good.
I want to really thank you for being here with us today, for those of you here in New York and for everybody on the online webcast. It's a great day for AMD. It's actually an important day for us. It's the first time in over 3 years that we're giving a comprehensive view of our strategy. It's also my first time as CEO of AMD being able to address all of you as our stakeholders in the company. What I'd like you to take from today is we're going to go through a lot of information. We're going to go through products. We're going to go through technology. We're going to go through financials. We promise we're going to go through financials, but you have to wait for Devinder at the end. Most importantly, I want you to get the spirit of AMD.
We are a technology company, and we are here to bring great technologies to the market, and we're going to show you what those are over the next couple of hours. I'll start with a little bit of a view of where we're. We have a few slides and a few agendas here. I'll start with my discussion today will be a broad overview of where we're going as a company. I'll start with some of our progress to date, because we're really a couple of years into our transformation, and we have made good progress. The market has been changing at an unbelievable pace over the last couple of years, and I'll talk about how the markets have changed and how we, AMD, are participating in those markets. The most important piece is how we're sharpening AMD's focus.
Many of you have asked me, what are you going to do as CEO? There are a lot of wonderful things about the company in terms of our technology and our foundation. What I can do is sharpen the focus and make sure that we're spending our resources in the most critical places. Of course, Mark is going to go into detail about our technology and the exciting new technologies that we've been working on. Finally, all of us will talk about our execution priorities. These are the key things that I wanted to cover today. Now, starting from a little bit about our heritage. You guys know AMD, and you know us well, but I wanted to give you sort of my interpretation of AMD. We just celebrated our 46th birthday, actually, on Friday. Friday, May 1st was our 46th birthday.
If you look at the heritage of the company, I'm actually extremely proud of what we stand for. What we stand for is innovation, technology innovation at our best, ensuring that we do things in the industry that either other people didn't want to do, didn't prioritize, or didn't think were important. That's our history. What are some of those things? The world's first 64-bit x86 processor. That was AMD innovation. As we look at the GPU barrier, breaking a gigahertz, that was an AMD innovation. When you look at the vision that we had over the last five years of putting CPUs and GPUs on chip to really build APUs and really drive the next generation of technology. Now, if you look, all PCs, smartphones, tablets, they have some version of APUs. That was something that we recognized over five years ago.
If you look at game consoles and the technology that's changing, we talked about some of the demos out there. The fact that our technology is in all of these innovative products, these are things that are special about AMD. This is a heritage that I'm extremely proud of, but it's also one that we absolutely need to extend as we look over the next five to 10 years, what are those next firsts that are going to be uniquely ours? A little bit about our progress over the past couple of years. If you look at the heritage of AMD, a lot has been spoken about the PC market, and I'm sure we're going to talk a lot about the PC market today. The PC market is a big market. It's over 300 million units.
There's been a lot of ebbs and flows, sometimes good, sometimes bad. If you look at the AMD business, though, in 2012, our singular focus in life was PCs. Over 90% of our revenue was in the PC market. We decided at that point in time that we needed to diversify our businesses. We had great technologies, but the technologies could apply broader than the PC market. We started that diversification activity, and I'm proud of what we've done. At the end of 2013, we had over 30% of our business outside of the PC market in new applications in our Enterprise, Embedded and Semi-Custom space. By the end of 2014, over 40% of our business in that area.
What we're trying to do is build a company on a very solid foundation of both diversified markets, great technology that go into the places that are going to grow. This is a story of the last couple of years. We did make progress, and I'm proud of that progress. This is really only half the story. The other half of the story is what we're going to talk about over the next couple of hours in terms of really where are we going to complete this diversification of our business. No story is set without a foundation of where we are in terms of the current business. I think there's a lot been written about AMD, and there are positives and negatives, and I want to acknowledge those going forward. What are some of the challenges?
I said earlier there's been a lot said about the PC market, there will be a lot more said about the PC market. The PC market is a bit volatile. Last year was better than expected in 2014. This year, at least the first half of 2015, worse than expected. It has affected our business. We've also had some channel inventory issues that we have been correcting over the past couple of quarters. I'm happy to report, though, that is making good progress on track to what we thought it was going to do. I also want to acknowledge that we have had some market share losses in some of our key businesses. Many of you will ask, "How do we expect to reverse those market share losses?" We'll talk about that today for sure.
When I look at the positives, though, when you think about what we have done over the past couple of years, this is fundamentally a different company. We have grown an entirely new business in EESC that's over $2 billion at the end of 2014. We absolutely have gained share in profitable areas of the market. We've talked about commercial. We've talked about professional graphics. We've talked about gaming being a huge driver of our growth. As we look forward, embedded, and we're going to talk quite a bit today about data center will be growth areas for us going forward. Devinder is going to talk about the balance sheet and the importance of the balance sheet, ensuring that we manage our cash well and ensuring that our near-term debt maturities were pushed out. All of that was done in the past couple of years.
One of the most important things for today is to ensure that you understand where are we focusing our investments, because that is key. To ensure that we focus our investments in the places where we will get the most bang for our buck. I want to start with 2015 expectations because I feel like if I don't cover this chart, I might not get you to listen to the next hour or two. The 2015 expectations, we gave guidance a couple of weeks ago at our earnings call about the second quarter of 2015. Normally, we wouldn't give long-term guidance.
However, to ensure that we answer some of the questions about where our business is going and where we see things are going, I'd like to give you some high-level view of the financials, Devinder will go through more detail when we go into the financial section. First half of 2015, we've talked about. It is a weaker-than-expected PC market. We do have some channel inventory rebalancing that we are doing. From a product standpoint, most of our new products will actually launch in the second half of the year. We have reported, and we have guided that we'll lose money in the first half of the year. No one enjoys losing money. That's certainly something that we believe we need to correct. As you look into the second half of the year, what do we see in the market?
We do see seasonal improvement in both PCs, gaming, and our other markets. That's expected based on some of the factors in the market, as well as some of the product announcements, our product announcements, as well as our partner product announcements with Windows 10. We expect that that will make the second half of the year stronger. In terms of what we're looking at, we're looking at being profitable in the second half of the year. Importantly, on the cash question, Devinder is going to go through the cash drivers in quite a bit of detail so you can see the pluses and minuses as we go forward. We will be approximately at $800 million of cash at the end of the first half of 2015, we expect that to be somewhere between $800 million and $1 billion as we go into the second half of 2015.
Hopefully, that grounds you a bit on our assessment of 2015. Again, we have some issues in the first half of the year. We expect the first half to be the trough of our business, we expect the second half of the year to be better than the first. That gives you a little bit of a view of 2015. Now, let me talk about our game plan. Our game plan is really a three- to five-year game plan of how do we really change the mix of our business, ensure that we're growing in profitable areas, expand our margins, and expand our cash generation. That's really the thought process. How do we get there? It's a combination of technology, products, markets, as well as execution. We've talked about 2015, the second half of the year.
I'm going to talk in some detail about our 2015 products in the computing and graphics section, Forrest will talk about our 2015 products in the Enterprise, Embedded and Semi-Custom section. You'll get a view of the product ramp that we expect in the second half of the year. In 2016, what do we expect? We expect to gain profitable share in graphics In certain areas of the PC segment, I'll talk about those areas, in embedded, we had previously reported two semi-custom design wins at the end of last year that are in process and are in development. We expect those to ramp in the second half of 2016, that will be a revenue driver for us going forward.
As we go forward into the medium term, going into the 2017 and 2018 timeframe, we're going to talk about server and infrastructure. This is an incredibly exciting area for us, we're probably going to spend more time on it than we've spent in a long time to show you our story of why we believe this will be a long-term both revenue, margin driver, and profitability driver for the company. That gives you a couple of steps that we will go through over the next couple of years. Now, in terms of the market, I want to define what are the markets that AMD is uniquely capable of participating in. Let's just start with a broad view of the market. You guys know this really well. Everybody's talking about the Internet of Things. What are the Internet of Things?
It's the idea that you have a bunch of endpoint devices, both sensor and compute driven, that are connected through a network into the cloud, where you're able to sense, analyze, and process all of that data. Incredible growth. Cisco says over 50 billion devices by 2020. Incredible use for computing horsepower in every part of it. I will say, though, the last five years, most of the conversation has been about the endpoints and what the endpoints are going to do. As we look forward to the next five years, we see growth in a couple of areas. As they relate to AMD, what we are looking for are the places that use high-performance compute, high-performance graphics, visualization technologies, complex system on chips. Where do you expect that? We expect that the intelligent controllers will become more intelligent.
The idea that we're going to get much more intuitive in how we use devices. We expect that this entire back-end backbone, whether you're talking about the networking or cloud infrastructure, is going to be significantly increased, improved in terms of the ability to store, manage, analyze, use that data, and there's a lot of computing growth in there. When you look at the AMD TAM, and I know there are a lot of analysts in the audience here, my TAMs are approximate TAMs, we think there's about a $60 billion TAM in here. It's a really very strong area of growth. The areas that we're focused on are the areas that require high-performance compute, graphics, visualization technologies. Those are the areas that are uniquely suited for AMD.
I'm going to spend a few minutes on this chart, and I'm going to do that for the sake of as important as it is to say where we are going to invest, it's important for us to say where we're not going to invest. If you take that entire IoT market, I'm going to sub-segment it into three key market segments: gaming, immersive platforms, and data center infrastructure. We're going to spend a lot of time on those areas over the next couple of hours to talk about the technology that's suited for this, as well as our products and our differentiation. We like these markets. This is the $60 billion TAM that we're talking about. As important is where we will not invest.
Where we'll not invest, it's always hard to say because somebody's going to criticize, "Well, why aren't you doing that?" When you look at the technology that's needed, IoT endpoints, very interesting silicon consumer. More low-end microcontrollers and sensors, not really conducive to AMD technology. When you look at low-end tablets and smartphones, again, big silicon consumers, not necessarily aligned to where our technology specialties are. We won't be spending as much time in those areas. We will talk a lot about the technology, and I don't want to say it's all about the technology, but it's pretty much all about the technology. It's extraordinarily important to ensure that we have competitive high-performance cores. I'm going to use those words, competitive high-performance cores, and ask you to listen to Mark Papermaster when he talks about our roadmap.
In terms of other decisions, we've also made some other decisions that we've talked about before, but I'd like to go into a little bit more detail. We have reduced our low-end PC exposure. What does that mean? When you look at AMD's historical business, we were very, very heavily concentrated in the consumer low-end PCs. That was actually our specialty. When you look at that market, there's been so much volatility, especially at the intersection of tablets and PCs, and the differentiation isn't really there. We've reduced some of our exposure in that area, and some of the market share losses that we've seen in consumer PCs have been related to that. We also looked deeply at the data center business and said, "Where do we think we can really uniquely add differentiation?
Which part of the market should we go after?" Forrest Norrod is a data center specialist. He's going to talk about it in great detail. We made a decision to exit the SeaMicro dense server system business. One, because microservers were not growing as fast as originally thought, and two, we really aren't a systems company, this was an opportunity for us to focus. On the silicon side, very, very clearly, we are an x86 company. We have tremendous x86 heritage. We have an opportunity to really lead in that area. We're absolutely going to invest in high-performance x86. On Arm, what we're adding is additional TAM. There are certain segments of the market that have high affinity to x86, like PCs and traditional servers.
There are certain segments of the market that have high affinity to Arm, like networking and storage and other embedded applications. That is one of the reasons that we've decided to invest in the Arm side. We're also simplifying our roadmap, and simplifying our roadmap means there are certain things that we were doing before that we're going to not do now. For example, last year, we talked about the idea of doing socket-compatible x86 and Arm devices. We called that Project SkyBridge. As we talk to customers really want x86 and Arm from us, but they don't necessarily need them in socket-compatible versions. We're going to focus on differentiated solutions in each one of those areas. That's what I mean by simplifying our roadmap.
A little bit about the markets, and as I said, we're going to spend quite a bit of time on these. We love gaming. Gaming's a great market. Immersive platforms, I'll explain a little bit more, data center, we will lean into as well. First on gaming. Gaming is a great market. AMD has been investing in this for the last decade. When we look at how much horsepower, both on computing and graphics, are needed in gaming, it's really, really amazing. It's not just where we are today as we go through some of the key inflection points, moving up in the display capability, going to 4K, as well as some of the other applications. It's some of the things that we're seeing in virtual reality.
As Ruth said, if you haven't seen virtual reality in action, you actually must because it is an amazing thing. In terms of where our business is today, this is the biggest segment of our business. It's about $3 billion in annual revenue. That's including game consoles, that's including PC gaming, graphics, as well as our embedded gaming. This is a nice market for us. We're going to continue to grow in this market because we believe gaming will continue to grow going forward. Immersive platforms. What are we trying to say here? In immersive platforms, we're choosing those portions of the IoT connected device world that really bring to life natural user interfaces. The idea that we want to communicate with our devices in ways that are much more intuitive and much more interactive going forward. There is an incredible use of computing technology in this area.
Again, Mark and I will cover it a little bit more later. It's a broad set of applications that goes beyond entertainment, but really goes into industrial applications, engineering capability, analytics capability that Forrest will talk about. Again, this is about a billion and a half of our revenue today. It's primarily in the PC portion of the platforms, as well as some of our embedded business. We see this market going to over $20 billion over the next couple of years as well, really led by some of the work in virtual reality and augmented reality that we've been talking about. Finally, data center. Data center, as I said, is a place that we're going to spend time today. This is probably the single biggest bet that we're making as a company.
What we'll lay out for you today is a comprehensive data center strategy. Why we think it's important, why we win, why we have the technology, why we have sustained competitive advantage. This is a place that AMD has absolutely been successful. We have not been competitive the last few years based on some of our investment choices. We will be competitive in the data center market. This piece of our business is less than $300 million today. If you look at the total TAM of data center and add networking and storage components of it as well, that's a $15 billion TAM. That's margin accretive business that plays directly into where we're investing our capabilities.
To try to now give you a little bit more deeper framework, and I know it's all about the details for this group, we've laid out gaming, immersive, and data center across our computing, graphics, and Enterprise, Embedded, and Semi-Custom business. We've sub-segmented the market as to where are the most important places for us to play. We're number one today in game consoles and gaming. We're going to continue to invest in that area. In terms of near-term opportunities, we absolutely have the technology to gain share in graphics. Mark will talk about some of that technology. In the PC platforms, we will be focusing our PC platforms on commercial as well as high-end consumer that can really utilize all of our visualization technology. Forrest will talk about some of the embedded opportunities as well. In data center, this is practically greenfield for us.
We know the market. We know what's necessary. This is our medium-term growth opportunity for the company. How do we win in terms of technology investments? I'm not going to steal all of Mark's thunder, but I will say a couple of things. Pay attention to what he says about CPUs. CPUs, both next-generation x86 and Arm cores, this is a multi-generational investment for us. We're extraordinarily excited about the technology in this area. Graphics technology, it's always been a horse race in graphics. It'll continue to be a horse race in graphics, but I will say that there are areas that we will double down and focus. We'll announce, for example, today, one of our first graphics products that will use high-bandwidth memory. That's a first for the industry.
We will talk about our performance per watt leadership, as well as some of the new features that we're putting in for virtual reality. That's leadership. On the platform side, Mark and I have been part of AMD for a couple of years, and some of you ask, "Well, what have you guys been doing for a couple of years?" The truth is, it takes a while to really transform both the R&D capability, the technology, and the modularity. Mark's going to talk in detail about that. These are the places where we're really focusing our technology investments. I want to talk about market share because market share is key. What is the most important thing for us? It's gaining profitable market share. Where are the areas that people truly value AMD technology?
In the near term, it's going to be our second half product ramps that I'll talk about in a little while. We will see the seasonality improvement, as I said. In the midterm, in 2016, we have our semi-custom and embedded design wins that we've won that will be in the process of ramping. I believe we will gain graphics and commercial share. We have our commercial lead here, and there's been a lot of focus in this area. As we go forward, it's really about our first strong data center portfolio over the last five-plus years. I think that will be a key driver. I also wanted to give some additional color on our semi-custom pipeline because we often get asked about our semi-custom pipeline. Our semi-custom pipeline, you should think about semi-custom is a business model.
It's a business model for how you build deep relationships with customers. We've now had a chance to go through several cycles of semi-custom, and what we've seen is customers who start with us as standard products over time see the need for semi-custom as a way to specialize. I won't give details about the semi-custom pipeline, but I will say that we expect to sign another semi-custom design win this quarter, in the second quarter, that will ramp revenue in the 2017-2018 timeframe. That gives you a picture of these are the markers that we're looking for in terms of returning profitable share growth. Now, in terms of the long-term financial model, again, Devinder is going to go through the long-term financial model in gory detail as we get towards the end of the presentation.
I just wanted to give you a flavor of how we're thinking about the business because it'll give you a flavor of some of the decisions that we're making in terms of where we invest. We've talked about the transition of our business from a PC-centric business to one that's more balanced between enterprise, embedded, and semi-custom. We're going to continue that diversification of revenue. We expect, and when I say target, I mean target as, let's call it, 3 to 5 years. We expect that our revenue will exceed 50% in our new businesses, our enterprise, embedded, and semi-custom. PCs will continue to be a large piece of our portfolio, and there are segments that we think we can grow there. In terms of where we're targeting, it really is about targeting earnings per share growth and overall profitability of the company.
We've set a $0.50 EPS share target. Devinder will go through a number of the different ranges in other things like OpEx and gross margin and so on and so forth. That's what we're looking for as our 3-to-5-year model going forward from a financial standpoint. I want to make sure as we close this introductory portion, that I also give you an idea of what we're trying to do at AMD beyond the technology. We've talked about technology, we've talked about markets, we've talked about products and market share drivers. I want to spend a few minutes on the culture, the culture of the company. When I took over as CEO, everybody asked, "What do you want to do?
What is most important to you?" Yes, there are many business things that are very important to me, but it was very important to me that I set the culture of the company. The culture of a company is really a little bit of a return to our heritage. We are a technology company. We have been at our best when we have delivered technology on the leading edge, and we've taken bold risks, and we've come out with something that nobody else has done. Our first and total priority is around building great products. It's building great products. If you ask any engineer in AMD what are they doing today, it's about building great products and understanding that's what we stand for. Those products need to be informed by markets and deep customer relationships. We want to make our customers incredibly successful, and we've done that.
When you look at the work that we've done with Sony and Microsoft, I'm incredibly proud of what we've done there. That's the culture that we're trying to drive. The last point is, hey, it's not that complicated. Okay? We don't need to do everything. We need to pick the things that we can do and do them really well. This is the culture, this is the mission, this is what we're trying to do from an overall AMD standpoint. Let me close by telling you a few things. I'm very excited about what we're doing. I'm very cognizant of the near-term challenges that we have, but also the long-term opportunities.
I think what you're going to hear from us over the next couple of presentations is detail on where those opportunities are, where we're prioritizing, why we believe we have a roadmap for profitable share growth. Thank you very much for your attention. With that, I'm going to introduce Mark Papermaster, our CTO. Mark?
Thank you, Lisa. All right. Look, I have fun talking about technology, there's never been a more proud day than today to talk about what we've been at, what have we been working on the last several years, which we think will be quite impactful on the industry. You know what? I'll connect it to what Lisa talked about, our focus, gaming, immersive platforms, data center. Really, when you look at these trends, you can say, well, there's something just very simple in terms of what is a driving force and the need for technology. First is in front of the screen. It's the perception we have. It's how we're interacting with computing so that it can be useful. Of course, it's what's behind the screen, the computing. When you look at what's driving this change of need and requirements of visualization, it's pretty clear.
You see it around you because it is the billions of connected devices. What is that doing? It's creating a massive amount of information. It's the sensors which are embedded in every device you deal with. It's the cameras you see everywhere and the flood of video information transport. It's the analytics that businesses in almost every vertical are creating a massive amount of data that you need to then analyze and make useful. It's fundamentally changing the needs and the requirements to high-resolution displays and multiple displays, and whole new categories of how you deal with information from both entertainment to really using that information in virtual reality and augmented reality, where you can actually manipulate that data. The needs of visualization are absolutely in an inflection point. Of course, how do you analyze that data?
That flood of data that you have needs that heavy lifting, and it's seamless between what you're using in your device. That's where you need to do some computation device, but it can't accomplish that heavy lifting without an assist from the cloud. Those are the markets we're focusing on, and that is where we've been doubling down in terms of the building blocks at AMD. No surprise, it starts with the engines. You look at the CPUs, we have redoubled our focus and new design work on CPUs, both Arm and x86. We've redoubled our focus on graphics. We've always had leadership. It's always a leapfrog race of leadership in graphics, and we've maintained that pace. We're uniquely positioned to have a leadership CPU capability, graphics capability, and the ability to put them together in a very seamless way. It's not just the engines, right?
That's really a significant change in culture in AMD as well, because it is about delivering solutions at the end of the day. We've been investing in the software and the accelerators, to actually unlock the value, right? It's not about just putting out great engines and saying, "Here, industry, have at it." We've got to enable that. We've made significant advances, APIs that have driven a tremendous improvement in graphics performance. We have hardware codecs to allow seamless 4K video streaming. We have a software enablement for emerging workloads, that really can use both the CPU, GPU as standalone or in fact, together on an APU.
One thing that should be very clear to all of us, when you look at some of the data breaches, security breaches that have occurred in just the last year and a half, is the fundamental need for security. That's a core competence. You can look at the work we did with game consoles, and what we've done across our roadmap. It's a core competence we have, and security has to be a part of every product which we deliver. With that, what I'd like to do is dive down a little bit. I'm going to walk through each of these and tell you what we've been up to, and I'm going to start with the cores. You think about it, Lisa said it. It's hard to, and it's unfortunate, it's the nature of our technology business. It's not software.
You don't rewrite a line of code. It's not just hardware. It's software and hardware coming together, and the longest lead time is the core development. What we've been doing the last, I'd say two and a half years at this point, we have put together a new x86 core. This is a ground-up new microarchitecture. We're very proud of it. We're going to share a little bit of the information on it today. What is this core? This core is about high performance. We don't back away at all from the energy efficiency that we've been designing into our x86 CPUs, but we've doubled down on the performance characteristics of the Zen core. It's got features for very high throughput, like simultaneous multithreading. It's got a very efficient design. You have to feed the core.
You have to be able to feed it with high bandwidth, with low latency. We have a new cache subsystem and a new memory subsystem to be able to feed this new core. Very key is FinFET technology, because we've married all of the design management techniques we've had with FinFET, and it gives us excellent scalability. We can take the Zen core from battery-dependent mobile devices right to enterprise-class compute capability. This is a core design for the workloads of the future. Availability, next year. We're completing that design this year. We'll have availability next year. I'd like to explain to you a little bit about the performance and how we built that performance. It starts with the previous family, because what have we been doing with the previous family from the Bulldozer core introduction in 2012 to the Excavator core that we just released?
It's coming up shortly in our 6th-generation APU. That's the Excavator core. We've been doing incremental instruction per clock, incremental microarchitecture changes, and driving the energy efficiency in a big way of this previous generation core. It again culminates in that Excavator core release this summer. It has a 40% improvement in energy efficiency, and we packed the density by almost a third of previous generation. Tremendous advancements as we've ratcheted the previous generation. That wasn't going to get us back to where we needed to be in high-performance cores. What we did with the investment is there's not any one thing you can do.
Jim Keller came on board, heads up our cores team, he rallied and brought the team together and focused on not one silver bullet, but a number of elements to drive the microarchitecture improvement and deliver what I've not seen in the industry before, a 40% improvement in instruction per clock. It's quite remarkable, it's a result of many elements done quite well to deliver that type of microarchitectural improvement. We're not stopping there. One of the key things that we have to do at AMD is have sustained innovation. We're not going to do a flash in the pan with the new core that we come out with next door next year. We want regular updates. We want to stay on the pace. We have leapfrogging design teams.
Our next-generation core is already well underway, what you'll see is a family of cores over time. This is a huge commitment from AMD and a commitment to have sustainable innovation. We're very excited about our core work, you'll hear more of the presentation as to how that rolls out in our product roadmap and the impact it has on the markets that we can serve. Okay, I want to talk about Arm, give you an update. We announced the K12 core at a session we had last year. We said we had a ARM v8 architectural license. We've made great progress. The team has leveraged that license. We brought the core up. We have it up and running in simulation. What we've done, to give you the update, is we've chosen to really leverage the synergy with our x86 core work.
It's all one core team. We don't run it as two separate efforts. What we're doing is fully leveraging that synergy. We have put the availability in 2017. We've aligned it with our customer needs, the market needs, and the Arm ecosystem for enterprise applications coming together. That's the update on K12, it's on track for 2017 sampling. All right. The other side of the core equation is, of course, the graphics. This is where AMD has such a long history of graphics innovation. I'll go back to, not by any means the beginning of that long history and the ATI heritage, but I'm going to go back to 2012, where we rolled out Graphics Core Next, new architecture for our Radeon graphics.
When we did this, it was quite revolutionary because it didn't just bring a new high-performance bar, which it did with the R9 290, but it did more than that. With Graphics Core Next, it provided an architecture to make it easier to write optimizing compilers, to make it easier for programmers to unlock the value of our graphics, both for visualization and compute. Really a groundbreaking architecture. You can see the adoptions. It's, of course, our discrete graphics business, but much more than that. It's in the game consoles. You see it in the Mac Pro. You see it in the iMac with a 5K Retina display. You see it in leading workstation designs, server installations. It's scalable from mobile applications to high-performance computing applications, and it's got an install base of over 100 million for developers to leverage.
This has been a very successful architecture, we've built on it. We built on it in the last 2 years by really accelerating the performance with APIs. Mantle shook the industry as we showed that you could eliminate intervening layers of software and rapidly increase the draw calls, the visualization capability of our graphics core. We've added advanced memory with compression techniques. We've added virtualization of our graphics. We've added a set of capabilities in virtual reality that we're very excited about, and we're going to talk about over the course of the day here. We've added high-bandwidth memory, which I'll talk about in a little bit more detail in just a second. But Raja Koduri and his team on graphics are not stopping. Like the CPU cores, we're working always on the next generations. What you'll see is we're taking our graphics also into FinFET technology.
We're going to push our graphics to have all that same leverage, we're tweaking the GCN to make it even more energy efficient. In fact, as we release our third-generation GCN, it will have a 2X energy efficiency improvement, we're going to drive yet further optimizations on workloads that we are already seeing excel on GCN, such as machine learning and a number of artificial intelligence applications. Graphics technology, we've been delivering. We're very excited. There's more innovation to come, again, I want to dive down into high-bandwidth memory for just a moment. This is quite a feat. We've actually been working on high-bandwidth memory for almost 7 years. Why? It involves, frankly, an ecosystem. It's what are we doing at AMD with our technology, what are the memory vendors providing, and the packaging ecosystem to put it together.
The whole supply chain has to come together to deliver HBM. When you do, it provides dramatic results because what HBM does is it stacks the high-bandwidth memory DRAM using through-silicon vias to allow a 3D stacking of that DRAM technology. We then connect it to a silicon interposer to our discrete graphics chip. It's not going off-chip in those connections from the memory to the graphics. Well, what does that do? What that means when you're not driving off-chip is you are more compact and you require far less power. That's why we're able, with HBM technology, to deliver significant improvement. We'll see over 3X performance watt improvement versus graphics DDR5 memory, which was deployed in our AMD products and across the industry. It's a 50% power savings.
This is quite impactful. We're excited as we roll it out this year in our discrete graphics roadmap. More on that to come as we talk about our products shortly. All right. Let's keep going. Let's talk about putting this together, putting the solutions together. The engines are critically important. They're the foundation that we build everything on with our CPU and graphics. Again, we've doubled down our focus on the enablement. What I want to do is start with, really, a discussion on energy efficiency. In every segment that we play, energy efficiency matters. It's the battery life in mobile. It's the total cost of ownership in the data center. You can't compete if you don't have the technology to bring energy efficiency in every market you play. We've made a lot of progress.
We have brought a 10x improvement over the last six years. If you look up to 2014, it's 10x improvement in energy efficiency in our mobile designs. That's not where we're going. We're committing to a 25x improvement in energy efficiency between now and 2020. We're well on the way with Carrizo, our sixth-generation APU. It's driven by the enhancements in the Excavator core and much more than that because we drove design elements. Of course, it's hardware design elements with clock gating and power islands and all the techniques of the IP elements, but it's much more than that. It's across the software stack. It's through managing power on the display. It's all the elements you need to bring together. Excellent step with Carrizo, much more to come. Energy efficiency has and remains vital and front and center in our R&D development efforts at AMD.
Next, I want to talk about security. I mentioned it earlier. It's simply fundamental because what you're seeing is software alone is insufficient. You can have various techniques and software layers, but you have to have is a combination of the software working seamlessly with the hardware. We have with the AMD Secure Processor established across all of our designs going forward, our new designs that we're working on. We're embedding in each of them the same AMD Secure Processor which establishes a root of trust upon boot. Every access that you do from boot on is authenticated. It leverages a Cortex-A5 embedded controller. It leverages an AMD cryptographic engine. From a software standpoint, in partnership with Arm, it's leveraging TrustZone.
It's creating a software ecosystem that's then easy to build software applications that ride on that hardware root of trust that we established from the moment you boot up an AMD processor. It's a requirement going forward. You have to provide security. We do with the secure processor. Next, I want to talk about our programming model. This has really been a huge investment for us. I'll go back to three years ago when we rolled out the Heterogeneous System Architecture Foundation. AMD was a founding member. We invited other companies to join us. Over 40 did because we're committed to open platforms.
What we said is we wanted to unlock the value of CPUs and GPUs working together and CPUs and accelerators working together because it takes more than just the foundry node to keep pace with the Moore's Law rate of improvement on performance. You have to provide other acceleration methods, and that's what we're doing at AMD. We have delivered capability on track. If you look at the end of last year, the compiler for the Heterogeneous System Architecture was delivered. Our OpenCL 2.0 support for AMD products was delivered. Excellent progress, again, on an open approach with HSA to unlock the value. This is coherent memory access across CPU and GPU. This is context, preemptive switching, so you can immediately context switch between the CPU and GPU. What it means is performance.
You'll hear again over the course of today how we're realizing that performance in real-world applications. For those that are with us here in New York today, we have demonstrations of that capability in our products. Machine learning. I am very, very excited about machine learning or deep neural nets. Why? It's an inherently parallel operation, and it's actually proving to be a transformation effect on a number of industries. I've worked with some startups that were working on various algorithms, various optimizations, and they were running into basically dry wells. They were running out of the computing capability they needed to solve their difficult problems. What they've realized is, with machine learning, it uses an algorithm. It's a gradient descent analysis. It's an optimization algorithm. It's well-known in the technical community, and it's highly parallelizable.
When you take applications that can leverage machine learning, we can provide a very rapid speed up with our AMD technology. Of course, you have to provide the software enablement. What's out there today is widely deployed Torch7 and Caffe. Actually most of the implementations today are in proprietary code, and that's not what we're about. We have mapped Caffe and Torch7 to an open-source approach, putting it out there to run on our Radeon and our FirePro graphics, and we'll be releasing that through the open source. We'll have that available in the second half of this year. LiquidVR and all of the work that we're doing around virtual reality. This is a new category. When you look at what we've done, is we're leveraging capabilities that we have in Graphics Core Next, and we're building upon them.
I talked about asynchronous compute engine. It was actually in our first generation of Graphics Core Next when we released it in 2012. We are leveraging that in a way that's turned out to be essential for virtual reality. If you're in a virtual environment, you can't break the presence. If you have too long a delay, if you see a break in the image, you break presence. You have a very unsettling experience rather than a virtual immersive experience. The asynchronous compute engines provide a capability we call TimeWarp. As you traverse and you move around in a virtual environment, while we're rendering the space that you're looking at, we are actually, in parallel, calculating the images and doing a matrix shift as you shift, as you look elsewhere.
It's quite impactful, when you see this in the demonstration, you'll see it provides a very smooth experience in virtual reality workloads. There's more. We'll talk more later about some of the additional features that we have in virtual reality. Double down investment in software and our programming enablement. Okay. Last piece I want to talk about in terms of our platform enablement is our modularity. You can say, "Why is Mark Papermaster talking about modularity? Isn't this sort of just blocking and tackling kind of stuff?" It's actually not. It's very hard to do because it affects every piece of how you put a system on a chip together. It's a technological challenge, frankly, it's a cultural challenge because it drives every team to do their work in a different way than they have for years. Frankly, modularity has transformed every industry that's adopted it.
Look at the automotive industry. Look at Ford. Ford adopted modularity fundamentally changed their ability to get products to market and to leverage technology across their product lines. You can look at other industries and see that same impact of modularity. We've brought modularity to high-performance system on a chip, it is transformative because what we've done is actually put a protocol, a standard way of interconnect around each of our building blocks, each of our IPs. We leveraged our experience in coherent HyperTransport, we built upon it. We had a fabric, an interconnect that was proven, we built upon it to create a high-performance, scalable interconnect for our IP, that we put it together, third-party IP as well, then we've actually made it extensible off chip. This is a very capable AMD Network on Chip.
What it really allows us to do is have tremendous flexibility in how we put our solutions together. It can be Arm, it can be x86, it's ISA-agnostic in terms of how you put this together. Again, third-party IP. Very flexible we've implemented modularity into our entire methodology, our power management, our test. All of the elements that have typically been in centralized units, we've decentralized to, again, allow us to be much more agile in putting these solutions together. It's a quicker time to market, it's actually a higher quality at the end of the day, as you can leverage more reuse across your portfolio. That's what I want to talk to you about because this is how now we'll be putting our solutions together, is in this modular approach.
When you have a chip that we talk about, Forrest Norrod will talk about his business in the data center. Obviously, that's a very high demand on CPU technology, you configure with a very high CPU content on that system on a chip. Likewise, for client and embedded SoCs, many applications that leverage heterogeneity, they're going to have CPU cores and GPU cores in combination. Then the discrete graphics, where you load up as many GCN cores as you fit, given the target market that you're after. You need to be able to scale across the graphics stack, it takes several implementations, we can very rapidly attune our implementations for that market. What's common across all these, of course, is the software. It's all of our drivers.
It's all of the capability that we have that marries with each of the hardware blocks, and its accelerator capability and its security. Those are constant across those capabilities while we can change the other elements. What's particularly exciting is what this modularity means for our semi-custom business. As I said, we designed our AMD Network on Chip to be able to fold in not only third-party IP, but a customer's IP. It allows us to be more agile as well in our semi-custom business, and lower our time to market with our customers in this business. Modularity can be transformative, and that's what we're doing with this approach at AMD. Okay. We're out of time. I want to wrap up.
Look, I don't know how to say this other than, we were really clear on our goal with the Zen design, and it's getting right back into that competitive high performance CPU. You look at the market out there, it's a wide-open space as you heard from Lisa, in terms of bringing competition back to high performance x86. Graphics, we're going to continue to pedal to the metal. It's always been a fight. It'll continue to fight. We like our path. We've invested, we will deliver. Of course, that modularity that I just talked about, the flexibility that it brings, and the solutions that we put around it. In summary, what I'm going to tell you is, I'm an industry veteran, I've had the opportunity to work, and the pleasure of being able to work on a number of impactful products to the market.
It's been a great number of years I've been in the industry, I've never been more proud of the innovation and the tenacity of any team compared to the team I have here that we're working with in AMD. We have been focusing on reloading our war chest of technology, we're ready for the fight. Thank you very much. With that, Lisa, I'm going to turn it back to you.
All right. Well, it's hard to follow Mark with all of that great technology, what I'll try to do in this section is both Forrest and I are going to talk about the products. How do we take that technology into our product roadmap to get to profitable share gains? A lot has been said about the computing and graphics business. I said it earlier. The PC market is a big market. It's a good market. It goes through volatile swings from time to time, we're in one of those swings at the moment. That being the case, it is one of the markets that truly values computing and graphics technology, that is what we're investing on. What are the key things that we're trying to do with this business?
The overarching goal is profitable market share, as I said, Devinder will go through some of those drivers when he goes through his pieces. From a strategy standpoint, it's really focused on four areas. It is building great processors, okay? We can only do that with the technology that Mark is enabling and that Mark's team is building. Those new CPU cores are going to give us a step function improvement in our competitiveness going forward. Winning in graphics. We love graphics, we love gaming, winning those key inflection points going forward. Perhaps two areas that are slightly different. When you think about AMD, I want you to think about the designs that are uniquely designed to our IP. When you look at, for example, in game consoles, Sony and Microsoft designed their systems around our system-on-chip, and those are phenomenal systems.
In computing and graphics, sometimes we're part of second-party designs, so it'll be a design that same system, you can use our chip or our competitor's chip. Or there are places where we are uniquely differentiating. For example, the Apple Mac Pro system or the 5K display in the iMac. Those are areas where when people design to our technology, that fully leverages and monetizes AMD IP. We are absolutely going to lean into those areas when we talk about the computing and graphics business in terms of changing the game and how technology is utilized. The last point is around open platforms and industry innovation. Okay, we are not confused. The world is not all about us. We're one piece of it. We provide silicon, we provide solutions. We need the ecosystem partners to be there as well.
Our approach is going to be about open innovation, and that's why we talked about some of the things in Mark's presentation. Forrest will talk about it more, but it is about enabling the ecosystem like we've done in the past. Okay, the markets, again, these markets we talked about, gaming, immersive platforms, big markets, over $40 billion of TAM in these markets going forward. They greatly utilize our technology. What I'm going to do now is give you a preview of our 2015 products. These are not product launches. The product launches will actually be in the coming weeks at some of the industry events. I will give you a view of where the products are going and what are differentiating things in our products. Let me start first with graphics. Winning the gaming inflection points.
Leadership in gaming, leadership in visualization, this is really critical. We are focused in our 2015 product lineup on some of the key inflection points: DX12, 4K display resolutions and performance there, and virtual reality and everything it takes to enable virtual reality. In terms of our 2015 graphics lineup, today we are announcing our mobile product lineup. We are refreshing our entire mobile line with the M300 Series of graphics processors. These are appearing in our OEM systems today, and you'll see over the next couple of weeks announcements from our OEM partners. Very focused on improved power efficiency, very focused on DX12 enablement and leadership there, and also on dual graphics capability. These are out and in the marketplace. Coming later this quarter is a big change for us, introducing our new desktop GPUs in AIB.
We talked about picking up the inflection points, investing in technologies, knowing that they're going to make a huge difference in the marketplace, and then picking the right time to introduce them. That's always the important thing. We are going to introduce a full lineup in our AIB channel starting later this quarter, and it is going to focus on the key areas that I talked about: DX12, VR enablement, superior 4K gameplay. But the most exciting piece of it is what we're doing on the enthusiast side, and it is really the unveiling of our next-generation graphics technology using high-bandwidth memory. I'm not going to name it today because, again, I'm not doing a product launch. We will launch this later in the second quarter over the next few weeks. But this technology is really truly revolutionary.
When we think about what's important at this timeframe for enthusiast gamers, it's about enabling that next generation of experience. Okay? 4K displays are coming down in price points. They're enabling lots of people to have them. We want to lead in that area. What does high-bandwidth memory do? Mark gave you the speeds and feeds about the power efficiency and the memory bandwidth. But when it comes to something like this, it changes the game for how you can put this level of performance in this type of card. And if you think about it, with the memory now on die versus on card, it allows you to enable new form factors. We have a bunch of exciting form factors that we're going to talk about as we launch this product, but we think it will really bring a new level of performance to the industry.
This technology is really just at the starting point. We're going to talk about graphics today. There are also conversations about HBM2. But when you look across the industry, we have the computing and graphics components that can come together, and HBM can be utilized to bring up that capability. So, very excited about what we're doing there. I also want to talk about software and the inflection points around software. If you remember, a couple of years ago, as we started working with the game console manufacturers, we were amazed at how much performance they could get out of our graphics technology that the PC gamers were not experiencing. It was like night and day when you have programmers that are programming to your hardware. We decided we can bring some of those features to the PC gaming market. That's when we introduced the Mantle API.
The Mantle API was about making the hardware easier to use with a thinner software layer so that we could really unlock the performance for PC gamers. It was really an industry first in thinking for the PC market. Now, in the industry, others have adopted similar technologies: DX12, Metal, as well as Vulkan from the OpenGL group. We think this is really good. This is about enabling the ecosystem to solve new problems, and we are absolutely optimizing our products for the DX12 ecosystem. What's next? What's next for us is what's the next big problem that needs to be solved? It is the virtual reality problem. Mark started this, and I'll probably say it in a little bit more, I'll call it plain words than the techie words. What is the problem that we're trying to solve in VR? It really is about presence.
We are trying to create the new form factor. We're trying to create a new way of experiencing devices, and it's the idea of full virtual reality. When you want to experience a 3D image, you want to be able to keep up. It's all about the latency. If you can't keep up, and if you move and the image doesn't move, you'll actually get motion sick or something like that. This is probably the toughest problem to solve in graphics. It's the toughest problem to solve. When you think about rendering that much information over a 10-millisecond timeframe, that's a tough problem. Our engineers saw that this was a tough problem, and we said we want to develop a solution for it, and we call it LiquidVR by AMD.
We announced this at the Game Developers Conference in March. It's about helping software developers solve the problems. Our graphics cards and our compute capability will really be focused on that. Some of the features that we're looking at, Mark talked about some of these things: asynchronous shaders to minimize latency, developing technologies that really allow head tracking. It's hard for me to describe in a chart. It's really important to experience. This technology has advanced faster than anybody would have expected. The last six months has been an incredible advance in capability. I think the next 6, 12, 18, 24 months will be key to bring this forward, and AMD is investing heavily in this area. Next slide, please. Let me talk about computing and what we're doing in computing.
Later this quarter, we will also be announcing our 2015 computing lineup, and this is our AMD sixth-generation A-series processor that we call Carrizo. Carrizo is a step function improvement in terms of mobile capability. Mark talked about the fact that we were optimizing our roadmap for mobile with Excavator. In the same 28-nanometer technology, we got significant power reductions as well as performance improvements based on design. What does that mean for us? From a graphics side, 40% performance improvement over our competition. We will win handily on that. What have we also done? What are the applications that really need this type of technology? We're really looking at video streaming and how important that is when you look at the form factors. This will be the first APU in the industry that has hardware-assisted HEVC.
That will really improve our battery life and improve the smoothness of play. We have up to 2x the battery life here in this form factor. Carrizo is ramping very well. We started first volume shipments in the first quarter at a small level, significant volume shipments in the second quarter, and OEMs will be announcing systems so that they're ready for back to school this season. Very aligned to the overall PC market season. Commercial. Commercial is an exciting area for us. You might have seen on my earlier charts in the CEO deck, I talked about commercial as being a strategic area for profitable growth. We like the business. We like the fact that there's a longevity to the business. We started really focusing on this in terms of design wins and customer over the last, I would say, 12 months or so.
We've doubled our commercial client design wins. We've increased our pipeline. We've also adopted to things that are important in the commercial business in terms of security, in terms of image stability, in terms of platform longevity. We're also introducing Carrizo into our commercial lineup as well in the second half of this year. You're going to continue to hear more about our commercial successes. We're going to continue to invest in this area, and it's an important area for us going forward. Finally, workstation graphics. Workstation graphics, again, this is an area where I would say we have been traditionally underrepresented. We have lots of hardware, good hardware in this area. What we haven't done is spent as much time on the enablement. Some of the key wins powering the Apple Mac Pro, as I said, beautiful system.
We now have a very strong lineup with HP and Dell for their workstation offerings, and they are the strongest workstation players out there with FirePro. When you look at the ecosystem, and Forrest is going to talk more about it is about enabling the ecosystem. Enabling the ecosystem with the top ISVs that are servicing the workstation market. We've made progress in this area. We're going to continue to make progress going forward. That gives you an idea of our 2015 product focus for computing and graphics. Let me spend the last few minutes talking about future innovation and what we're doing in this business going forward. I did say earlier it is about the technology, and it is about the technology. Mark talked about the investments that we're making in cores, and cores on the CPU side and the graphics side.
What I'd like to do is talk to you about how we translate that into products. There are a couple of key areas that we're focusing when we look at the computing and graphics business. First one is very simple platform infrastructure. We want to make it easy for customers to adopt our technology on the OEM side and on the end-user side. I'll show you how we do that. The second is Zen, bringing Zen throughout our product portfolio. That is really critical for us to gain profitable share. The third point is perhaps an evolution of our thinking in terms of where to target products in this area. I'll talk about scalable platform design points. A little bit about our roadmap. Our roadmap in 2015, I talked about at the low end, we have Mullins APUs.
Our Carrizo lineup will be launched in the second quarter this year, and you'll see OEM systems in market. On the desktop side, we have Kaveri and Vishera as the two players on the desktop side. If you take a look at this roadmap, it's a strong roadmap, particularly in the mobile segment. It's a strong roadmap. It also has four different infrastructures going forward. When you look at 2016, you'll see that simplicity that we're looking for, how to make it easy for customers to develop with AMD. On the mobile side, we'll do our seventh generation A-series APU. It'll have the same platform infrastructure as our 2015 products, so OEMs will be able to seamlessly move to our next generation APUs. Very, very important at the high end, we are reinvesting in desktop FX processors at the high end with introducing the Zen core.
That's a big change for us. When we looked at the market and looked at the opportunities, high-performance compute makes a difference. High performance and compute in desktop makes a big difference. We will introduce Zen first at AMD in the desktop space. We will also have a new infrastructure with AM4 that will share both our APUs and CPUs so that that infrastructure can be shared across the board. That gives you an idea of how we're thinking about enabling our customers with our product portfolio. Just a little bit more detail, and I know you guys want a lot more detail, just a little bit more detail on our 2016. As I said, leading with our FX CPUs, it's going to be really exciting to be able to demonstrate Zen cores next year, and that's really critical for us.
They will be high core count. They will include multithreading. They will have the platform infrastructure supporting DDR4. As we go to the 7th generation APUs, the mobile market wants performance per watt. We have demonstrated performance per watt with Carrizo. We're going to continue to demonstrate performance per watt with our 7th generation APUs. As we move over to our high-performance GPUs, again, the high-bandwidth memory introduction in 2015 is very significant. However, it's not just about one data point. It's about what we do going forward. Introducing FinFET process technology on the GPUs will give us substantial performance per watt improvements in addition to the architectural improvements that Mark talked about. Ensuring that we get 2nd generation HBM. We've now gotten experience with the 1st generation, really extending that to higher bandwidth and more performance.
With the VR and AR acceleration that we're talking about. This gives you an idea of the portfolio that we will have in 2016 going forward for computing and graphics. I want to give you a little bit of how we're thinking about the market going forward and how we think about this business. Again, computing and graphics is a critical piece of the AMD business. Why is it critical? All the technology that we talked about that Mark is developing, it services both the computing and graphics business as well as the Enterprise, Embedded and Semi-Custom business. It's actually a very synergistic product portfolio. If you look at what we've done over the last five years and some of the decisions that were made, we focused very heavily on mobile, and especially the mobile notebooks and driving the APU roadmap.
We made good progress. As I said, APUs are now everywhere. They're in PCs, they're in phones, they're in tablets, they're everywhere. This is one vector of where innovation is going, though. If you look at the vectors around computing innovation and graphics innovation, that entire IoT picture is going to need lots of different combinations. There are going to be some vectors that are going to need more compute performance. We need to have an answer to that. Our answer is the FX CPU with compute leadership. There are certain vectors that are going to want more graphics, gaming, or GPU compute performance. We talked about our next-generation technology we're launching in 2015. We're going to extend that leadership into 2016. When you think about heterogeneous computing leadership, it is about APUs that are monolithic.
It is also about putting together larger scale APUs that Forrest will talk about. We will have, with this technology, vectors to address much more competitively all of these areas of the market. That is what really makes me excited about the computing and graphics business for AMD. Let me finish with a couple of things. There is no question that we need to understand the market backdrop, the market backdrop will continue to be something that we look at very closely. We are proud of our product lineup in 2015. As I said, the next four, five, six weeks, you will see a lot of information from AMD, details about our products, details about our customers and our ecosystem. We are excited about that. I think leading in the key technology inflection points, making sure that we are really monetizing our IP and monetizing our platform investment, very clear.
We are going to focus on profitable share growth. That means that it will be a little bit bumpy at times. We have lost some share. I totally recognize that. We will gain that share back in more profitable segments. Our future innovation is really enabled by the roadmap that Mark talked about. It is an exciting roadmap. It really is just the beginning, we have many more aspects of it that we will share over the coming quarters and the coming years. With that, I think it is actually time for us to take a break. For those in the room, we have about a 15-minute break with some refreshments in the back. For those online, we will be back shortly. Thank you very much.
Good afternoon. Welcome back after the break. Earlier today, Mark and Lisa painted, I think, a compelling picture of the technology investments that AMD is making and the way that we believe they are going to drive our client graphics business going forward. I would like to take the next 25 minutes or so to talk to you about the Enterprise Embedded in Semi-Customs business, which has been a principal driver of growth for the last few years, and one that we think is central to the growth story of AMD going forward. In introducing EESC, I first have to start correcting a misperception. A misperception that quite frankly, we had even within the company just a year or two ago. The Enterprise, Embedded and Semi-Custom businesses are not really three discrete businesses. They are a set of businesses that share a great many things in common.
They share technology. They share, in many cases, customers. Our embedded customers are many times also our semi-custom customers. They share a perspective around the best way to showcase AMD technology, which is in all of these businesses, our customers are building products around the technology ingredients that we give to them and bringing differentiated solutions to the end customer that leverage AMD unique IP. We really think now at AMD of EESC as being a continuum of offerings, leveraging technology, customer relationships, and the modular design approach that Mark talked about earlier at both the chip as well as the systems level. Looking forward, the central tenets of these businesses are common and ones that we are focused on driving over the next several years.
The first one being that we are focused at AMD on addressing vertical markets where our unique high-performance compute and graphics technology allow our customers to build those differentiated end-user experiences and systems. We're going to enable them to use those technologies by providing them vertical development kits, by providing them software, platforms where appropriate, and an open ecosystem of other elements from partners and the rest of the travelers in the technology industry to allow our customers to quickly embrace our technology. Very important in that approach is that word open. Mark talked about our commitment to openness in our technology, and particularly around our software. We think that's so important because in today's markets, one of the things that really drives innovation is the ability to have open ecosystems. We are completely committed and double downing on our commitment to openness.
Lastly, Lisa also mentioned the semi-customs business, not necessarily just as a business, but as a business model. I think that's absolutely true and something we've really recognized over the past year. We bring to this set of customers that continuum of standard products, lightly customized products, and full custom products, enabled by the technology that Mark and the teams have developed to rapidly assemble custom chips, and supported by that ability to do so with standard products and custom products sharing the same drivers, the same technology, but optimized for different points of scale. Let me talk about each one of the market segments in turn. We're going to talk in the same framework that Lisa introduced earlier in terms of gaming, in terms of immersive devices, and in terms of data center.
The one that is most familiar, I think, to people about the EESC business over the last couple of years is certainly gaming. Our flagship wins with all of the console makers in this latest generation of game consoles has been a huge driver of revenue for AMD. The reason we won is because our technology allows us to give the users that compelling visual experience that they need, that they want, that draws them in to the game and into the product. That compelling visual experience is a product of that combination of CPU, GPU, and software capability. That combination is what allowed customers like Sony to provide the compelling experience of the PS4.
If I cast myself back into Mark's role for just a moment and talk like a geek, I could talk about the nearly two teraflops of computing power that our graphics units provide. I could talk about the eight x86 cores providing the oomph to run the game logic and the physics of the game. I could talk about the audio and video capability that we've added and the IP that we've created to make that a compelling experience. I could talk about the memory innovation that even in the current console generation, before we embrace HBM, has led to such a seamless and fluid experience in these consoles. I could talk about all of that. The end result is what's important. The end result is that compelling experience.
We are going to take the technology that we've created for this console generation, that CPU performance, that cutting-edge graphics, and very importantly, the fact that we have now developed a set of game developers across the industry who understand and are optimizing for the AMD graphics architecture to showcase the capabilities of these game consoles, to showcase the capabilities of the GCN technology in PCs as well. We've got a great asset with a set of developers that now know AMD. We take those technologies and those assets, we deliver them in a set of embedded CPUs and APUs, discrete graphics, and that capability to do a full custom, or we call it semi-custom, but let's face facts, it's full custom design where necessary, and we're attacking now new markets. The first one that we really pursued was casino gaming.
Casino gaming is something we really got underway a few years ago and has really borne fruit. Lisa talked about the number ones in our business, and I'm proud to say we're number one in casino gaming worldwide, and we're growing our share rapidly in that market, both in traditional, what you guys think of as casino gaming, as well as pachinko and pachislo in Asia. We're expanding the aperture. We now have significant design wins in cloud gaming. We're addressing military, police, and aerospace simulation. We think about that, by the way, as gaming, because in both gaming as well as simulation, you're trying to create an immersive world and pull the user in. Lastly, we do think that our technology, our core IP, is appropriate. With the memories, innovations, and the power efficiency is applicable across all form factors.
We'd like to say from handhelds to holodecks. Gaming is going to remain a key driver of the growth in the EESC business and for AMD overall going forward. Immersive platforms are the next key area for EESC. On immersive platforms, we're really talking about platforms that are rapidly moving from just having a compelling front-of-screen experience, great visualization of medical imaging, great high-performance graphics for thin client applications, be they on the desktop or be they served remotely from the data center, to one where people are going to have truly two-way interaction with their device. Lisa talked about IoT user interface evolution, and we think there are compelling evolutions coming, revolutions coming.
When your immersive device can look at you just as you're looking at it can discern what you're doing and your intent, and then it can act upon it without you having to go through the rigmarole of tapping something into a keyboard, drawing a finger across a tablet, or holding a game controller. Let's talk about that a little bit more. Next slide, please. That first part of immersive experience is that front of the screen experience, is pulling a compelling visual experience out to the user. If you look on Times Square at the largest, which is where we are, for those of you that are watching online, we're just off of Times Square. Which, if you walk it, is one of the most amazing places in the world. Everywhere you look, there's visually rich experiences, incredible imagery coming at you from digital signs nowadays.
The largest one, a block and a half north of here, is one of the largest screens in the world, powered by embedded Radeon graphics. We've focused on providing advanced digital signage solutions for the last few years. We're growing our market share rapidly there. That is coupled with our leadership and number one market share in thin client, where we continue to grow our share and enjoy the growth of thin client devices as they become more and more prevalent in commercial installations around the world. What's fascinating to me is we're taking the APU technology that we've innovated in support of our mobile devices. We're bringing that to medical imaging. We're bringing that to medical imaging to allow advanced diagnostic instruments, ultrasound, portable X-ray machines, to have a set of capabilities that hitherto had been confined to very large fixed installations.
Now with the power of the CPUs and the GPUs integrated together, you can bring that into small form factor medical diagnostic devices. We think that's going to be very interesting to see how they evolve over time. Again, all of those enabled by our vertical development kits to allow our customers to more rapidly innovate. When we look beyond this one-way experience into the two-way experience, I said a moment ago that the system is going to look at you, and I mean that literally. Many of you may be aware that over the last two to three years, there's been a fundamental set of breakthroughs in the field of computer vision, machine learning, and machine understanding, where the ability of computers to recognize faces, the ability of computers to recognize what objects and actors in a scene are doing, has increased by an incredible degree.
That revolution, those advances have been made possible by the incredible power of the vector compute engines inside of our graphics units. They mean that in the future, that our machines will be able to look out at us, recognize us, see as we walk up to a digital sign, is the person walking up a fashionista or a tourist? Tailor the offering appropriately. They'll be able to see us. They'll also be able to see in augmented reality situations what's going on in the world. Then combine that with the analytic data that will make that a much richer experience. One example of that is, say, an engineer working at a car factory, watching as a car frame is being assembled and looking at the stresses and any diagnostic information on the materials as the frame is being assembled.
In order to do that, you need several things. You need the horsepower in the back end to run that analysis, to do that finite element analysis to say, "Hey, what the heck is going on?" You also need the horsepower to understand. The computer to see, oh, there's the car. I don't want to fix the engineer in one space so that they-- In that case, I know exactly where the car is at. So I can just assume and overlay the data. No, I want the engineer to be able to walk around, to be able to inspect from any angle. That requires that machine understanding. The horsepower to understand is something that AMD is one of the very few companies can provide in the industry.
It's that graphics, that GPU-powered compute engine that is so important for enabling these new machine learning algorithms. When you combine that with the HSA, Heterogeneous Systems Architecture, software and platform innovation that Mark talked about, you have an open innovation framework that allows users to harness the power, developers to harness the power of CPUs and GPUs and to use them much easily than they ever had in the past. We're going to couple that technology to a delivery mechanism where we can deliver embedded APUs for CPUs, GPUs, or semi-custom. We're going to fire, like a laser, an intelligent signage. Not just advanced signage, intelligent signage. We're going to fire like a laser at augmented reality for commercial applications. We're going to fire at advanced medical diagnostics that not only present a picture to the clinician, but they also provide analysis.
Hey, this might be a tumor. This might be a break. Look here. Of course, for industrial heads-up display or automotive cockpits, where self-driving cars need this machine learning capability in order to understand and interpret the world, AMD is really one of only two companies in the world that have the technology to power that fundamental engine. We're hugely excited about the wins we've already had in immersive platforms, but what we see going forward. I tell you, I joined AMD six months ago, and the reason I joined AMD was not because of the exciting opportunities we have on gaming and immersion. I didn't fully recognize and appreciate those at the time. Now I'm pretty psyched by them. I'm even more excited at the opportunity that AMD has in the data center.
The opportunity that AMD has to get back into the enterprise infrastructure and to really do so at a time when enterprise infrastructure is undergoing a number of significant changes. The first one is, look, everybody in the room has heard about software-defined networking, software-defined storage, software-defined this, software-defined that. I like the one wit who came up with, we should just have software-defined software and close the loop. Really, it's software-defined everything. Proprietary hardware is dying across the enterprise, not just in the traditional data center, but in telcos, in just about every field of infrastructure. The server, one. As that transition is occurring, that means that performance demands are unbounded. The companies that are going to win are the ones that can provide the right solutions to meet those incredibly difficult performance demands.
AMD is one of the few companies that has demonstrated the ability to service this market. We're building upon our heritage of innovation in the data center when we introduced the first 64-bit x86. We introduced the first hardware virtualization support. We introduced the first multi-core CPUs. We're introducing the first APU server relevant technology. We're going to take that heritage, we're going to take the technology that Mark and team have created, we're going to couple it with an approach that, number 1, provides the ecosystem innovation and choice. Seven months ago, I was a customer of AMD's, I can assure you that as a customer in the enterprise space, the need for a choice to drive competition, to drive innovation, and to drive the economics of the market has never been more dire.
AMD is going to get back into the x86 server market and provide that choice. We're going to do so by offering our customers a choice of technologies as well. x86 is obvious, Mark and Lisa already talked about that. Arm, we're going to offer as well. Arm will be important over time in the general purpose x86 server market. In the intermediate term, perhaps the more important and more interesting opportunity is to capture some of that enterprise infrastructure that's moving to the server base. In many cases, the embedded processors inside of that proprietary networking or storage silicon are MIPS or PowerPC. Those architectures are dead.
There's an inflection point that customers are already going through where they need to migrate their software base, even as they're extending it to take advantage of the performance that you can run on a general purpose server. Arm is hugely important for some of those markets, we think that much of that volume will be captured by Arm-based server infrastructure. The ingredients that we have already been revealed to you somewhat by Mark. Let me turn up the contrast one click. High performance x86 and Arm CPUs that have the performance level that are competitive with the highest performance level in the industry are absolutely critical for playing in this market. We've made the multi-generational commitment to introduce those cores and to maintain that core roadmap over time. Our GPUs are going to become more and more important to the enterprise infrastructure.
Machine learning, machine understanding, computer vision, all are going to be incredibly high growth areas for GPU acceleration, as well as the more traditional HPC applications of GPU that we already know. We're going to continue the memory innovation that AMD has introduced in the past. We were the first people to introduce the memory controller onto the CPU. We're the first now to introduce HBM to the graphics realm. You'll see us continue to innovate and provide more innovative and higher bandwidth solutions than our competitors. We're going to be open. Not just open on the software side. We're going to be very open for partnering and innovation on the hardware side as well with other folks in the silicon ecosystem.
That's something where you see more and more some of the competitive ecosystems are trying to close off, which I think is antithetical to the fundamental tenets that have driven this industry over the last 25 to 30 years. The opportunity, obviously, is to offer choice in that $10 billion-plus server processor market, that $4 billion-plus networking processor market, and that $2 billion-plus storage processor market. To lead in machine learning, particularly by taking that open approach to enable more rapid innovation across the ecosystem, also to take advantage of the fact that much visualization and workstation deployments are moving into the data center. We're uniquely equipped to take advantage of that as well. Short term, in 2015, you're going to see a couple of things from us.
First off, you're going to see us introduce the Opteron A1100, which you may know better as the chip called Seattle, code name Seattle. It is a true enterprise class chip powered by Arm A57 cores. It has all the redundancy, the reliability, the manageability, and the deployment features that people expect and demand before something goes into their data center. We've been working very closely with the ecosystem to get ready for Arm in the enterprise for quite some time. We first sampled this chip in the first half of last year, and we've been working very closely with our ecosystem partners to get the software, both from the operating systems, the drivers, the tool chains, the developer kits ready, as well as to get the right development platforms and end systems utilizing this technology ready to go.
You'll see us in the second half of the year, along with our ecosystem partners, introduce really the first true enterprise class Arm server powered infrastructure in the second half of this year. The ecosystem is finally ready to go. You'll also see us actually double down. I say continue, we've had the GPU technology for some time, you're going to see us double down on GPU in 2015 and beyond. We've got some great chips already. The S9150 FirePro card has 33% more memory than the highest memory capacity NVIDIA card, and its baby brother, which is coming, will take that lead up quite a bit. We have 77% more double precision floating point performance than the highest end NVIDIA card to date.
If you look on the supercomputing list for the Green500, number one spot since November of last year has been and continues to be an AMD FirePro-powered supercomputer that has the highest energy efficiency and performance on that list. One area we're going to double down is that open area. Again, I don't want to overly belabor the point, but I'm going to overly belabor the point. We think that innovation in the enterprise nowadays, particularly in these new areas, is fostered and accelerated through open source communities and open collaborative innovation. We're going all in. Open source drivers, open source libraries, open communities. Where we do have additional new features, we're going to introduce them in a standards-based way. So GPU virtualization, for example, we could have taken a proprietary approach.
Instead, we're providing the industry's only standard SR-IOV virtualization support and bringing that into GPUs. Looking forward beyond 2015, next slide, please. Looking forward beyond 2015, we've got an even richer roadmap. The first is, of course, the next generation Opteron, the chip using that Zen core. No, I'm sorry, I said that wrong. The chip using many of those Zen cores to deliver a level of performance and competition to the x86 server market. High core count with full multithreading, disruptive memory technology and memory bandwidth to provide extremely high performance, high native IO capability. What do I mean by those last two? I mean we're not just going to produce high performance cores. We're going to do the same thing that AMD has done whenever it successfully disrupted markets and made an impact. We're going to innovate. We're going to differentiate.
We're not going to just do the same thing. We're going to provide differentiation that allows our customers to build demonstrably superior systems for particular workloads and particular tasks. The opportunity for that has never been better. I've been in the server industry for a long time. I can tell you it's becoming the car industry. 15 years ago, you maybe needed a sedan and a minivan to service the needs of 99% of the x86 server market. That's no longer true. The new workloads, the new applications, the new scale deployments mean it's never been more important to have different points of optimization, different types of systems, and that gives us great opportunities to differentiate and provide unique solutions. We're going to bring out Arm into the data center, and using K12 offer the highest performance Arm server CPUs to the market. It will be true enterprise-class devices.
Again, we're the only guys in the Arm ecosystem that have experience delivering and supporting enterprise-class server CPUs for an extended period of time. We're going to bring that APU concept fully into the server realm, have high-performance server APUs offering not just high performance few cores and memory, but multi-teraflop GPU capability, providing a level of performance for machine learning, a level of performance for finite element analysis, a level of performance and memory bandwidth for the reverse time migration algorithms that the oil companies use to do reservoir simulation. We think we've got unique and compelling technology that only is possible by wedding together that CPU, that world-class GPU, and combining them with an open standard HSA software interface. Across all of these, we're going to have security. Mark talked about one aspect of that security, which is our Arm TrustZone processor and infrastructure.
We have many other innovations as well, some leverage, ironically enough, from our game console technology. Finally, we've got that semi-custom capability, that semi-custom business model that allows us to much more rapidly assemble the right standard product for that bifurcating server market, that car-like industry, but also provides us an easy path to optimize for cloud service providers or for large customers of all type that may need or may benefit greatly from a fully customized part that incorporates our IP and theirs. Again, that's a capability we've demonstrated now, and that's a capability we're going to bring to the data center market. If you put it all together. Next slide, please. If you put it all together and look forward, over the next few years in the data center, the opportunity for AMD is tremendous.
I am incredibly excited, and my team is energized by what lies ahead. We're going to re-enter the server market, providing choice and competition to our customers, to the ecosystem, and providing innovation to help accelerate the market. We're going to drive GPU and APU technology into the mainstream, lower the barrier for developers to use GPU acceleration, and make it much more accessible and enable rapid innovation through that open and partner-friendly approach. Like I said, the data center opportunity, and the ability, the opportunity to disrupt there is why I came to AMD in the first place. You add to that, in my set of businesses, the opportunity to continue to expand our gaming franchise and to help lead and innovate as devices truly get immersive and two-way.
I think EESC will not only have been, but will continue to be the principal growth driver for AMD going forward. Thank you all very much. With that, I'll turn it over to Devinder.
Good afternoon, also to those on the webcast. Today is really a day of firsts. This is our first time doing a financial analyst day from the Nasdaq, where we started listing on this stock exchange earlier this year, on January 2nd, to be specific. The other thing is it's the first time since 2012 AMD is having a financial analyst day. For many of you, I know you've been waiting for a long three years. We're going to lay out for the first time in that time period a long-term financial target model that I'll get into in a few minutes. Having heard from Lisa, Mark, and Forrest, I'm sure the question in your mind is: how does this all add up financially? How does this all add up financially?
What I'm going to do is, as Lisa said earlier, talk about 2015, because reading what is in the press these days and some of you writing about AMD, I know there's a lot of concern about 2015. We'll lay out 2015, especially the second half, then get to the longer term in terms of what we think we could achieve with everything that you have heard this afternoon. Let's get started. The things I'm going to cover today is PC challenges have been there since the PC market shifted pretty dramatically in the second half of 2012.
The fiscal management that we have had over the last few years has been challenging, no doubt, I'll cover that, I'll go back a little bit in the story that you heard in terms of the current to show you what we have done and how we have managed and what we can do if, for whatever reason, the PC market continues to be challenging. The second thing is the things that we did in 2014 in particular, what financial flexibility we have created for ourselves as we got ready for what is ahead of us for the next few years. The 2015 financial expectations, in particular the second half, the capital structure and liquidity profile from my standpoint over the next few years, as I mentioned earlier, the long-term financial model.
As Lisa said earlier, to tell a story, you have to know where you are in the current, I want to go back a little bit. I want to go back because what we went through in 2012 and what we're going through right now, there are some similarities. The PC market continues to be challenging, AMD is still here after three years, getting ready for the fight for the future. If you were at AMD's Financial Analyst Day in 2012, you would have heard 90% of the revenues is from PCs. $600 million of OpEx on a quarterly basis. High near-term debt maturities coming due in 2015 and 2017. Complex R&D processes, you heard Mark talk about how he has significantly changed the R&D processes over the last few years. Finally, inconsistent profitability.
If you go back and look at the revenue, what did we do? The first thing we did, actually, we started it even before 2012, we needed to diversify AMD's revenue, we were able to do that very successfully by having about 40% of our revenue coming from the Embedded, Enterprise and Semi-Custom segment that you just heard Forrest talked about. That growth businesses, the Enterprise, the Embedded in particular, and the game console semi-custom businesses, it was driven us to that transition. On the OpEx side, you've seen us taking actions. I talked about $600 million per quarter in the early part of 2012. We did a lot to bring the OpEx structure down from an affordability standpoint and also to reduce the cash burn. We've taken it down in 2014 alone to the $350 million level.
That's a 36% reduction in OpEx from Q1 of 2012. How did we do that? We had, obviously, a lot of restructuring actions. We took real estate facilities-related actions. In the PC space, when revenue goes down, there is significant amount of marketing dollars that are deployed in winning PC business. We chose not to pursue some business that was not profitable, didn't make sense for us, especially with the long-term roadmap, therefore, we were able to bring down our sales and marketing dollars and continued with SG&A efficiencies. In process technology, in particular with the R&D model that Mark implemented, we were able to reduce R&D costs.
On top of that, with the semi-custom model that we've implemented, we are deriving today, it doesn't show up on the R&D line when you look at AMD's P&L, we are getting about $100 million plus R&D funding in the semi-custom NRE business model that we have today that helps us obviously do some of the things in our semi-custom space. Managing through the PC challenges, one of the things we had to do in that timeframe was first half of 2012, we made money. PC market shifted, for the next 12 months, four quarters in a row, we lost money. We managed our way through that with some of the actions that I talked about earlier, really found a path to profitability, in the end, were profitable in 2014.
This allowed us in terms of taking advantage of the situation in terms of the opportunities as something like this presented. One of the first things we did in the 2014 timeframe, you go back and look at the 2013 timeframe, the debt profile of AMD, this is what it looked like. We had $1.1 billion of near-term debt maturities coming due, as I said, in 2015 and 2017. $1.1 billion. We were losing money in the tail end of 2013 until we got to the 2014 timeframe. The action we took in the 2014 timeframe with the debt transactions, as I stand here today, having paid off $42 million of convertible debt that was due on May 1st, just a few days ago. We have no term debt maturities for the next four years. No term debt maturities.
That allows us from a viewpoint of the flexibility to go pursue the transition and what's going on with the businesses that Lisa and Mark and Forrest talked about. The other thing we did is we were able to bring down our weighted average interest cost. It was about 8.4% before the debt transactions and 7.1% after that. 130 basis point reduction that allowed our interest expenses to go down on an annualized basis by about $16 million. Finally, we put in place a cash profile to manage the business. For the last five quarters, if you have seen AMD from a viewpoint of the cash, we have a target of cash from $600 million-$1 billion. That's our target zone. We have managed the cash closer to the higher end of the optimal range.
On top of that, we've put in place an ABL facility to the tune of $500 million. We have this facility available to give us flexibility when we need it. In fact, just a few days ago, actually last month, we renewed it. We renewed it for a full five-year term all the way to April 2020. The last thing I'll mention on cash, which comes up a lot, is 90% of AMD's cash is located in the U.S. Switching gears, I'll go to the 2015 financial outlook. The financial outlook for the first half, if you take what we did on an actual basis in Q1 of 2015, and you add to that the midpoint of guidance that we provided on our earnings call in April, that's what the first half is.
About $2 billion of revenue, gross margin to the tune of 32%, operating expenses slightly north of $700 million. We had net income, or actually a loss, as if you add up the numbers, you get a loss in the first half. Cash, Lisa said earlier, first quarter, we ended up at about $900-plus million. We think we end up at about $800 million at the end of the first half, which is Q2. The free cash flow is obviously negative. In the second half, we are expecting revenue to grow 15% ±3%. Gross margins should improve, especially with the newer products that we're introducing in the next few weeks, as Lisa mentioned, and that should help with the gross margin profile. Also the second half, semi-custom business, PC business is always better from a second or first half.
We are expecting that's why you see the uplift in the revenue. Net income-wise, we expect to be profitable. With those numbers, we expect from where we end up in Q2 of 2015, that cash should grow, and we should be again at the higher end of the optimal zone between $800 and $1 billion and be positive on a free cash flow standpoint. The question on cash comes up a lot. Let me address that. We have run different models for 2015. We have run longer-term models from a cash standpoint. Any models that I've looked at recently in terms of revenue up or down, if revenue is down 10% and even up 20%, we think we can manage between the $600 and $1 billion of cash levels.
That's something that I know is top of the minds in many of you who follow AMD. We've done a lot of models, that's the way it comes out, every model, every sensitivity that we run. 2015 focus, if you were to summarize it, our goal is to stabilize the PC business, invest in everything that Forrest talked about in the enterprise embedded semi-custom space. Continue the OpEx discipline, which frankly, we have gotten very good at that with some of the actions we have taken, taking down the OpEx from $600 to $350. Very important that we fund the roadmap that Mark and Lisa and Forrest talked about today. We want to fund that roadmap because that's the future.
It's very easy to go ahead and sacrifice short-term, cut your way to profitability, and not be able to fund the future roadmap, that's not what this is all about. The other thing is managing the cash and liquidity, which is something that we have done over the last several quarters, going back to 2012 when the PC market shifted. Then most importantly, laying the foundation for growth and profitability. Let me switch gears here and get to the long-term outlook. Let me cover the capital structure and liquidity first. On the liquidity side, we believe we can fully manage cash at the higher end of the optimal zone. We can manage cash at the higher end of the optimal zone. The optimal zone is $600 to $1 billion.
As we have managed for the last few quarters in the last couple of years, we feel we can do that with our long-term target. Flexibility with the fine demand laws ABL, I talked about earlier, fully available up to $500 million if we needed all the way to April 2020. On the debt side, when opportunities arise, we'll continue to de-risk the maturity profile like we did in the 2013 and 2014 timeframe. We want to reduce interest expense. That's a drain on the P&L. If opportunities arise to do that, we'll definitely go ahead and do that. Then on the capital structure, for the excess cash that we get over $1 billion on the balance sheet, the expectation is go ahead and deploy that to reduce the debt.
Because having more than $2 billion of debt on the balance sheet, from my standpoint, is still a lot. By doing that, generating the cash, reducing the debt, our goal is to get net neutral net debt position, get to a neutral net debt position. Today is $2.2 billion, $900 million of cash, we are fully cognizant of the work we have to do, we are fully getting ready to do the work with all of the things you heard earlier today from an AMD long-term standpoint. Finally, from a leverage standpoint, to get our target optimal leverage ratio to about 2 times. Today is significantly higher than that, to work that down, in particular, with all of the actions that I talk about on this slide. Can you get me some water, please?
Before I get to the long-term model, let me cover some of the business drivers for computing and graphics and the enterprise. Thank you, Ruth. In the Enterprise, Embedded, and Semi-Custom space. Going from left to right, if you go to PCs, in the PC space, we've been very consistent the last couple of years. We see the PCs, although it's a significant market, Lisa talked about 300 million units PC market. It's big, it's important. It's very important for revenue base, we see that from our standpoint for our model, flatter down over the next few years, the PC part of it. On the graphics side, with some of the new products you heard about today, mid-single-digit percentage growth on the revenue standpoint for the GPUs.
Professional graphics, server, and embedded with all the exciting stuff, I don't use the word exciting very often, all the exciting stuff you heard from Mark and Forrest and Lisa earlier today, we think we can get double-digit percentage growth in those areas. Obviously, the big one there is the data center with the server part that you heard about today. Then semi-custom, we've proven the concept. We've got a good game console business. There is a pipeline of design opportunities. We talked about even announcing new opportunities or signing one within this quarter, that gets better for us as the semi-custom model matures and more customers show interest. Because when you have done it and proved it for a high volume, high ramping product like the game console, others are interested in at least understanding or engaging with AMD in the semi-custom model.
We think from the base that we have today, we can continue to grow that in a mid-single-digit percentage growth. From a financial standpoint, what you're probably more interested in is how does that translate to the operating margins? How does that translate to the operating margins? The Computing and Graphics segment today loses money. It loses money. If you look at the financials, except for going back to Q1 of 2014, it lost money. What you want to know is, even if the revenue is down, can we get back with the model that we have to have profit in the business? We think we can. The profitability for the PCs and GPU space, we think is in the mid-single-digit percentage. Then for the rest of the businesses, at least double-digit percentage, which would be at least 10%. At least 10%.
That's the way we laid out. The thing that Lisa said earlier, a couple of times earlier, I want you to pay attention to, is we are focused on profitable growth. We do not want to chase revenue share or unit share where it doesn't make sense. It only makes sense if it allows you to be profitable. If you can grow the business from a profitable standpoint, it makes even more sense, and that's where you want to grow. AMD's revenue, no doubt, you add the numbers in 2015, if you're doing the math from 2014 to 2015, it's a revenue down year compared to 2014.
From where we sit today and looking out for the next three to five years, these are our assumptions in terms of the assumption for revenue and the growth drivers for revenue and for the margins, and in particular, more importantly, operating margins, because in the semi-custom space, as all of you know, the gross margin in the semi-custom space, because of the NRE funding up front, is lower than corporate average. The operating margin is where it gets made up, because in the semi-custom space, by the time the product is introduced, you hardly have any R&D left. You don't have to spend in sales and marketing dollars. In fact, in the game consoles that are out there from a Sony and Microsoft standpoint, we don't spend marketing dollars. We don't need to spend any marketing dollars. The marketing, in some sense, is done for us.
What I'll get to next is the long-term targets. In fact, somebody told me that showing this slide, because you're showing it for the first time in three years, a lot of people are going to be waiting to see this slide because they've waited for three long years. This is kind of how it lays out for us. Gross margin with the products we are introducing, with the revenue mix that you will see from the various businesses that we have, and the products that we have in the space, in particular in the data center space, in server, embedded, and pro graphics, we think the gross margin on a long-term basis is somewhere in the 36%-40% range. On the OpEx-revenue ratio, this is something we are good at controlling. We're good at managing that. We've done that.
Anything, the OpEx falls between 26% and 30%. In particular, we will continue to fund the R&D roadmap while staying lean in the SG&A area. SG&A area will be lean, but R&D roadmap, where necessary, even for the longer term future, we want to fund it, but in all cases, maintaining the OpEx ratio within these targets to be somewhere in the 26%-30% range. Operating margin, as I said earlier, greater than 10%. The margin-accretive businesses that we have is really going to fuel that. The margin-accretive business that you heard Forrest and Lisa talked about and the products that we're getting to will fuel that.
In the semi-custom space, as we mature the business, as we diversify across beyond the game consoles into other businesses, we think we can get better operating margins, and that also helps the operating margin of this particular target that you see here. Bottom line, as Lisa showed earlier, and that was really the punchline as we were discussing the presentations when we were putting this together, greater than $0.50 EPS is the target. To get to greater than $0.50 EPS. If we do that and we are focused on everything we talked about from a product, revenue, gross margin, operating margin standpoint, and focusing on the bottom line for profitable growth, we think there is a free cash flow model that yields free cash flow in the first place.
From a modeling standpoint, you can assume that the free cash flow of the company is approximately equal to AMD's net income. That's the models. The net income, essentially, from an operating margin improvement, the OpEx leverage fall-through is important. Then we will continue to stay focused on the financial execution as we have for the last three years. I always kid when I talk internally at AMD, I took on the CFO role at AMD in probably one of the most difficult times in the history of the company. October of 2012, I took on the CFO role. It was a difficult quarter. I showed you the profile in 2012. We have managed the last three years, and we are fully ready, willing, and able to manage the next three to five years as we deliver the long-term financial targets that you see here.
In summary, from my standpoint, it's really about sharpening not just the financial, but strategic execution at AMD. The drivers or the tenets for success are exactly on this page. If you pull it all together in terms of everything you have heard, in terms of the technology, the vision, the roadmap, and the products, driving growth opportunities, gaming, immersive platforms, and data center is the key. Market share gains when it makes sense and we feel that we can have margin improvement for profitable growth. That's important. The second thing is continuing the expense control discipline. Continuing the expense control discipline that we have implemented over the last few years and making sure that we fund the future on the engineering investments for the long term, and also achieving consistent return on investments on IP investments. The IP investments that we make have to return investments.
If it doesn't, we'll make a different decision. For example, in the SeaMicro systems business, as Lisa talked about earlier, we got out of it. It didn't make sense, at least for the time frame, and from the system standpoint, we said we can't get an ROI, so go ahead and get out of the business. That's what we did to make sure that we can get an ROI from an investment standpoint. The first two things are important, then we want to bolster the capital structure. A lot of you have written about the debt load on AMD. I'm very cognizant of that. Having a lot more debt than cash is not a good thing. We need to go bolster the capital structure. Manage the working capital is something we've done well. The last few years, actually, we've gotten good at it.
We managed the inventories, we managed the working capital, and we definitely need to reduce the leverage from a viewpoint of the debt on the company. Finally, in all cases, not just for the company as a whole, but for every business that we participate in, we want to manage for profitable growth. In the end, by doing all this, from AMD's standpoint, is generating consistent free cash flow. With that, I'm going to round out here. I think you've heard a lot today, and we have, as Ruth said earlier, an extensive Q&A session. Looking forward to the questions that you might have that Lisa, Forrest, and Mark, and myself can answer in the next session. Thank you.
Okay. If you'd like to join us here on stage and we can kick off. Great. Okay. Chris here and Devinder. One second, we'll get you a microphone.
Chris Rolland, FBR. Perhaps talk about the movement to 14 FinFET timing. We know it's 2016, but maybe a little bit more. What products are going to migrate there, and what are effects on margins? Thanks.
Okay. Maybe, Chris, I'll let Mark start with the technology question, and then I'll follow up on the others.
Sure. Look, there's good progress with FinFET in the industry, and what we've been focused on is migrating our next generation of products in 2016 into FinFET.
We're working closely with the foundries. We have test ships. We track the progress. As usual, we don't release specific details there, but I can tell you I'm fully confident with the foundry of FinFET ramp in 2016. I know some are going out early in 2015, and they'll be part of the early ramp. We're out in 2016. Looks quite robust.
Chris, just to give you a little bit more color on that. What we've talked about today is that you'll see FinFET in our Zen-based roadmap. That's starting first with desktop and going to server. You'll also see FinFET in our graphics roadmap. In terms of margins, you would expect server and desktop at the higher end to be certainly margin accretive. Graphics is right in line with our corporate average.
Great. Ed, Sharon, there's a gentleman here with his hand. Thank you.
Mark Lipacis, Jefferies. Thank you very much for the presentations. If I may, a question for Lisa, maybe a question for Forrest. Lisa, the semi-custom business, you started ramping two design wins on gaming in 2013, two you expect in 2016 and one in 2017. It's about a one a year cadence. Is that a reasonable cadence for us to think about going forward? A question for Forrest. The data center business that you're excited about, to the extent there's upside in that business, do you think it comes more on the semi-custom side for your customers, or do you think it'll be more standard products? Thank you.
Great. Thanks for the question, Mark. On semi-custom, I would say it's still a new business model for us. We saw the two big game consoles, as you said, ramp. We announced two new semi-custom wins at the end of 2014 that will start ramp in the second half of 2016. I mentioned earlier today that we expect to sign another semi-custom design win this quarter that will ramp in the 2017 timeframe. I would say it's on the order of a couple a year, and depending on the market, depending on the size, it'll be a lumpy sort of business. The important thing that we talked about, though, is the continuum of the business. The fact is, whether it's embedded or data center starting as a standard product going to semi-custom, it's really an opportunity for us to deepen the relationship with customers.
I'm not so focused on the sheer number as much as how are we differentiating in the business. Okay, Forrest?
Yeah, on the data center side, I think that I'm going to reflect back to that continuum message again. I think that having both standard product as well as semi-custom in the portfolio is a very powerful combination, and quite frankly, I think that we're very optimistic about both sides of that business. I think you'll absolutely see standard Zen-based product from us in 2017, and I think there's great opportunities on the semi-custom side as well.
Great. John, we have a question here in the middle of the room.
John Pitzer with Credit Suisse. I've got one question for Mark and then one for Lisa. Mark, just on the core improvement on Zen, 40%, can you help me better understand how much of that is the move to FinFET versus sort of the change in the architecture of the core? As you think about predictability of your roadmap, you talked about it from the design side. I'm kind of curious from the Moore's Law side, how should we think about your progression down Moore's Law beyond 14 nm FinFET?
Sure. That's a great question. It's really why I showed the metric of instruction per clock, because what that does is it actually separates out the process variables. Process is very important. Look at the industry's ridden on, in its history, a traditional huge bump you get node-to-node, and we've seen that be more variable in recent years. FinFET provides a very big jump, and we see a good opportunity in 10 nm beyond it. When I showed the 40% instruction per clock on the Zen core, that's really taking out the element of the FinFET process, and it is talking, in fact, about the microarchitectural changes we made to the execution units, the cache subsystem, the memory, and the bandwidth of how we bring and actually feed in necessary data and instructions throughout the core. It's very much separate from the process technology dependency.
Lisa, qualitatively, you've been talking about the second half as being kind of seasonal, but when you look at the numbers Devinder kind of gave quantitatively, that implies a much better than kind of normal seasonal second half. Is that your expectation of share gains, or do you actually think given the low base effect in the first half that the overall industry is just going to have a better second half, or is this other revenue streams outside of PCs that are getting you comfortable with that half-on-half growth projection?
Yeah, John. Good question. Let me provide just a little bit more detail there. In terms of where the PC business is going to be, I think my comments on the earnings call a couple of weeks ago and my comments today will be, let's call it roughly seasonal. Is it a little bit better, a little bit worse? It's really too hard to call right now given where we are in the quarter. However, when we look at sort of the dynamics of our business, we were taking channel inventory corrections in the first half of the year. Those were very deliberate actions. We do expect the channel inventory levels will be more healthy in the second half of the year. That's one driver for us. We do expect our products to give us a lift.
When we talk about sort of our next-generation graphics products and Carrizo, I think that's another driver. The third thing is, because our business is better diversified right now, we have very good visibility into what semi-custom will deliver in the second half of the year, based on our current visibility, that gives us the range of guidance for 2015.
Great. We have another question here in the Just the row here. Thank you.
Sanjay Jha here, Nomura Securities. Mark, I have a follow-up on the IPC question. You talked about 40% improvement, clearly a significant improvement. My understanding is it's not enough to close the gap with Intel, on that two-part question, where all it allows you to gain share, is it enough to gain share in the high-end servers? Part B, your slide indicated it's a second half launch for Zen. That's what at least it looked like. How long would it take it to trickle down in your roadmap?
Sure. I didn't show a specific comparison of the Zen core to a competitive core in the industry. I can tell you that the reason we say that Zen core brings us back to high-performance competitiveness is because that's what it does. We've looked at that market, we've looked at the competitive landscape, we feel quite confident it puts us in that zone of competitiveness that we need to compete in the market, in desktop, in the data center, and we'll bring that core forward as well in 2017 into our APUs. Timing, you really saw it in Lisa's roadmap. It is shown in 2016 to be going out in our desktop products. We haven't given specific dates of the particular product launches.
I think if I just add to that question, I think the important thing that Mark mentioned, that I will reemphasize is, this is a multigenerational commitment we're making to high-performance CPUs. The first data point is Zen. I think we have teams deployed on the next core revisions as well, that gives us good confidence in terms of the competitive nature of the roadmap.
Grace, we have some questions over here.
Thanks. Two clarifications for what we saw or didn't see. When you were talking about your future cores, you highlighted the Zen core, traditionally, you've run two separate cores, a high-end, I guess those were the earth-moving equipment, then the Big Cats. Are we not going to get any further Big Cats cores going forward? What happens to, say, game consoles, which are based on those things?
Mark, do you want to.
Sure. One of the things that we're doing, and it's quite accurate what you recognize in the roadmap. We had the Cat family and the Bulldozer family of a performance optimized with Bulldozer and actually performance per watt optimized with the Cat family. That's what we've done historically. What we're doing is leveraging the scalability we get with FinFET. What we did with Zen was we optimized it so that it could leverage that FinFET scalability. You can run it at a higher power and very high performance and actually scale that core down right into APU, into the mobility space, and leverage that at a much lower power, yet still leverage that efficient instruction per clock, that microarchitecture gains that we brought into Zen and realize that benefit. It very much brings us a much more focused and efficient approach going forward.
Again, we've leveraged that efficiency to add Arm into the portfolio. What we've got is one x86 core that's scaling very well across the markets that we're targeting. You saw that in 2017, we'll have with the ARM v8 license, we'll follow it with a performance-optimized Arm core as well.
Great. Just one quick one following on that. Talking about the Arm cores, I guess on the roadmap, we saw a server product for our Arm-based server product end of this year, a follow-on next year. I think, Lisa, you also mentioned the microserver market hasn't been as big as you thought, and you've also divested SeaMicro. Do you expect for the indefinite future to keep developing standalone Arm server products, or is it just these next couple of years that we're going to have them to run out what you've got in development?
Sure. One clarification I'd like to make, and then I'll let Forrest answer the roadmap question. There has been a little bit of confusion about whether the SeaMicro systems business actually is a statement on whether we believe Arm is or is not important in the chip side of the server. I think those are two separate decisions. One was a decision on systems business and systems investment. Relative to the Arm chip investment, I think as we said during the presentation, it's an important business and an important market for us. I'll let Forrest comment on the details of the product roadmap.
Yeah, certainly, we anticipate introducing the A1100 Opteron, the first real Arm server chip, in the second half of the year with the ecosystem fully ready. We do think that there is an opportunity there in the long term in the standard server market, I think that is taking longer to develop than some of the forecasts. We think there's a tremendous opportunity for Arm as we look at many different applications or different types of enterprise equipment migrating away from proprietary solutions onto software server-based solutions. You should anticipate that we're going to come out with the A1100. We're going to follow it up with chips based on K12 at the upper end of the performance spectrum. We'll certainly stay participating in the broad enterprise infrastructure space with Arm.
Great. There's another question here next to David. Thank you.
Hi. Stacy Rasgon with Bernstein Research. Two questions. First, maybe a bigger picture one. You've outlined a pretty broad strategy here. You've talked about a big focus on high-performance computing cores. You've talked about a big focus on data center and infrastructure. You've talked about a focus on embedded and IoT. How is that different from Intel's overall arching strategy on an R&D budget that's 15% as big?
Yeah. Stacy, let me take that and perhaps try to narrow it a bit. I think IoT is a very, very big market, clearly that's where a lot of talk in the industry is. We are not going to address the entire market. We're really going to focus on the places where we can uniquely contribute, it has its foundation in our IP. What are we good at? We're good at high-performance computing, we're good at high-performance graphics, we're good at workloads that utilize that kind of capability, I think we can be much more flexible in how we choose to run the business model. We are not going to be in the endpoint space. We're not going to be in the smartphone space. In fact, we're actually not competing really in the low-end tablet space, where there's a lot of silicon consumed.
If you have a fab, you need to focus on silicon consumed. We don't have a fab. We have to focus on where can our IP be monetized. It is very much synergistic subsegments of these spaces versus we're going to attack the entire IoT market.
The broader question around data center, I don't want to be blunt, you talk about being good at this. At the same time, in servers, you have 1.5% market share, down from 10% over the last 4 years. In PCs, you talk about a long-term target to be flat to down. You've been down 20% a year for years. How should we think about these targets, I guess, in the context of that history, where you're going with this? What gives us confidence that those improvements are now happening when this has sort of been the plan for a while and they haven't happened? How do we get confidence around this?
Yeah. Let me try to frame it in this way, again, I want to address your question head-on because it's an important question. In data centers, the reason we have 1.5% share today is we decided not to invest. If you take a look at our core roadmaps, we decided not to invest in high-performance computing. Now, when you want to get back into that market, the truth is we have the technology, we have the IP, we have the people to do that. We need to set the design point there, that's what Mark was talking about. When you set a design point, a brand-new microarchitecture from scratch, it takes 3-plus years. We've been working on it, from that standpoint, that's what gives us confidence.
If you think about just all of the various places that we've been able to compete, it is really about the technologies, especially in the data center. I've spent an enormous amount of time with customers. I think Forrest can also talk to the customer aspect of it. The data center wants AMD. Customers want us to offer the technology. What we must do, though, is it can't be just we're the cheaper solution. That is not the right way to do it. It is to start with technology that's competitive and ensure that we're wrapping around it things that others won't do, and that was the multiple workloads that Forrest talked about. That's the data center market, and maybe let me let Forrest comment, and then I'll come back to PCs.
I think one of the things that's very important for AMD, first off, it's a very large market, and we're not going to try to cover every space in that market. I think we're being very thoughtful about where we see opportunities for the parts that we have, the technology that we have, the flexibility we have in the modular approach that Mark outlined to be able to target specific segments of the market and produce differentiated solutions for specific segments of the market. In a way, I'll tell you something I've told people internally. You're absolutely right. We've got 1.5% server market share right now. I'm telling my teams that's an asset. When you have 30% or 40% share, your strategy has to be centered on broad participation, general purpose, very broad design points.
When you have 1.5% market share, you can think differently, and I think that's where AMD has had success in the past, where we thought differently, where we identified. Quite frankly, the 64-bit's a good example, or even the integrated memory controller. Things Intel doesn't want to do or things Intel is not doing. There's a broad enough disparity, a broad enough range of workloads and evolution in the data center now. There's plenty of opportunities for us to focus and optimize and substantially grow that 1.5% share.
Okay. Stacy, let me get back to the PC question because I think that's an important one as well. The PC market has really undergone a tremendously volatile time for the past couple of years, and there have been various different aspects of it. AMD's PC market is, if you look back in 2012, was not reflective of the overall PC market. We were heavily concentrated in low-end consumer PCs. That's our heritage. That's where we had the majority of our share. That's where we played, let's call it, the cost-performance game. That portion of the market has seen the most volatility. We have definitely lost share there. Some of that share we've decided not to take because it was unprofitable given some of the marketing development funds that were in place.
What we have done is we have really made progress on changing the mix of our business. If you take a look at our ASPs in PCs, they've actually held their own. If you take a look at subsegments, like for example, the high-end mix, competing against Core i3 and Core i5, we've actually done quite well in those segments. We've replaced some of that low-end share, let's call it, with better ASP business. It's a process. We're going through a process in that PC business. There's no question that I want to take share in the PC business, but we're going to do it with profitable share. That's on the computing side. On the graphics side, I think share has gone back and forth a number of times. We've actually gained share in OEM graphics. We've lost share in desktop graphics.
As our technology comes out, I believe we will show share gain possibilities. I'm not asking you to trust me, but I'm asking you to understand that we understand our business. From that standpoint, we've laid out what are the segments that we think we can compete. That's where we have to lean in. That's where we have to invest.
Thank you.
We have one here in the middle. Ian, there's one here with glasses.
Thanks. Ian Ing, MKM Partners. First question is for Mark. Performance per watt expectations across a lot of your roadmap. Could you talk about low-power design where you're advantaged? You and Jim Keller have a track record here, obviously, competition's out there also. Where is your exact recent innovations and advantages? Thanks.
That's a great question, you think about it, look at our APU roadmap. In fact, I included a slide in the presentation showing the APU roadmap and that 10x energy efficiency we gained over those six years up to 2014, a big gain that we have with Carrizo, the Excavator core, then the overall sixth-generation APU. What we're doing is clearly leveraging that. When I describe what we did with Zen from instruction per clock, that's about performance, that's about the microarchitecture for performance. We didn't back off at all, continuing gains in power efficiency that we'll drive. We've invested in the modeling, the detail that we have, because it's all driven by data. You have to do exhaustive analysis. You have to run applications, traces. You look at where is the power being consumed, frankly, it's a lot of hard work.
You ratchet your way through, you manage those sources of power dissipation in the workloads that you care about. It's the same thing in graphics. What we're doing is we've created really a synergy that we didn't have before at AMD across the teams. We've broken down silos. We're taking the learning that we have across our APUs, we're applying it on the other segments. What you're seeing in that commitment that I showed you in the graphics roadmap is applying those techniques we've been really leveraging on the APU side and bringing it directly into our discrete graphics. It's blocking, tackling, hard work, analysis, simulation, managing the power down with, again, techniques that we've talked about in technical papers of really managing the switching activity, getting work done. We have a philosophy, return to zero.
Get the workloads done, get to where your power management's kicked in, you have no power dissipation nor leakage where you can turn off power islands. It's very much engineering focus and application.
Thanks for the details.
Great.
Devinder, just a quick clarification. Second half gross margins on new products up a bit. Are the traditional second half headwinds still there, like game console ramps and things like that, or is that abating at all? Thanks.
From a revenue standpoint, if you're talking about the second half-
Gross margins
essentially the semi-custom, typically, if you go back and look at 2014, the behavior was second half was stronger than the first half, and we expect the same. From a revenue standpoint, if you look at PC market, we have new products coming up. Typically, the second half is stronger than the first half, and we're fully expecting that. In the space of new products, we think the new products allows us to become more competitive, and therefore, it benefits us across some of the other areas that we play in.
I think, Ian, your question was about some of the gross margin drivers as well for the second half. Yeah. Do you want to comment on second half gross margins?
Sure. Gross margin, if you look at it from a new product standpoint, that benefits. The new products are more competitive, therefore you get better gross margins. Obviously, you have the mix of the business improving in the second half, and that helps the gross margin. That's why we think the gross margin can be better in the second half than where we ended in the first half. I'm calling it flat to up in the second half compared to where we ended the first half of 2015.
Great. Next up is a gentleman here in a white shirt just in front to the left. Thank you, Chris.
Matt Ramsay, Canaccord Genuity. One question for Lisa and then a quick follow-up for Mark. Lisa, maybe a different way to ask Stacy's question is, you had a slide where you listed what we're not going to focus on, right? I think it was smartphones, tablets, IoT endpoints, or microcontrollers, right? I don't think any of those things, when any of us walked in this room today, we really thought you guys were focusing on. Devinder's guiding down OpEx in the back half of the year slightly. Maybe you can just talk about the focus and where those cuts are coming from.
Yeah. We have made some key decisions relative to simplifying what we're spending time on. For example, when we decided to exit the SeaMicro Systems business. That is a source of OpEx dollars that we can now reinvest in the data center side of the business that we've talked so much about because there is a reinvestment that's necessary there. When we look at simplifying the roadmap and how we've talked about the roadmap simplification, we've looked at the products. Over the past six or seven months, it has really given us a chance to re-look at the products and say, "Okay, which are the products that are truly going to return on investment because they give us strong market positions, and which of the products are not?" There are other things that we've done to simplify our roadmap.
For example, in the past, I've talked about 20 nm being a node that we did some designs in. We started some initial designs. We've run some silicon. Those products are probably not going to go into production because we think we can get much more bang for the buck out of the FinFET technologies going forward. A lot of, I would call them micro-optimizations in terms of ensuring that everything that we're doing is paying off. Mark talked about the x86 and Arm leverage. That's very unique to AMD. Okay? The fact is that we're doing x86 and Arm, and we keep getting asked, "How could you do both given the ecosystems that both require?" He is creating leverage between how we build those CPUs together, and we've created more leverage in the way we put the roadmap together.
I will say that with Forrest joining AMD, we've had a chance to re-look at some of the things that we've done as well. The thought process of that particular slide was there are some large markets that we won't invest in. I still get asked very often, "Why aren't you investing in mobile?" Because that's where all the units are. I think it's important for us to restate that. This concept of simplifying and really ensuring that our development dollars are spent on the highest leverage opportunities is something that is throughout the company.
Great. That's helpful. Mark, I was kind of excited about the High- Bandwidth Memory slides that you guys put up. How applicable is that into your higher volume businesses, be that PC or eventually server?
That's a good question. We're on HBM Gen 1, and the discrete graphics are rolling out. We're very happy with the performance we're seeing. We're coming up on-ramp, obviously, as Lisa said, in weeks. We'll be announcing product. It's in the high end of our product space. Again, details to come. What you're going to see with HBM is like you've seen with any other new memory technology. It's all about scale. As the volume comes up and the costs come down, that opens up new markets. Any of the segments, and you follow it closely, so you know where they can benefit from that type of lower power and a higher bandwidth. When the cost comes down sufficiently, it'll likely make sense to extend its footprint beyond the discrete graphics.
We have a very back of the room in the middle there in the checked shirt.
Okay. Two quick clarifications, two questions. The clarifications, Devinder, what was the amount of cash in the U.S.? Nine zero or one nine or nine?
More than 90%. The cash that we have on the balance sheet, more than 90% of that is resident in the U.S.
The second clarification, Lisa and yourself have both talked about, I think, $800 million in cash for the June quarter and between $800 million and $1 billion for the December. I assume that doesn't include any ABL draws or any other debt additions, that's a net cash.
At this point, no. We're planning with the cash management actions that we have, and we've done that, $800 million on a, I don't want to say natural basis. There was a draw on the ABL in Q1. Just to clarify, we did pay off the $42 million of convertible that I talked about, but that's just trading lower interest from an ABL standpoint on the convertible note. Beyond that, no plans to draw on the ABL on an incremental basis.
That $800 guide does not include a draw on the ABL or any additional debt?
Does not include incremental draw.
Right. It includes what was drawn in the March quarter.
That's right. Other than the $42 million that we paid off, there's no incremental draw to get to that $800.
The two questions, the first one, any way you can answer this, how should we think of inter-quarter working capital swings? Or maybe asked another way is inter-quarter on numbers that we don't see, do you ever end up going into the ABL?
Inter-quarter is a difficult question in an evolving or transitioning business model, right? I think the takeaway you should have is, we manage our working capital on a very tight basis from a viewpoint of serving the customers. For example, when we gave guidance for Q2, one of the things I said is, inventory from Q1 to Q2 will grow $100 million. That's primarily because we see, first of all, the stronger revenue in the second half, then in Forrest's business to satisfy the higher ramp in the semi-custom products. We have to get the inventory, produce the units, get it to our game console customers ahead of time. Therefore, you have from a timing standpoint. Inter-quarter working capital is typical.
You have obviously the cash collections, and you have the inventory management coming in on time. I'm not too worried about that from an overall standpoint.
The last one, sorry. About two management teams ago, there was talk of ending the take or pay options with GlobalFoundries. This year, the wafer supply agreements again are take or pay. Is this a decision that you're making in order to get better commercial terms, lower pricing, or is it just that all that's available to you? How do you sort of weigh potentially lowering the take or pay amounts or even removing the take or pay nature of the contract versus pricing?
I won't comment on the two management teams ago, what they said. I'm not familiar with that concept. I think if you go back and look at the WSA, which has been in place since 2009, we signed the WSA in 2009, and since that time, in 2011, for the first time, we did an annual contract on a take or pay basis. How that helps us, if you look at it from an AMD standpoint, is you get into a discussion typically at the end of let's say end of 2014, we got on discussions about what's our view of the market, the mix of product. We make at GlobalFoundries today not just the PC-based products, but starting in 2014, we also made the semi-custom product and also the graphics product.
We look at that, obviously capacity is allocated for AMD's needs, and that discussion becomes a take-or-pay. There's two aspects of that. There is an obligation to take, but there's also an obligation to make. From my standpoint, sometimes in a situation that we live in in the foundry world, that's a good thing because it's a assured supply. From a customer standpoint, we can go, for example, to Sony and Microsoft and say, "This is how much product we're going to make, and this is where the supply is going to come from." The take-or-pay, I wouldn't read too much into that. It's an annual contract we have done now. We've done amendments to the contract. It's an annual contract.
It's very good for us to have a discussion with one of our key foundry partners to say, "This is what we need," we go ahead and take that. The other thing I would add is if you go back and look at 2013, in particular, as the year came to an end, the take-or-pay agreement that we had with GlobalFoundries early in the year was higher. As the business evolved, as demand profile changed, we did not take as much as we signed at the beginning of the year, with no special trigger payments, no penalties or whatever. That's just, I think, a function of the relationship as it's evolved between us and GlobalFoundries.
We have a question here in front of the camera, directly in front of the camera.
Quick question for Forrest on accelerators. I wanted to clarify. Have you already developed accelerators in-house? Do you have that product or not?
I'm sorry, do we already have?
Accelerators.
In terms of APUs, we have not yet announced a server-centric APU product.
Yeah, because I saw it on the portfolio, I was wondering. It would seem like you'd be in a unique position having a CPU and a GPU. You'd be one of the few companies that could address Opteron and then having an accelerator in-house. Is that kind of the plan maybe to differentiate yourself from-
Yeah, I think I mentioned in the 2016-2017 roadmap slide our intention to do not just a server APU, but a very, very high-performance server APU. Absolutely. We see one of the key assets for us in the enterprise business, particularly as, again, these computational demands continue to increase, being able to wed the two together gives us something that's worth more than the sum of the parts.
Last question from me, just when you talk to customers and having come from Dell, et cetera, does it look like customers now? Is it mandating now almost accelerators? How broad is this going to be as a trend, maybe as an add-on feature within servers? How important is it becoming?
I think it's very workload specific. It's a little bit hard to predict what it's going to be in three to four years. There's many traditional HPC workloads which are very amenable to GPU or other types of acceleration, and that's been growing at a pretty steady clip. If you had asked me that question about two years ago, I would have given you the answer that, well, I think accelerators will always be important but never be dominant by any stretch of the imagination. However, as over the last couple of years, we've seen a number of new algorithms really burst onto the scene and burst onto near-term commercial application. I think that has tremendously increased the interest and the potential commercial demand for accelerators.
Those were workloads that quite frankly, we and others didn't predict a couple of years ago because, quite frankly, it was a fundamental breakthrough. I do think that acceleration technology of all types will become more and more important going forward in the enterprise. I think equally important is you have to have the right software and systems architecture to be able to make that acceleration easy for the developers to use. The whole HSA Foundation and that whole approach is all about creating standards around that to lower the hurdle rate for developers to actually develop and utilize that acceleration facility.
Thanks.
All right. There's one just in front of you here.
Yeah, thanks. Thanks for taking the question. If you look at your sort of 10 or 15-year history in the past, you see that there have been moments where you have done pretty well in certain segment of the markets. For example, in 2007 in the server market, recently in the console market. It seems like it's just never enough. It seems like in that sense, do you really need to think about your business model? Do you need to think about potentially partnering with some bigger organizations in terms of your go-to-market strategy perspective? Seems like what you have been doing for the last 10, 15 years, some periods of nice momentum, but then you cannot sustain it. How should we think about that going forward?
Yeah. I think that's a very good question, and it also gives me an opportunity to maybe just step back a little. AMD is a company that I will say, most people will say we have a lot of good technology, we have a lot of good people, we have a lot of potential, but we haven't seen sustained momentum. When Opteron was great, we were at 25% market share, we had better technology. We missed the next node, we lost that technology momentum there. I believe that what needs to be different is really a different way of thinking about how we attack the market. Okay? Building the products, and they will come, that's sort of one aspect of it. Building products in markets that are fundamental growth markets and that we have differentiation beyond just the speeds and feeds is critical.
It's about not just the technology, the technology is very important, but it's about the customer relationships that we generate. Why are we so proud of the game consoles?
Winning them, very hard. Once you win them, that business is with you for five to seven years. When we look at the segments of the data center that we're going after, going after the traditional enterprise space is interesting, but going after the new workloads where we can really provide differentiation beyond just the speeds and feeds. The speeds and feeds are kind of table stakes. It's, as Forrest said, what can we do around the memory technology? What can we do around workload optimization, combining our CPU and GPU technology? That will give us longer standing, stickier differentiation. It is absolutely true. My goal in life is not to be a flash in the pants. It is to figure out how we build a three-to-five year consistent profitability, cash generation, and sustainable operating profit. I believe we have a great technology platform for that.
We have to marry it with the right customer relationships and the right differentiated products to do that.
Yeah. If I can just add a quick comment. That's what's also quite fundamental to the technology roadmaps I discussed earlier is, yes, we're very focused on Zen. We've got great products this year and next generation of products in FinFET. What I emphasize, and I want to underscore that point relative to your question, is we have the roadmap beyond that. We didn't do a point investment on the key engines we've generated. Always will, in fact, right? This is the new model of our R&D development in AMD, is to be doing a next generation and a leapfrog team that's working on what follows. We'll do that across the elements of our portfolio, our IP and technology portfolio. That's fundamentally different.
Okay, we have time for a couple more. There's a gentleman in the very back of the room in a blue shirt.
Hey, thanks. Can you talk about the gaming opportunity? Not the consoles, but the casino games. Where is that opportunity coming from? Is it you're gaining share in the games, or what's going on there? I'll probably have follow-up.
Yeah. Clearly, we saw casino gaming as close adjacency given the technology portfolio we'd put together for the console gaming. We've been attacking that one now for a couple of years, primarily with standard products, although there's some interesting semi-custom opportunities there as well. Our initial focus was, I'd say, in the traditional slot machine and traditional casino game market. We've recently opened the aperture of that and are now attacking pachinko and pachy slot as well. We think we've got number 1 market share in what you might call the traditional casino gaming already, in very short order, and now we're expanding into some of the adjacent game categories.
Is that end market growing in terms of number of slot machines that are being ordered, or is it growing because there's more content per machine?
I think it's a little of both. I think, quite frankly, on that second part, we see multi-displays being added to many machines, so you're moving from a single display on the poker machine to three or four displays on the front and the sides of the machine to try to pull the consumer in to play that particular box. We see many more displays as well as much richer graphical content and a lot of 3D content coming to casino gaming as well.
Lastly, I apologize if this has been asked because I've been in and out, with regards to consolidation, can you discuss it? It's obviously been talked about out there. What are your thoughts?
When you say consolidation, you mean semiconductor M&A?
Exactly.
Yeah. Look, it hasn't been asked, we will certainly address that. Look, we just spent a couple of hours laying out a long-term strategy, when you look at the strategy, we feel really good about where we're investing and that we're investing in the right growth markets. That is absolutely our focus. When you look at the semiconductor landscape, there are lots of things happening. I think we view ourselves as having a very unique IP portfolio. When I say IP, I mean technology products and patent portfolio. I think from that standpoint, that's good. We'll always look at the landscape and have it continue to evolve, I think the single most important thing is executing on our strategy.
Great. We have a gentleman just here with his hand up with the striped shirt. Joe.
Thank you. It's Joseph Moore from Morgan Stanley. Can you talk about the high bandwidth memory? That's been an expensive product, it sounds like, from the silicon interposer technology has been expensive. Can you talk about your ability to penetrate sort of mainstream GPU price points? As I talk to your channel partners, there's been some concern that you're limited to four gigabytes of memory buffer. Can you talk about if that's a limiter?
Yeah. Maybe, Joe, I'll start and see if Mark has anything to add. We haven't announced details of the product, and again, I'm not going to get into the details of the product. High Bandwidth Memory, though, is a very differentiating technology. You shouldn't think about it in comparing speeds and feeds relative to today's graphics cards. If you look at the bandwidth that Mark showed, both between the memory and the GPU and the power efficiency of it's like comparing apples and oranges. We'll go into a lot of those speeds and feeds when we actually launch the product. The reason we brought it in is we thought that the enthusiast class graphics is the place to introduce this technology. Really, when you introduce new technology, you're testing it in manufacturing and testing all aspects of the ecosystem.
We think you're going to think about graphics cards differently, and I think that's the excitement around it. How do you bring what used to be a very high-end graphics card that would only go into big gaming rigs. Put it into different places. That's where we see the growth going forward. I don't know if you had any-
Well, you asked about the cost point, and it really goes back to my comment earlier. It's about economy of scale. Like any new technology, as you ramp, you get manufacturing efficiencies, and you bring the cost point down. Again, that's my expectation with HBM.
Just more broadly speaking, your GPU business last year was worse than your CPU business, and I feel like the CPU shortcomings you had, losing share in the low end are pretty clear, but I'm still a little unclear on why GPU was so difficult for you. It seemed like maybe you got leapfrogged a little bit at the high end, but your product portfolio was pretty good. What happened from a revenue standpoint last year?
Yeah. Let me talk about the GPU business. We don't typically break out in granularity the GPU business from the rest of the business. The way to think about it is the GPU business is, I would say, a back and forth typically with technology. When we announce a really strong product, we gain share. When others announce a product, there's a little bit of share going back and forth. Specifically, in the GPU business, we actually gained share over the last couple of years in OEM graphics. If you look at some of the design wins that are out there now, particularly with the Apple iMac, and that's an example of an OEM graphics design win that we won and gained share over the last couple of years. We did lose share in desktop graphics, particularly in the AIB market.
A lot is going to ride on our product portfolio. The second half 2015, the launches around our top-to-bottom portfolio in both mobile and desktop are important. Mark talked about the power efficiency that we are bringing in 2016 with the FinFETs. That will certainly improve our graphics competitiveness. The portion of the graphics business that I would say is still the most untapped portion is the professional graphics workstation and the GPU compute that Forrest talked about. It is about, I would say, where do we focus our efforts in the graphics business. We are going to focus on the areas that are higher margin because that certainly is back to profitable market share. We feel very, very good about our competitive graphics portfolio.
We need to invest more in software, for sure, and we need to continue the investment in the new technologies going forward.
Thank you.
Great. Well, that actually concludes our questions and answer session. We will say goodbye to the folks on the webcast and thank them very much for joining us today.
Thank you very much.