Also available via the app that I mentioned a little bit earlier. If both of those fail, we'll also have hosts in the hallways and in the lunchroom as you're seated, and we can guide you to the right table. All right. With that out of the way, I'd like to invite our Chairman of the Board, Andy Bryant, on stage.
Good morning. Welcome to the 2015 Intel Analyst Meeting. As I begin, I would like to actually thank, I have two independent directors here, Jim Plummer and Frank Yeary. Later this morning, we will have a third join us, which will be Dave Patrick. It's always good when you have them in the room, and they're learning about the company and actually watching how we do, I guess. Those of you who have heard me talk in the past, one of the things I say about Intel is, we're a technology company. We invent things. I know at times we've said we're the PC company or the Internet company or the all sorts of other company, but bottom line is we have always been a technology company. We are one of the most innovative companies on the planet, and that's who we are.
We also know we have an obligation not just to invent stuff, but to turn it into business value. One of the things you find about this management team is they all get that, and they understand the goal isn't just to find technology. The goal is to get a return on our efforts. Today, I'm pleased to say, in the last year, I think we've had good progress against the things we committed a year ago. If you start with the Data Center, we certainly have strengthened, I'll call it, the CPU franchise there. We've done more than that. We've expanded investments into adjacencies. You look at Fabric. You can look at silicon photonics. You can look at Diane's team starting to do customization for certain customers. It's no longer just make a CPU and see what happens.
You're looking at a broader Data Center environment instead of just a server CPU. I think this is our most significant potential growth opportunity. You've heard us talk about the Internet of Things business. Had strong growth in the past. Takes our embedded history, a sub business we've been in since I've been here, and builds on it. Takes it, looks into the cloud, looks at the new capabilities, starts to connect itself to different parts of Intel, not just Intel, but the market in general. What you find is, by the way, it's a business that we talked last year about common IP. It uses a lot of the IP generated elsewhere, so it's a low incremental cost business to be in. The other thing that happens is they invent things that actually come back into the other parts of the company as well.
You may find ultra-low power needs in that business, which becomes useful certainly in a client business. Again, it's one of the things that not only creates value for the shareholder, it creates value for the other businesses as well. One thing we've talked more and more about recently is the memory business. The memory business continues to innovate, continues to be a strong, profitable growth business for us. Basically in NAND and SSDs. However, you saw the announcements this year, 3D NAND. You saw the announcements on 3D XPoint. You're starting to see new technologies there, new technologies that not only help the memory business but connect, for example, into the data center business. You're seeing this value come across different business units as, again, innovation takes place.
Of course, I'd be remiss not to talk about the PC business, which is a constant state of conversation. It continues to be, quite frankly, a very profitable business for Intel. Vital source of technology, volume to amortize into the factories. It's absolutely a critical foundational piece of business for Intel in a lot of different ways. Certainly, we've been optimizing returns in what I would consider a very difficult environment. If you look at the worldwide GNP being a little less than we expected, a lot of the business is tied to that. On the other hand, you look at what Kirk has been doing, he's bringing innovation into that business. He's bringing innovation in terms of new capabilities that may be of interest. He's bringing innovation in terms of integrating, which adds value.
Because of that, they will not only support his ASPs, average selling prices, but they may even go up a little bit. You see the innovation even in the core business, the more mature business making a difference. The other thing that comes out of that particular business, it's the biggest investor of the, I'll say, the common intellectual property. A lot of what everybody else is using, from Diane's server business to Doug's Internet of Things business, starts there, and that's where a lot of the expense is carried. Again, it shows how the common IP spreads across the different businesses. The foundation for all this, of course, silicon technology. Hopefully, you just watched the Gordon video, which I thought was interesting. Moore's Law has shaped Intel's past, quite frankly, it will shape Intel's future.
I know there's a lot of talk these days about, are we losing our advantage? Is it still of value? Bill will be here today. He'll address those, I don't have to, thank goodness. It will be on today's topic. Bottom line is Moore's Law is still a fundamental differentiator for Intel, it will be for years to come. I'm chairman of the board, the board has this responsibility to continue to monitor the company, to review the strategies, we try to do that. In fact, I would say most of our strategic discussions in the last year have been around how to best allocate capital. What you should see through the day is, I won't say this is the board driving this, but we do get to review and approve the plans.
You'll see the company investing more in the growth and emerging business, more in data center, more in Internet of Things, and more in memory. You'll see the company beginning to invest less in either more mature or unprofitable businesses. The investment in the traditional PC is coming down, and phones, and software. Again, trying to make sure we allocate capital to the best use for the shareholder. Of course, as you follow us, the largest allocation of capital right now is the purchase of Altera, assuming that does happen. When you buy Altera, you do get a separate business. Again, as we talked earlier, how it wraps into the other parts of Intel, it should help the Internet of Things business, it should help the data center business.
What we see is the ability to take that acquisition, a lot of capital being deployed, and use it in a variety of places. You recently read we're increasing our capital commitment into the memory business. Hopefully, you read that. It was in investments in Dalian. Again, we see that as an important future business. The memory business should be more profitable on its own, but it also should tie very closely into the data center business and help that particular business. Of course, we always invest in leading-edge technology. It's one of the things we do. It is Moore's Law, and it is a differentiator for us. Now, when we talk Moore's Law, this was the 30th? 50th. 50th anniversary of Moore's Law was this last year.
We actually had a nice little celebration in San Francisco, and I was amazed at how widely and how positively it was viewed. We thought it would be a nice little event. Probably millions of people followed it, thousands of stories in 39 countries, and the overwhelming theme was how people were encouraged about how technology has changed lives and how it will continue to change lives. Having been there, it was actually one of those nice events you get to have in your career. When people think about Moore's Law, of course, we all think technology. It's really an economics law. If you read the paper, which flashed up there briefly, it's something I understand. I'm an economist. I was an economist, I guess. Now I'm a chairman.
What's interesting is with our founders, each of them made the transition from scientist, inventor, to very good business leader. It's incredible when you look at the history of Intel and how we have made those transitions, and that's what we have to work on today. If you ask me what's most important for Intel right now, we focus on execution, we focus on getting results. We understand what matters is creating shareholder value. My view is we're on track. We've had a decent last year. I would acknowledge the last two years has actually tested the impatience at times of our employees, as change is not easy, and our shareholders at times. It's necessary to get to the place we want to get into the future.
I think we have seen decent progress this year, and we expect to see continued progress in the year to come. The good thing is, I just get to say what I expect. I now get to bring on the stage the guy who actually has to do the work. Let me introduce CEO, Brian Krzanich.
Is everybody excited, energized, ready to go? Okay. Let me see if I can get to my agenda. I thought what I would do this morning is go through a little bit of the 2015 results. As Andy said, we have some pretty good results, we think, relative to the market and relative to what's been going on. I thought we'd spend a little time on the overall corporate strategy. We'll put it into context on the big picture. As Andy said, we've really spent the last two years really trying to refine that, put it into motion, adjust how spending and the organization is structured, and how people get things done here day to day at Intel to align to that strategy.
I thought we'd take the segments apart, so the clients foundational business, the data center, the IoT, and the memory, which are the growth engines. I want to spend a little bit of time talking about the future investments. There, I thought we'd talk about Altera, and put a little bit of context around that, and then our investments in wireless. I know that's on a lot of people's minds. We feel like we've made a lot of progress there technically this year. Then close it up with a summary. That's my agenda for today, and hopefully, it'll meet everything you guys want to know about the company. Let's start by taking a look at just how 2015 has been. We started with our commitment last year, at this meeting, and talked about what are we going to do.
First, we said we're going to pursue Moore's Law with relentless further, and we continue to do that. I think we all feel pretty good about our progress there, 14 nanometers, and Bill's going to get up and talk, and part of the reason we wanted to keep Bill on the agenda and have a special discussion about that this morning is, you read the press, there's a lot of speculation, Moore's Law is dead. What's going to happen? Is the termination, the yields? We're going to try and pull the covers off all of that today and really talk about that. We can tell you without a doubt that Moore's Law continues and is alive and well, like Gordon. By the way, I have to tell you, it was the 50th anniversary, and I had a great opportunity.
I happened to be in Hawaii with my family at the same time Gordon Moore was there. I actually got to do that interview with him. It was one of those special moments. I actually brought my daughters with me. It was great to see an icon like that. We say we develop products that enable the best computing experience. I think if you've seen the discussions on the web about Skylake, if you've tested one of the systems, if you have one of the systems, hopefully, you've seen the performance that we've continued to drive. Really, Skylake, we call it our best processor ever, the best improvements ever we've done in a processor to date. We really try to do a pragmatic market-driven industry view.
We're not saying that the market's going to go the way we want it to go. We're really trying to understand the way the market's headed and move our products and our R&D that way. Open the foundry to companies that are able to utilize our technology. We've continued to make progress there. Bill will talk a little bit about that. Really, we have to wait till our customers bring the products out for us to really talk too much about that. Create platforms for enterprise. This is what you're going to see, especially during Diane's talk. I'll talk a little bit about it. We really see that you have to develop a platform.
You have to design the whole architecture as the cloud. We'll talk about this. As the cloud grows, you really have to think about the whole cloud infrastructure. You can't just sell silicon. We're really focusing on bringing innovation to market quickly. There are several examples over the last years. If you take a look at Quark. Quark was a product that wasn't even on our roadmap a couple of years ago. It's out in a number of products now. One of the latest examples is Curie. We announced it at CES this year. It's already in market. We're really able to spin products at a much faster rate in some of these segments now. Growth in the data center, we'll talk about.
Our real foundation is that if it's smart and connected, it's going to work best with Intel. We think we have a lot of really strong examples for that. Let's get right into that. Our results for 2015, the full-year revenue, the target is roughly around $55.2. You've seen our forecast for Q4. It's got a spread around it. We'll be in the $55.2 range. Full-year margin of 62%. We're very proud of our ability to continue to maintain our revenue, also our profitability. As you put those together, that gives you a full-year operating profit of about $13.5 billion-$14 billion. Throughout today, I think as Andy Bryant said, you're going to see evidence that Intel is transforming that the growth in revenue sources. I'll talk a little bit about that later.
If you take a look at the mix of this revenue, though, the underlying performance of that has really shifted over the last three years, and we want to really talk about that today as we go through this. Let's go right in from that to the strategy and just talk a little bit about what our corporate strategy is. It hasn't changed a lot since last year, but we have refined it. We have continued to improve on it. You're going to see it, you see it in our agenda today. You see it in some of the time you're going to be able to spend, and you'll see it in my discussion as well.
Our strategy is really based on 3 pillars or 3 fundamentals, and those are really that Moore's Law is critical and that we can continue to drive it, we can continue to lead it, and it differentiates what we do. The second thing is really tied to Moore's Law. What Moore's Law does is allow you to continue to integrate more, to continue to bring in more of intellectual property into your product. What that does is lowers the cost, lowers the square footage of the product that you're trying to build. We see examples of this day after day in our lives as products continue to get smaller, more compact, better battery life. That is this combination of Moore's Law and integration.
The last thing is this shared IP, which Andy kind of alluded to in his discussion, where it's really about we generate intellectual property at, say, the PC, where the PC is able to afford it's able to fund it, but we spread it everywhere. One of the best examples I can give you is RealSense. We originally created RealSense, which is our computer vision system, to actually unleash people, this part of this no wires, no connections concept. We wanted to unleash people from the keyboard and mouse of today. We really thought there had not been innovation in that space for quite a while. We developed this computer vision model.
What we found as we pushed that IP into other areas is probably the largest areas of growth are going to be in places like robotics, in places like drones, in a variety, in industrial applications, automotive applications. These are all going to actually take that computer vision probably at a much higher rate than even the PC, and it's a great example where the PC funded this intellectual property, but it's being spread across a variety of technologies and business applications. We believe it's a space where it differentiates Intel again. I'm going to spend just a moment on Moore's Law because you've got Bill, and I think other than Gordon, there's probably not a better person to talk about Moore's Law than Bill.
For me, what I wanted to stand up here and tell you guys is that we continue to believe that Moore's Law is an advantage. It's alive and well. In our transparency efforts, we took a look this year, we actually looked at 32 nanometers, 22 nanometers, and 14-nanometer transitions. We saw that we'd moved to roughly two and a half years. Actually, if you take the history of Moore's Law over time, actually, we have studied it for probably 25 years of history. It's floated up and down, and it wasn't until 10 or 15 years ago that we'd actually pushed it down into two years, again, as the lithography systems caught up to it. It's expanded out and contracted over its history of life.
What we saw was that the trends, the fact that we didn't have EUV, the complexity of the technologies, some of the new materials, the last technologies had taken two and a half years, and we didn't see 10 nanometers being that dramatically different. The lithography systems weren't going to change. The complexity was about the same. We really said, "Hey, we want to give you the best data we have. We want to adjust our plans and spending. We want to insert products and really be clear to our customers so they could bank on that commitment." That's why we adjusted it to two and a half years. Our goal is always to get it back down to about two years. It's always the most profitable, the faster you could do Moore's Law, and we've brought up discussions in Investor Days of the past.
You can go search them up. Andy did one of the best ones a few years ago, where we showed that if you could do Moore's Law at an infinitesimally small time period, it would be the most financially beneficial. You have to bring all the elements together that keeps us at, we think two years is probably about as fast as you could possibly do. We'll always be pushing back to two years, and trying to go for that, but we wanted to be transparent and clear, and that's why we moved it to two and a half years. There wasn't really anything else in there other than that. Let's go to the foundations. The client business is what I believe a strong foundation. You see the products up here on the stage, and you're going to hear from Kirk later today, who runs our client business.
We continue to make great progress in our SoC capability in this space. We're very proud of our latest product, Skylake. I want to tell you that our growth is not dependent on this business. It's a stable, good, profitable business. What we've done this year is taken a very pragmatic look at the industry. Now, internally, Kirk's got the direction. You need to grow. We've also taken a look at the market dynamics, the replacement rates that we're seeing in the market and said, "Hey, the probability though is we need to have a growth strategy that is not dependent on growth on this segment." We've tested this, as you take a look at our growth strategy today, to high single-digit declines even, and we're comfortable that we can continue to grow this business, even with a high single-digit decline of the client business.
I want to take a little bit of a deeper look at how we look at the client business, though. We really see a strong tailwind as we move into 2016. We believe, as we said earlier, Skylake is truly our best product that we've ever produced, and it generates great performance improvements. If you take a look at the form factors and the products, and Kirk will be up here to talk about some of that, truly our partners and us have worked together to produce what we believe are just world-class products. The 2-in-1s that you'll see are really, if you take a look at that, there's not a reason you'd have to go to a tablet-like device. You really have that mixed world now, long battery life, all of that.
You combine that with Windows 10, it's a very unique situation that we believe is being set up for 2016, where you have a new operating system. The operating system's pretty strong, you combine that with a very strong product, and you have a combination of what we believe is a great environment to get this business to be a growth business. As I said, what we've done is we've been very careful as we forecasted next year, and Stacy will talk to you about that, and as we've looked at our strategy for this business to not be dependent on this being our growth engine. Additionally in that client is the mobile space. We said that we'd hit over 40 million tablets last year. We achieved that number. We said that this year we'd grow at roughly what the market did.
The market of tablets actually declined this year. Our guess is right now we'll end up shipping somewhere in the 30 million units. Right now, Q1 through Q3 volume is sitting at about 26 million units so far this year. We're going to ship less tablets this year than last year. To me, that's reflective of the business. The tablet business is just not a growth business. It's a declining business now. More importantly, we said that we were going to go and take the additional spending, that 2014 was a year that we needed to get into the market, and that 2015 was a year that we needed to start to make the improvements in our profitability. What we went out and did was committed $800 million. I can tell you with very high confidence we'll exceed that $800 million number. Stacy will actually come up here.
At the investor meeting, I always make sure he gets all the good fun messages. He'll give you the actual number, but I think you'll be impressed and surprised that the team's done a fantastic job. He'll also give you some insight into next year as well. We're committed to continue to improve the profitability of this business. As Andy said, we understand that is our job. We're in these businesses to make money. Phones we've been very careful with, right? You can get into the phone business and lose a lot more money than tablets if you go in there with either not the right products or the right partnerships. What we've done is really started to build a partnership with two very key partners, Spreadtrum and Rockchip. Those are beginning entry into phones. You can also buy an Asus Zenfone.
I can tell you my daughter uses one. You can buy them at the AT&T store, Intel Inside, and it's actually a very good phone. I want to talk about how Intel is becoming really stronger through this diversification. I thought these two charts were a pretty nice representation of that. What you see is that if you take the non-PC revenues, the DCG, IOTG, and the memory groups, and you take a look at them, they are now 40% of the revenue, and the PC has come down quite a bit. What's more important is they're growing at a faster rate than the PC declining. What's equally as important is if you take a look at the operating margin, we're now 65% operating margin as these groups. This is a combination of their growth businesses. They have this shared IP. They benefit from Moore's Law.
All of those pillars of strategy that we talked about, Moore's Law, integration, shared IP, are what allows these to not only grow, for Intel to have a very strong presence, and for them to have this kind of operating margin. We're very proud of this transition that we've really been driving over the last couple of years. Let's talk about those growth engines. I think that's really the most important topic we can talk about today. We talked last year about really three growth engines, data center, IoT, and memory. What's really important is that together, if you take a look at these, they feed upon themselves, and this is important. I see it every time I go visit a customer, and I'm in doing a discussion on IoT. I end up talking to them about our memory footprint.
I end up talking to them about their cloud needs, their data storage needs, their big data needs. All of these things end up. It's amazing to me how I never end just one discussion with a customer. I'm always in with this combination, this virtuous discussion about all three of these. We believe that our strategy is really built on these three things feeding each other. I'm going to walk you through just what that means and how they do that this morning. Let's talk about the cloud first, and we'll just start with the data center and the cloud. The first thing is that if you take a look at today's cloud, today's cloud is really what I call a people-driven cloud. Most of the data is either from people or about people.
It's your emails or your Facebook posts or your Twitters or whatever, or it's the data that's about you, your sales data, your health records, your HR data. It's really driven by you. It's input by people. It's a people-oriented structure. Our belief is that the cloud of tomorrow will be much more driven by things. The multiplication, and just stop and think about. I'm going to give you an example here in just a second, but you can stop and think about the number of people in your house versus the number of things that could be connected. The average house has maybe four people in it. You start taking a look at all the thermostats you have, all the IP cameras you could have for your security system, all the refrigerators and other things, vacuum cleaners, yes, they're coming connected as well in the future.
All of these things you start looking at, you start looking at it's a multiple number of people that are in the house. The data rate of those devices ends up being quite higher. They're sending data every so many seconds, every so many milliseconds, in some cases. You can't type that fast. We thought we would pull one example that we are actually doing with some partners. We're supporting a clinical study of 500 healthy volunteers, so these are not people who are ill, these are healthy people, in conjunction with a cardiovascular research group. The interesting thing is they're doing 24 by seven, they're wearing things like Basis watches. They've got scales that report up their daily weight, so you have to be a little gutsy to be in this study, if you want to enter your daily weight.
500 people are generating 3 million data points a day. There's no way you could type that fast, right? It's a great example of the multiplicative effect and just how fast this cloud is going to grow. One of the questions I get as I go out and visit many people is the cloud growth done? Are the number of Facebook locations and posts and all, are we kind of reaching a steady state? I say no. In fact, the explosion's just beginning because it's going to transition from a people-oriented cloud to a thing-oriented cloud, a machine-oriented cloud, and the multiplier is much, much higher. This just simple example, for me, is a great one to demonstrate that. 500 people, 3 million data points a day.
We think of the data center, though, as much broader, and no longer can you think of it just as a data center, as a singular entity. It's really about cloud, enterprise IT. You have to look at each one of these segments, comms, and the service providers and high-performance computing. Inside comms and service providers are networking and storage and some other things. You really have to take a look at each one of these segments. Diane's going to come up this morning and really walk you through the strategy of each of these segments. You have to think about it in that structure. If I take a look at the cloud, the cloud, we believe, is just one of the best growth areas that we can think of for the data center. In 2013, the cloud was 20% of DCG's revenue.
We expect in 2016, it will exceed, and pass over, the enterprise section. The data center has always been in the classic form, the enterprise. The traditional companies like Intel, having their large server rooms, keeping their large databases of factory and employee and design work, whatever your company was, that was always the fundamental area, the largest segment of this business. By next year, the cloud should pass over that. It's another example of just as the cloud continues to grow, it'll become a more and more significant portion of our business. The other area that is interesting is the connection between these devices and the cloud. This is another huge growth opportunity for us. That networking section, both in the data center side and at the thing side or the connected device side through the gateways, is another growth area.
The amount of intelligence that's occurring in both of those segments, both of the networking at the data center and then the networking down towards the device itself, is continuing to grow, and it's an opportunity for us. What's happening is that as that intelligence requirement's growing, because the amount of data is growing and they want to filter the data, they want to have analytics occur at the site, all of that stuff that's occurring there is a huge opportunity for us to come in with our general-purpose CPUs and really lower the cost, improve the capabilities, and drive a synergy across their systems. This is a segment that we have less than 10% market share in so far. It's an area that we're growing. It's one of our fastest-growing segments, but it's also one of our lowest market shares.
As we look forward and we look at the data center business in general, it's another huge opportunity for us to go and really capture market share and growth. I'd like to now shift a bit to the second part of that virtuous circle, that's the Internet of Things. As Andy said, we've been in the embedded business, gosh, many of us remember about 25 years ago when we were making all kinds of embedded devices. What's new and what I tell people is different about embedded is if you took a look at those embedded devices 20 years ago, 25 years ago, it actually had quite a bit of intelligence. Whether you were building a fish finder or a brake controller or an ATM machine or whatever it was, there was quite a bit of intelligence at those points of sales and all of those devices.
What's different in the Internet of Things, and what makes it, is that connection back through the network and into the cloud. You need to have, and this gets back to why is the connectivity and mobile so important to our long-term strategy, is that connectivity needs to happen through multiple channels. If you're sitting on a system that's sitting out on a pump in some place in Texas in the middle of nowhere, you need a 2G or 3G connection. If you're sitting inside a building like this, you may use Wi-Fi, right? If you're sitting in a mobile device, a car or a robot or a drone, you want to have the ability to transition through all of those connectivities based on the performance that you're searching for. You really need to have all of those. That's why connectivity is so important to Intel strategy.
That's what separates Internet of Things from the classic embedded world. It also means that you need to have much more highly integrated and much more of the IP in order to be able to do this. You have to, depending on the device, have sensors, you have to have AI engines, artificial intelligence engines that can understand motion and activity. All of those things are critical for these IoT devices. We believe that we're unique, and the real power that we bring to this section is that we have everything from end to end. We have Curie modules that can go down to the lowest level of things, up through Atom. That's typically what we see in the things space. If you take a look at most of the managers here today, they're wearing their TAG watches. Those have Atom-based products inside them.
If you take a look at some of the pumps in our factory, they have Quark-based products in them. All of those systems have sensors. They have a certain amount of artificial intelligence in them, understanding those motions. We then have the ability to go through the network, and then we can build out their cloud infrastructure as well. That's another example where almost every time I sit down with an Internet of Things provider or a customer, and I'm trying to work out a platform with them, I end up having the end of the discussion be, "Now, what are we going to do in the data center?" It's an important part of this virtuous cycle . We want to be there. We want to be a part of this.
We need to lead it because it's part of this virtuous cycle that's going to push the data up into the cloud and feed the data center. That's our Internet of Things strategy and why we're there and how important this is. I want to talk about the final piece of this virtuous cycle , and that's memory. As Andy said, 2015 has been a great year for memory. We believe if you take a look out over the horizon, memory's going to continue to really be a profit and growth generator for Intel. The improvements in memory technology are truly enabling faster and more efficient processing. If you take a look at our processor getting so fast now, that we really have to make some fundamental changes in the memory architecture.
If you just take a look at 2015, the revenue from our SSDs, most of the SSDs we sell are enterprise class SSDs, grew double digits when compared to the previous year. This is a strong growth engine. What we're really looking forward to is some of the phenomenal architectures that are coming in the future. We've talked about our strong partnership with Micron and the development work we've been doing there over the last few years. We've got two technologies that we believe are really driving the growth engine of our memory business out over the horizon. The first is our 3D NAND. 3D NAND provides us high capacity, low cost using the current architectures of NAND cells. It really allows us, we believe, an advantage over the competition in that density and cost cycle.
The one we're really excited about, the one that many of us have been working on for 10 plus years, though, is the 3D XPoint technology. You've heard it 3D XPoint, you've heard it Optane. Optane is the platform name. What that really does is bring a new class of memory, and I think that's important. A lot of people keep asking me, "Is this a NAND replacement? Is this a DRAM replacement?" I keep reminding everyone that this is a new class of memory, and we're really excited about this. We believe it's probably the most significant change in memory in the last 20 plus years. We're going to bring Optane, and Rob's here today. You'll be having lunch with him. You can spend a lot of time if you really want to get into the details.
I'm impressed at how he's a product guy who's become a factory guy, so in my mind, he's finally earning his pay. He's really, I think he's got a great business strategy. What's important about this is, this isn't a product that's 10 years out still or five years out or two years out. We're going to have products in market next year on Intel Optane technology. What this does is truly generate, as I said, a new class of memory that's really between NAND and DRAM, right? It's hundreds of times faster than NAND, and much, much cheaper than DRAM. It allows opportunity to re-architect how memory and storage is thought of. It's important to not think of it as a replacement.
It's important to think about it as a re-architecture, a chance to change the way we think about memory and storage. You also saw us make an investment in our Dalian factory around memory. This is important. This is probably one of the first memory investments we've made in an Intel factory in quite a few years. I probably couldn't even tell you how far back it was. I actually worked in one of our original memory factories when I got out of college, so that's probably how far back it is. We said that we'd invest up to $5.5 billion. It is initially built out to do 3D NAND, so it's really bringing that into high volume and driving the cost down there. We've built it such that in the future, if we want to do 3D XPoint there, we can.
We'll do that just based on capacity and as the industry requires the capability. I've walked you through the virtuous cycle. I've walked you through the three growth engines of that virtuous cycle. Hopefully, you see how they're connected, right? That faster memory and faster storage will really fuel the data center and the cloud. It also has great IoT applications as you start thinking about artificial intelligence and think about automated driving, autonomous driving, the ability to have all of that at a much faster rate right next to your CPU is a shift in the performance of those devices. I want to talk to you about, in the last few minutes here, our two investments that we're really building on that are technology-driven, and that is the Altera acquisition and then our wireless investments.
Altera, we believe, is going to really feed and basically speed up this virtuous cycle. Think of it as an accelerator of this whole cycle of components. Every one of these devices, whether it's in the data center and it's a big data application because you've built out all of this cloud, and now you have all this data and you want to do something with it. That health study is a good example. I've got all this data on these employees now, and you want to be able to do something with 500 people. What do you do with that? You want to start to put big data applications against that and understand as their weight shifts, as their activity moves, how does their health shift, right? You want to understand all of that, and you really need to do a big data application.
What you find is that if you put an FPGA next to your CPU or, in fact, eventually embedded into your CPU, that you can take those algorithms and put them basically, instead of off the board in memory or in storage, you can put them right on to silicon. Performance improvements can be 25%-30% improvement. Just like any other improvement, you can either use that for lower battery or longer power, lower usage of power, or you can use it for higher performance at the same power. Okay?
When you talk about IoT, you put this into that, you put an FPGA onto a CPU for IoT, and you can take all of the autonomous driving algorithms that you're using, all of the sensor fusion that you're trying to take, all the data from the sensors, plug those through the algorithm that then says, "Put the brakes on, steer left, do whatever you need to do." Right? Those algorithms can now be in silicon and perform at a much higher performance rate than they can if that system has to go off silicon, onto the board, and back. This, we believe, will be an accelerator of this virtuous cycle, both at the data center level and the IoT level. That's why we made this investment. Along with the fundamental business, is fed by Moore's Law. There are three fundamental reasons we did the Altera acquisition.
The base business is fed by Moore's Law. We believe in Moore's Law. We believe we have leadership in Moore's Law, and we're going to continue to lead. Second, it feeds IoT. I gave you the autonomous driving as an example. We can go through many other examples. Third is the data center. Once you have all that data up there, you want to be able to do big data applications. FPGAs are great usages of that. The last section I'm going to talk about is our investment in wireless. Hopefully, it's pretty obvious based on the conversation. If you just take a look at IoT, we believe those devices need to have every form of connectivity and the ability, in some cases, to move through that connectivity spectrum as they travel, right? When you start talking about robotics and drones and things like that.
These devices have to be connected. They have to stay connected. They need high-speed connections to the cloud in order to download data. We think that's important. Also, there's a transition coming. 5G is coming, and 5G is an opportunity for Intel to work with both the base carriers, the Ericssons, Huaweis, Nokias, to build the base stations and build the infrastructure of 5G. What we're finding is you need to be in the other end of 5G in order to be credible and to have a presence in the standards bodies and in the customers. We think 5G investments and IoT are just two examples. That, along with our ability to get into the mobile business as well. As we said, we're being very careful and cautious about that in order to do it in a profitable way.
We really believe that we are positioning Intel for the future of mobile by having these leadership positions. I've got 2 minutes and 25 seconds left. I'm right on time. I want to just wrap up our conversation and end with my commitment. I've tried to end each session of these investor meetings each year with a commitment that you should be able to walk away with and hold us accountable to. The 4 key takeaways and commitments I want you to walk away with is that we truly believe that this strategy of having a base business of the client that's stable, profitable, and innovative is core. The virtuous cycle of these growth engines of IoT, the data center, and memory all being connected through a wireless strategy is truly going to be the growth engine.
You've seen that in the data center case, it's growing now, so it's a higher amount of our profitability and an increasing amount of our revenue as a corporation. We are seeing strength in client. We believe we put the right investments in place and have the right innovation. We've worked with our partners like Microsoft and the OEMs to bring great form factors, Win 10, Skylake, with Kaby Lake next year. We believe the innovation engine in that space is strong. We've not required that for our corporate strategy of growth. We've tested it to high single-digit declines and still believe we're a growth company in that model.
You've seen this DCG, IoT, and memory virtuous cycle. It's important that we have good positions in all three. When we think about them, we think about them as a larger platform. In almost every customer visit, it's never just a singular discussion. It ends up being a discussion on all three of those. Just my last commitment is around our leadership team. We believe that we continue to grow. You see that the leadership team of Intel today is quite a bit different. Roughly about 40% of my direct staff is new to the company from outside of Intel. We've really transformed Intel into what I believe is a modern-day, very aggressive and transformative company. You have our commitment to keep driving that transformation. As the industry and the world changes, we're going to continue to change as a company.
With that, I think I'm done. I now have 16 seconds. Oh, no, I'm 16 seconds over. I apologize. I'd like to invite my friend and the person who knows Moore's Law better than anybody other than Gordon himself, Bill Holt, on stage to talk to you.
Good morning. I'm not sure whether I should be pleased how many times people have said what I'm going to talk about today, but there certainly were a number of them. Hopefully, I won't disappoint you too much now that you see what I'm really going to say. It's very nice to see Gordon's video again, but it is sort of intimidating to follow an icon like that and talk about the same topic. Hopefully, I also won't embarrass him in that sense. Let me get started. Make sure I know what's what up here. I have four things I want to talk about today. I want to give you an update on 14 nanometer, and most importantly for me, how we're doing on continuing to deliver on the cost per transistor trend.
That is the fundamental economic law that Gordon was referring to, that Andy referred to. Without that, there really is no justification for continuing to advance technology. There's lots of other good things that come from it, the amazing improvement in performance. If that didn't come with a cost reduction, we wouldn't have been able to afford to do this. I want to show you that. I want to take a little different look at the economics of Moore's Law and take a little bit of thinking about what does it take to be able to afford to continue doing this? Clearly, there are less people doing this than there used to be, and that obviously relates to the affordability of continuing to do this. I want to look at that for a moment.
I will take some time on competitiveness, and then very briefly, I want to just give you some indication of some of the things that are coming down the pipeline, things that give us confidence that, in fact, there is a future in Moore's Law, in the economic aspect of it and what we're going to be able to deliver over time. Okay. Last year, I showed you this chart on how the yield was progressing, and if you look carefully at the chart, we were predicting that early this year, we would have achieved equivalent yield with the 22 nanometer technology. We also noted that our 22 nanometer technology is probably the best ever we've had for yield. If I update that chart for today and stretch it out for the rest of this year, what you'll see is we didn't quite achieve those goals.
We have been showing steady progress, in fact, in improving the yield, but we have not yet closed the gap to 22 nanometer. That gap results in a number of things. It results in a tightness in the supply, though we have been meeting all of our commitments in supply. Some of the recent reports about shortages are because people are actually ordering more products than what we had predicted, and that obviously is a good thing, but it is a shortage. It also means that the products that we were producing are somewhat more expensive, and Stacy will give you some data on that.
The good news is we have been seeing improvement, we have good data to suggest that we will in fact close this gap, that we do not see a sustained delta occurring that would in fact remove some of the value from continuing to do scaling. We see that while it's been harder than we thought, it's taken us longer than we would have liked, we do not see that long term, there will be a difference between what we were able to achieve in the past and what we expect to achieve going forward. This is another chart I showed you last year. It's probably the most important chart as far as I'm concerned.
The cost per transistor is the fundamental metric that we need to drive down and to continue to make each new technology, one affordable and also valuable to pursue. This year, I want to add onto that chart some estimates of where we think our 7-nanometer technology will be. Last year, I hinted to this, but I wouldn't actually put anything down, and I'm not going to put an exact point down because there is still some degree of uncertainty in these various parameters. On the left-hand side, you see the increasing trend of wafer costs, and this is cost per millimeter squared, but the same shape with wafer costs. This is coming due to some of the items that Brian mentioned. The technology is getting more complicated.
We have not seen a fundamental improvement in our patterning capability, and that has led to an explosion in complexity. The net result is that the cost per wafer is going up faster than it had traditionally. As I said last year, our response to that is to improve the scaling faster than traditional. When Gordon mentioned that basically we're seeing a 2x increase in the number of transistors per generation, and that's fundamentally this curve. What we've done is we've actually focused on that to a higher degree and used increased scaling to be able to offset that increased wafer cost. The net result then is the curve of the combination of these two, which is the cost per transistor. That's the fundamental building block of all of our devices. That's what drives the cost reduction on products.
In fact, we saw very good results on 14, and we're projecting even better results on 10. We can see enough of 7 now that we expect that it will, in fact, continue to stay below the curve. Exactly where below it is some of the things we'll work out over the next two to three years. We have defined enough of the technology that we do know that we expect with high confidence to be with inside that oval on the right-hand side. Let me move on and talk about Moore's Law. Many of you have seen us talk about this. You've seen Andy talk about this, as Brian mentioned. I'm going to talk about it in a slightly different way today.
Fundamentally, Moore's Law enables us to take the same functionality and make it in a smaller and smaller die, and as I showed on the last chart, a cheaper and cheaper product. You can use that benefit to continue to use a product that's basically the same cost of providing increasing amount of functionality. Those are basically the two choices. Over the last few years, we've had a model, Brian referred to it, that we've used it in the past to show you why it's better to go faster. We've built this model, and we have a fair amount of confidence in it. It allows us to take a look at what is it that we have to spend over a 10-year horizon for our manufacturing development. We can then perturbate that model with varying parameters.
When we last did this in 2011, we showed you that if we were to look at what it's going to cost us in this model over the next 10 years, we'd spend about $104 billion in the combination of red, which is the developing of the new process, and the blue, which is the manufacturing cost of the factories and actually building the products. What we show you is that there's a substantial benefit to continuing to do these new technologies as opposed to remaining on this old technology and attempting to build the same products. Now, this is a theoretical model. You obviously can't build the same products on a 10-year-old technology. You can calculate what the cost would be, but you wouldn't have a product that you could actually make unless you just made the same product.
If you continue to increase transistors, make them faster, if you were somehow able to do that, they would not fit in the package, they would not be able to fit in the power envelope. From an economic point of view, you can make this comparison, and you can say the benefit of, in fact, advancing technology for Intel in 2011 estimate was over $200 billion. Okay? This is the model that we're using. We've updated this because trends do change. In 2015, we went back and reran the model. What you see are things that we've talked about. The cost of development has gone up. The red bar is, in fact, larger. That's because of the same thing that's driving up the cost of wafers.
The complexity is exploding on us, and the number of tools it takes, the number of wafers it takes, the amount of people it takes to develop the technology are, in fact, increasing. The amount of money you have to spend on that new technology development is, in fact, larger. You also see here a slight change in the case for not doing any new products. That comes from a slightly different projection on what the unit growth rate would be. You basically see the same fundamental benefit that comes from advancing technology. What it did tell us is that you can perturbate this model, and there are things that can actually make visible changes. Then we set out to look at what are the most important ways that you can, in fact, test this model. The first one is, in fact, lower unit demand.
As I said at the beginning, there are less people doing new technology. Why is that? Well, there are less people who can afford it. What is the fundamental driver of whether you can afford it or not? Whether you have enough revenue to, in fact, justify the investment. If you perturbate unit demand, it's very likely that you can probably affect the robustness of this development model. The second one, as I already mentioned, higher development cost. If that red bar got sufficiently large, then it would swamp out the benefit that you would get on the manufacturing side. The third one is cost per transistor.
If we're not capable of delivering a sufficient advantage or a sufficient improvement in the cost per transistor, then in fact, the blue part of the new technology model will, in fact, grow, and it will again become approaching the size of the no technology improvement case. Let's go look at these. If we looked at the original 2011 model, we were projecting a given growth in demand. We began to change the parameters to the model, and we looked at what if it was flat. I'm not saying it's going to be flat. Obviously, Brian has showed you a lot of reasons for what the real projections will be, we looked at this, and there was no problem in terms of those two bars still being very different.
In fact, when we took the model to the limit to find out when would, in fact, the cost of developing a new technology and deploying it equal the cost of staying on the same technology in our theoretical model, you have to, in fact, have a decrease in unit demand of 25% per year. Now, the background of this graph is wafer starts, and you can see that this is somewhat of an absurd case, because the number of wafers that we would be building out there in time is approaching zero on this scale. The conclusion from this is that there's a very large margin for us in terms of our ability to afford to actually generate enough volume to continue to benefit from advancing Moore's Law. The second one I also mentioned was increasing R&D cost.
In our original version in 2011, we looked at the historical rate, and the cost of new technology development was increasing at about 10% per generation. This was a metric that we actually used to try to control our spending, and we tried to keep within this window. In fact, that isn't the case, and if we look at what our projection was in 2015, and we look at the more recent history, it's closer to 30%. That's what I said. That's what made that red bar get bigger. On this scale, if you wanted to make the red bar basically eat up all of the benefit, you'd have to grow to a rate of 190% per generation. I think the most significant thing is that, if our revenue was projected by the model, you'd have to be spending $80 billion a year on technology development.
I think long before that, Brian would have found other people to go work on this problem. I think it's fair to say that the increasing cost of R&D, at least for Intel, there is an increasing cost in R&D, but it's not going to be the thing that stops us from pursuing Moore's Law. Let's look at the last case, cost per transistor improvements. This is a stylized version of that earlier chart that I showed you, and you see the decreasing trend. Historically, this trend has been about 0.7x cost reduction per transistor per generation when you combine the increasing wafer cost and the die scaling. On this graph, I've shown some of the recent numbers, just to give you a context that we think we will achieve on 14 and 10.
Again, we adjusted the model until there was no benefit, so the two bars were equal. On this scale, we'd have to be at a 0.86 cost per transistor improvement trend as opposed to the 0.7. This one is the only one that actually is somewhat close, and this is still a fair distance away. This is a substantial amount of reduction, but the good news is that this is the one that we most directly control. This is what we're paid to do. This is the one that doesn't depend on Intel's volume. It doesn't depend on how much it costs us to do this. It just depends on whether we have enough good ideas and enough motivation to, in fact, focus on and drive this trend.
While this could challenge the economic benefits, we don't see any indication in the foreseeable future that this will, in fact, be a problem. I think our conclusion from this is that while economics likely will be the biggest limiter, and it is the limiter that is, in fact, reducing the number of people actively pursuing Moore's Law, based on our size and the benefit we have of our scale, none of these factors will be the ones that cause us to have any problems. Okay, let me shift gears and look at competitiveness. I'm going to basically focus on scaling or density because that is the important part of cost per transistor that we actually, again, can control. Last year I showed you this chart.
It was based on two very simple numbers. We used this chart last year because data was increasingly becoming available both on what people were doing and what we were doing and on projections going forward. It's not a very sophisticated metric, but it does generate the shape of the curve and give you a pretty good understanding of what you would expect for the entire technology. Actually, the year before, we took a slightly different tack, and we made an attempt to try to estimate what will we think the full chip scaling would be for a roughly equivalent kind of product as we progress through the technologies. What you see is the shapes are very much the same. After the investor meeting, we actually got more information on some of the competitors, so we adjusted the line slightly.
In fact, the shape is very similar to the previous one, it basically gives the same message. It's hard to get data for this graph because this requires you to know almost entirely what the technology is going to be or for someone to tell you what do they predict their scaling to be. What's happened recently now is we actually can look at products from Intel and from our most important competitors at the 14-nanometer generation. The easiest one to look at is to go look at the Apple products because we have a number of examples there of products that are quite similar. On here we're showing a normalized density for the Apple A8 at TSMC, the A9 at TSMC. Then that same basic product at Samsung. Those are all available.
We've actually bought phones, taken them apart, we, like everybody else, have been looking at the die to see what it actually does, gives us a good indication. Now, beyond just being able to look at two pitches, we can now look at many aspects of what the technology provides. There's also this base number is basically simple to calculate. You just have a die size, you can count roughly how many transistors are there. If you make that same measurement on an Intel part, on our primary parts that we've been selling, our high-volume products, what you'll see is a number that actually looks exactly the opposite of what I just showed you. What it shows is that the raw density as measured by just area and transistor count actually looks worse.
Since I'm sure that you've seen various reports of these kinds of numbers, you should be asking yourself, why did you show us that previous chart? What is going on here? You said that you would be better, yet when you look at the raw data, it's not. It's worse. Well, we've said in the past that that name that we had generate with technologies is not always the same, the same goes for the composition of die. I'm going to take a little diversion and talk about, what are the various kinds of blocks that you typically will see in a product? Now, what you find is that depending upon what the composition of the product is, you can make dramatic changes in what the transistor density or the transistor per millimeter are.
Now remember, for an entire technology generation, we're talking about an improvement of maybe 2X. If you change from most of the die being memory to most of the die being logic, you can make a 3X or even 5X, depending on what cell type you choose. What it tells you is it's very important if you're going to do a legitimate comparison to understand, are these products actually similar? Are they aimed at the same market, are they constructed from a similar kind of basic design blocks? Now if I go back and look at those same five products, what you will see is they are not really that much alike. Of course, the Apple parts are very similar, the Intel parts are very similar, they're addressing fundamentally different markets.
If you look at the details here, the one that is most dense, the red, which is the amount of on-chip cache memory, is almost three times higher, which adds to an almost 6x increase in those transistor counts, as opposed to the area that's allocated towards doing logic. If we take this data, we now know the composition of the parts, we now know the relative density that we can get by actually looking at the layouts. We go back to that previous chart and adjust it for this. What you see is that, in fact, that improvement is present if you were to build a similar part. These are no longer the parts that they were. They are parts that have the same distribution of circuit types. I'm sorry if this is complicated, but it's actually quite important.
If we now take this data and apply it back to that original graph, we can now actually update the points at 14 nanometers. Here's the graph that I showed you. If we now take the information from analyzing the actual released products, it has a slight movement. What you see is a number of factors that now we can validate with actual products. In fact, most people's technologies, as they moved to the 16/14 node, did not really provide any increase in density. That's evident in the raw data as well as in this chart. Also, you can see, just to validate the legitimacy here, is that the Samsung part is slightly smaller, has slightly higher density than the TSMC version of the same product.
When we take and normalize the parts of the technologies for similar composition, in fact, the data point for Intel's 14 nanometer does show that we have a substantial lead. Recently, people have begun giving more projections on 10, so we're able to update that data as well. Our projections for what other people would do and their more recent statements are very much aligned, and they show that this increased density advantage will continue into the future. Let me move on then and look a little bit out towards the future. This is a busy chart. It's intended to be busy. The point of this chart is that there are lots of very interesting, exciting, and good ideas that are being researched, both at Intel and universities, at our fellow travelers.
The size of the chart, or the number of things on the chart, is what is important. This is not to imply that all of these will actually prove to be beneficial. The fact that in spite of the concerns about the difficulty in advancing Moore's Law, there are a lot of really good ideas, and there's no question that you're going to have to make substantial innovation to continue. We've all agreed that the days of very simplistic scaling are long over, and the only way you can continue to drive economic benefit and all of the rest of the things that come with Moore's Law is, in fact, through significant innovation.
This shows that there are efforts of many different directions, either in increasing the function that you would get from a single device, simple scaling or scaling by using new materials to continue to reduce the size of things, or looking at innovative ways of doing integration. Probably the best for that idea that actually exists today is, in fact, 3D NAND. Stacking die in terms of die stacking doesn't advance Moore's Law. It certainly helps form factor, but it doesn't reduce cost. 3D NAND, where they've, in fact, on innovative ways to make many bits with a similar number of processing steps by stacking the bits up instead of across does in fact provide the kind of cost benefit, and it basically produces a cheaper cost per transistor for that particular kind of memory.
Applying those kinds of ideas to other areas and looking at ways to in fact stack logic devices. Some of the things that will help with our basic equipment set in terms of more selective depositions or even eventually the kind of holy grail of directed self-assembly, which would help eliminate the needs. All of these are being studied. Which ones will actually come true? I wouldn't tell you if I knew, but I don't really know. The fact that the page is full, and these ideas are all credible, and they're worth pursuing tells us that the future is actually quite robust. If I summarize, on 14 nanometer, the yields are improving. We do not see a fundamental problem in the ability that scale technologies will no longer be able to achieve the same yield levels.
We have a very high bar in our 22 nanometer technology, which is and continues to be the highest-yielding technology we've ever had. As you saw from the chart, while we are a little behind where we wanted to be, we are in fact closing that gap, and we don't see any reason that we will not be able to achieve those same levels. We do have some demand increases. Availability is not everything we might like, but we have been consistently able to meet our commitments, and the shortages that do exist are generally due to the fact that people like these products. Brian said Skylake is probably our best introduction of a product ever. It is a very good product and has very high demand.
If I look at the cost per transistor, it's clearly getting difficult, and it takes a lot of innovation and a lot of work. The progress really is promising, and the projection out in time also looks very good. If I look at the economics of Moore's Law for Intel, this is very solid. This is very expensive. It's very difficult. It takes commitment from the company to continue to invest. We have the ability to do that, and we project that, in fact, that investment will deliver a substantial benefit. When I look at the competition, basically it's exactly what we thought three years ago. It's kind of interesting that it is in fact exactly what we thought. Everybody has done exactly what they said they would do, and that's an indication of everybody being pretty much on track.
It doesn't change the fact that we project and we have good evidence that we will have a substantial benefit in density that can be translated into cost or functionality. Finally, when I look out into the more distant future, we've never said that we have much visibility beyond 10 years. 10 years is even getting kind of fuzzy these days. When you look at the kinds of things that you're going to need out in that time period, there's a lot of progress being made. While we don't know exactly which ones are going to come, we do have a high degree of confidence that the right kinds of technologies will be available to continue to increase the fundamental benefit of Moore's Law. Actually, I'm done a little bit early. That's very unusual. I limited my foils so that I wouldn't run over.
I guess that means you get to go have your break four minutes and five seconds early. Thank you.
Ladies and gentlemen, we will now take a short break. Please be back in your seats at 10 o'clock. Once again, 10 o'clock.
Thanks, everyone. Thanks, Bill. All right, we'll take a quick 30-minute break now. I'd like to invite you to join us out in the lobby, and we can mingle there. I'll again remind that we will be filing an 8-K, and we'll be issuing a press release momentarily. The outlook that's in that press release and 8-K will appear up on the screen here behind me. To those of you out joining us on the webcast, we'll be back in about 30 minutes. Thanks, everyone.
San Francisco, I was amazed at how widely and how positively it was viewed. We thought it would be a nice little event. It was probably millions of people followed it, thousands of stories in 39 countries. The overwhelming theme was how people were encouraged about how technology has changed lives and how it will continue to change lives. Having been there, it was actually one of those nice events you get to have in your career. When people think about Moore's Law, of course, we all think technology. It's really an economics law. If you read the paper, which flashed up there briefly, it's something I understand. I'm an economist. I was an economist, I guess. Now I'm a chairman. What's interesting is with our founders, each of them made the transition from scientist, inventor, to very good business leader.
It's incredible when you look at the history of Intel and how we have made those transitions, that's what we have to work on today. If you ask me what's most important for Intel right now, we focus on execution. We focus on getting results. We understand what matters is creating shareholder value. My view is we're on track. We've had a decent last year. I would acknowledge the last two years has actually tested the impatience at times of our employees, as change is not easy, and our shareholders at times. It's necessary to get to the place we want to get into the future. I think we have seen decent progress this year, and we expect to see continued progress in the year to come. The good thing is, I just get to say what I expect.
I now get to bring on the stage the guy who actually has to do the work. Let me introduce CEO Brian Krzanich.
Okay, good morning.
Good morning.
Is everybody excited, energized, ready to go?
Yes.
Let me see if I can get to my agenda. I thought what I would do this morning is go through a little bit of the 2015 results. As Andy said, we have some pretty good results, we think, relative to the market and relative to what's been going on. We'd spend a little time on the overall corporate strategy. We'll put it into context on the big picture. As Andy said, we've really spent the last two years really trying to refine that, put it into motion, adjust how spending and the organization is structured, and how people get things done here day to day at Intel to align to that strategy. I thought we'd take the segments apart, so the clients' foundational business, the data center, the IoT, and the memory, which are the growth engines.
I want to spend a little bit of time talking about the future investments. I thought we'd talk about Altera, and put a little bit of context around that, and then our investments in wireless. I know that's on a lot of people's minds. We feel like we've made a lot of progress there technically this year. Close it up with a summary. That's my agenda for today, and hopefully, it'll meet everything you guys want to know about the company. Let's start by taking a look at just how 2015 has been. We started with our commitment last year at this meeting and talked about what are we going to do. We said we're going to pursue Moore's Law with relentless fervor, and we continue to do that.
I think we all feel pretty good about our progress there, 14 nanometers. Bill's going to get up and talk. Part of the reason we wanted to keep Bill on the agenda and have a special discussion about that this morning is, you read the press, there's a lot of speculation Moore's Law is dead. What's going to happen is the termination, the yields. We're going to try and pull the covers off all of that today and really talk about that. We can tell you without a doubt that Moore's Law continues and is alive and well, like Gordon. By the way, I have to tell you, it was the 50th anniversary, and I had a great opportunity.
I happened to be in Hawaii with my family at the same time Gordon was there. I actually got to do that interview with him, and it was probably one of those special moments. I actually brought my daughters with me, and it was just great to see an icon like that. We say we develop products that enable the best computing experience. I think if you've seen the discussions on the web about Skylake, if you've tested one of the systems, if you have one of the systems, hopefully, you've seen the performance that we've continued to drive. Really, Skylake, we call it our best processor ever, the best improvements ever we've done in a processor to date. We really try to do a pragmatic, market-driven industry view.
We're not saying that the market's going to go the way we want it to go. We're really trying to understand the way the market's headed and move our products and our R&D that way. Open the foundry to companies able to utilize our technology. We've continued to make progress there. Bill will talk a little bit about that. Really we have to wait till our customers bring the products out for us to really talk too much about that. Create platforms for enterprise. This is what you're going to see, especially during Diane's talk. I'll talk a little bit about it. We really see that you have to develop a platform. You have to design the whole architecture.
As the cloud, we'll talk about this, as the cloud grows, you really have to think about the whole cloud infrastructure, and you can't just sell silicon. We're really focusing on bringing innovation to market quickly. There are several examples over the last years. If you take a look at Quark. Quark was a product that wasn't even on our roadmap a couple of years ago. It's out in a number of products now. One of the latest examples is Curie. We announced it at CES this year. It's already in market. We're really able to spin products at a much faster rate in some of these segments now. Growth in the data center, we'll talk about.
Our real foundation is that if it's smart and connected, it's going to work best with Intel, and we think we have a lot of really strong examples for that. Let's get right into that. Our results for 2015, the full-year revenue, the target is roughly around $55.2. You've seen our forecast for Q4, it's got a spread around it, so we'll be in the $55.2 range. Full-year margin of 62%. We're very proud of our ability to continue to maintain our revenue, but also our profitability. As you put those together, that gives you a full-year operating profit of about $13.5 billion-$14 billion. Throughout today, I think, as Andy said, you're going to see evidence that Intel is transforming and that the growth in revenue sources, and I'll talk a little bit about that later.
If you take a look at the mix of this revenue, though, the underlying performance of that has really shifted over the last three years, and we want to really talk about that today as we go through this. Let's go right in from that to the strategy and just talk a little bit about what our corporate strategy is. It hasn't changed a lot since last year, but we have refined it. We have continued to improve on it. You're going to see it. You see it in our agenda today. You see it in some of the time you're going to be able to spend, you'll see it in my discussion as well.
Our strategy is really based on three pillars or three fundamentals, and those are really that Moore's Law is critical and that we can continue to drive it, we can continue to lead it, and it differentiates what we do. The second thing is really tied to Moore's Law. What Moore's Law does is allow you to continue to integrate more, to continue to bring in more of the intellectual property into your product. What that does is lowers the cost, lowers the square footage of the product that you're trying to build. We see examples of this day after day in our lives, as products continue to get smaller, more compact, better battery life. That is this combination of Moore's Law and integration.
The last thing is this shared IP, which Andy kind of alluded to in his discussion, where it's really about we generate intellectual property at, say, the PC, where the PC is able to afford it is able to fund it, but we spread it everywhere. One of the best examples I can give you is RealSense. We originally created RealSense, which is our computer vision system, to actually unleash people. This part of this no wires, no connections concept. We wanted to unleash people from the keyboard and mouse of today. We really thought there had not been innovation in that space for quite a while, so we developed this computer vision model.
What we found as we pushed that IP into other areas is probably the largest areas of growth are going to be in places like robotics, in places like drones, in a variety, in industrial applications, automotive applications. These are all going to actually take that computer vision probably at a much higher rate than even the PC, and it's a great example where the PC funded this intellectual property, but it's being spread across a variety of technologies and business applications. We believe it's a space where it differentiates Intel again. I'm going to spend just a moment on Moore's Law because you've got Bill, and I think other than Gordon, there's probably not a better person to talk about Moore's Law than Bill.
For me, what I wanted to stand up here and tell you guys is that we continue to believe that Moore's Law is an advantage. It's alive and well. In our transparency efforts, we took a look this year, and we actually looked at 32 nanometers, 22 nanometers, and 14 nanometer transitions. We saw that we'd moved to roughly two and a half years. Actually, if you take the history of Moore's Law over time, actually, we have studied it for probably 25 years of history. It's floated up and down, and it wasn't until 10 or 15 years ago that we'd actually pushed it down into two years, again, as the lithography systems caught up to it. It's expanded out and contracted over its history of life.
What we saw was that the trends, the fact that we didn't have EUV, the complexity of the technologies, some of the new materials, the last technologies had taken two and a half years, and we didn't see 10 nanometers being that dramatically different. The lithography systems weren't going to change. The complexity was about the same. We really said, "Hey, we want to give you the best data we have. We want to adjust our plans and spending. We want to insert products and really be clear to our customers so they could bank on that commitment." That's why we adjusted it to two and a half years. Our goal is always to get it back down to that two years. It's always the most profitable. The faster you could do Moore's Law, and we've brought up discussions in Investor Days of the past.
You can go search them up. Andy did one of the best ones a few years ago, where we showed that if you could do Moore's Law at an infinitesimally small time period, it would be the most financially beneficial. You have to bring all the elements together that keeps us at, we think two years is probably about as fast as you could possibly do. We'll always be pushing back to two years, and trying to go for that, but we wanted to be transparent and clear. That's why we moved it to two and a half years. There wasn't really anything else in there other than that. Let's go to the foundations. The client business is what I believe, a strong foundation. You see the products up here on the stage, and you're going to hear from Kirk later today, who runs our client business.
We continue to make great progress in our SoC capability in this space. We're very proud of our latest product, Skylake. I want to tell you that our growth is not dependent on this business. It's a stable, good, profitable business. What we've done this year is taken a very pragmatic look at the industry. Internally, Kirk's got the direction. You need to grow. We've also taken a look at the market dynamics, the replacement rates that we're seeing in the market and said, "Hey, the probability though is we need to have our growth strategies not dependent on growth on this segment." We've tested this, as you take a look at our growth strategy today, to high single-digit declines even. We're comfortable that we can continue to grow this business, even with a high single-digit decline of the client business.
I want to take a little bit of a deeper look at how we look at the client business, though. We really see a strong tailwind as we move into 2016. We believe, as we said earlier, Skylake is truly our best product that we've ever produced, and it generates great performance improvements. If you take a look at the form factors and the products, and Kirk will be up here to talk about some of that, truly our partners and us have worked together to produce what we believe are just world-class products. The two-in-ones that you'll see are really. If you take a look at that, there's not a reason you'd have to go to a tablet-like device. You really have that mixed world now, long battery lives, all of that.
You combine that with Win 10, it's a very unique situation that we believe is being set up for 2016, where you have a new operating system. The operating system's pretty strong, and you combine that with a very strong product, and you have a combination of what we believe is a great environment to get this business to be a growth business. As I said, what we've done is we've been very careful as we forecasted next year, and Stacy will talk to you about that, and as we've looked at our strategy for this business to not be dependent on this being our growth engine. Additionally in that client is the mobile space. We said that we'd hit over 40 million tablets last year. We achieved that number. We said that this year we'd grow at roughly what the market did.
The market of tablets actually declined this year. Our guess is right now we'll end up shipping somewhere in the 30 million units. Right now, Q1 through Q3 volume is sitting at about 26 million units so far this year. We're going to ship less tablets this year than last year. To me, that's reflective of the business. The tablet business is just not a growth business. It's a declining business now. More importantly, we said that we were going to go and take the additional spending, that 2014 was a year that we needed to get into the market, and that 2015 was a year that we needed to start to make the improvements in our profitability. What we went out and did was committed $800 million. I can tell you with very high confidence we'll exceed that $800 million number. Stacy will actually come up here.
At the investor meeting, I always make sure he gets all the good fun messages. He'll give you the actual number, but I think you'll be impressed and surprised that the team's done a fantastic job. He'll also give you some insight into next year as well. We're committed to continue to improve the profitability of this business. As Andy said, we understand that is our job. We're in these businesses to make money. Phones we've been very careful with, right? You can get into the phone business and lose a lot more money than tablets if you go in there with either not the right products or the right partnerships. What we've done is really started to build a partnership with two very key partners, Spreadtrum and Rockchip. Those are our beginning entry into phones. You can also buy an Asus Zenfone.
I can tell you, my daughter uses one. You can buy them at the AT&T store, Intel Inside, it's actually a very good phone. I want to talk about how Intel is becoming really stronger through this diversification. I thought these two charts were a pretty nice representation of that. What you see is that if you take the non-PC revenues, the DCG, IOTG, and the memory groups, and you take a look at them, they are now 40% of the revenue, and the PC has come down quite a bit. What's more important is they're growing at a faster rate than the PC declining. What's equally as important is if you take a look at the operating margin, we're now 65% operating margin as these groups. This is a combination of their growth businesses.
They have this shared IP, they benefit from Moore's Law, all of those pillars of strategy that we talked about, Moore's Law, integration, shared IP, are what allows these to not only grow, for Intel to have a very strong presence, and for them to have this kind of operating margin. We're very proud of this transition that we've really been driving over the last couple of years. Let's talk about those growth engines. I think that's really the most important topic we can talk about today. We talked last year about really three growth engines, data center, IoT, and memory. What's really important is that together, if you take a look at these, they feed upon themselves. This is important. I see it every time I go visit a customer, and I'm in doing a discussion on IoT.
I end up talking to them about our memory footprint. I end up talking to them about their cloud needs, their data storage needs, their big data needs. All of these things. It's amazing to me how I never end just one discussion with a customer. I'm always in with this combination, this virtuous discussion about all three of these. We believe that our strategy is really built on these three things feeding each other. I'm going to walk you through just what that means and how they do that this morning. Let's talk about the cloud first, and we'll just start with the data center and the cloud. The first thing is that if you take a look at today's cloud, today's cloud is really what I call a people-driven cloud. Most of the data is either from people or about people.
It's your emails or your Facebook posts or your Twitter or whatever, or it's the data that's about you, your sales data, your health records, your HR data. It's really driven by you. It's input by people. It's a people-oriented structure. Our belief is that the cloud of tomorrow will be much more driven by things. The multiplication, just stop and think about I'm going to give you an example here in just a second, but you can stop and think about the number of people in your house versus the number of things that could be connected. The average house has maybe four people in it. You start taking a look at all the thermostats you have, all the IP cameras you could have for your security system, all the refrigerators and other things, vacuum cleaners, yes, they're coming connected as well in the future.
All of these things you start looking at, you start looking at it's a multiple number of people that are in the house. The data rate of those devices ends up being quite higher. They're sending data every so many seconds, every so many milliseconds, in some cases. You can't type that fast. We thought we would pull one example that we are actually doing with some partners. We're supporting a clinical study of 500 healthy volunteers. These are not people who are ill; these are healthy people, in conjunction with a cardiovascular research group. The interesting thing is, they're doing 24 by 7. They're wearing things like Basis watches. They've got scales that report up their daily weight. You have to be a little gutsy to be in this study, if you want to run into your daily weight.
500 people are generating 3 million data points a day. There's no way you could type that fast, right? It's a great example of the multiplicative effect and just how fast this cloud is going to grow. One of the questions I get as I go out and visit many people is the cloud growth done? Are the number of Facebook locations and posts and all, are we kind of reaching a steady state? I say, no. In fact, the explosion's just beginning because it's going to transition from a people-oriented cloud to a thing-oriented cloud, a machine-oriented cloud. The multiplier is much, much higher. This just simple example, for me, is a great one to kind of demonstrate that. 500 people, 3 million data points a day.
We think of the data center, though, as much broader, no longer can you think of it just as a data center, as a singular entity. It's really about cloud, enterprise IT. You have to look at each one of these segments, comms, and the service providers, and high-performance computing. Inside comms and service providers are networking and storage and some other things. You really have to take a look at each one of these segments. Diane's going to come up this morning and really walk you through the strategy of each of these segments. You have to think about it in kind of that structure. If I take a look at the cloud, the cloud, we believe, is just one of the best growth areas that we can think of for the data center. In 2013. The cloud was 20% of DCG's revenue.
We expect in 2016 it will exceed, and pass over the enterprise section. The data center has always been in the classic form, the enterprise, the traditional companies like Intel, having their large server rooms, keeping their large databases of factory and employee and design work, whatever your company was, that was always the fundamental area, the largest segment of this business. By next year, the cloud should pass over that. It's another example of just as the cloud continues to grow, it'll become a more and more significant portion of our business. The other area that is interesting is the connection between these devices and the cloud. This is another huge growth opportunity for us. That networking section, both in the data center side and at the things side or the connected device side to the gateways, is another growth area.
The amount of intelligence that's occurring in both of those segments, both of the networking at the data center and then the networking down towards the device itself, is continuing to grow, and it's an opportunity for us. What's happening is that as that intelligence requirement is growing, because the amount of data is growing and they want to filter the data, they want to have analytics occur at the site, all of that stuff that's occurring there is a huge opportunity for us to come in with our general purpose CPUs and really lower the cost, improve the capabilities, and drive a synergy across their systems. This is a segment that we have less than 10% market share in so far. It's an area that we're growing. It's one of our fastest-growing segments, but it's also one of our lowest market shares.
As we look forward and we look at the data center business in general, it's another huge opportunity for us to go and really capture market share and growth. I'd like to now shift a bit to the second part of that virtuous circle, that's the Internet of Things. As Andy said, we've been in the embedded business, gosh, many of us remember about 25 years ago when we were making all kinds of embedded devices. What's new and what I tell people is different about embedded is if you took a look at those embedded devices 20 years ago, 25 years ago, it actually had quite a bit of intelligence, whether you were building a fish finder or a brake controller or an ATM machine or whatever it was, there was quite a bit of intelligence at those points of sales and all of those devices.
What's different in the Internet of Things and what makes it is that connection back through the network and into the cloud. You need to have, and this gets back to why is the connectivity and mobile so important to our long-term strategy, is that connectivity needs to happen through multiple channels. If you're sitting on a system that's sitting out on a pump in someplace in Texas in the middle of nowhere, you need a 2G or 3G connection. If you're sitting inside a building like this, you may use Wi-Fi, right? If you're sitting in a mobile device, a car or a robot or a drone, you want to have the ability to transition through all of those connectivities based on the performance that you're searching for. You really need to have all of those. That's why connectivity is so important to Intel strategy.
That's what separates Internet of Things from the classic embedded world. It also means that you need to have much more highly integrated and much more of the IP in order to be able to do this. You have to, depending on the device, have sensors, you have to have AI engines, artificial intelligence engines that can understand motion and activity. All of those things are critical for these IoT devices. We believe that we're unique, and the real power that we bring to this section is that we have everything from end to end. We have Curie modules that can go down to the lowest level of things up through Atom. Right? That's typically what we see in the things space. If you take a look at most of the managers here today, they're wearing their TAG watches. Those have Atom-based products inside them.
If you take a look at some of the pumps in our factory, they have Quark-based products in them. All of those systems have sensors. They have a certain amount of artificial intelligence in them, understanding those motions. We then have the ability to go through the network, and then we can build out their cloud infrastructure as well. That's another example where almost every time I sit down with an Internet of Things provider or a customer, and I'm trying to work out a platform with them, I end up having the end of the discussion be, now what are we going to do in the data center? It's an important part of this virtual cycle. We want to be there. We want to be a part of this.
We need to lead it because it's part of this virtual cycle that's going to push the data up into the cloud and feed the data center. That's our Internet of Things strategy and why we're there and how important this is. I want to talk about the final piece of this virtual cycle, and that's memory. As Andy said, 2015 has been a great year for memory, and we believe if you take a look out over the horizon, memory is going to continue to really be a profit and growth generator for Intel. The improvements in memory technology are truly enabling faster and more efficient processing. If you take a look at our processor getting so fast now that we really have to make some fundamental changes in the memory architecture.
If you just take a look at 2015, the revenue from our SSDs, most of the SSDs we sell are enterprise class SSDs, grew double digits when compared to the previous year. This is a strong growth engine. What we're really looking forward to is some of the phenomenal architectures that are coming in the future. We've talked about our strong partnership with Micron and the development work we've been doing there over the last few years. We've got two technologies that we believe are really driving the growth engine of our memory business out over the horizon. The first is our 3D NAND. 3D NAND provides us high capacity, low cost using the current architectures of NAND cells. It really allows us, we believe, an advantage over the competition in that density and cost cycle.
The one we're really, really excited about, the one that many of us have been working on for 10+ years, though, is the 3D XPoint technology. You've heard it, 3D XPoint, you've heard it Optane. Optane is the platform name. What that really does is bring a new class of memory, I think that's important. A lot of people keep asking me, "Is this a NAND replacement? Is this a DRAM replacement?" I keep reminding everyone that this is a new class of memory, we're really excited about this. We believe it's probably the most significant change in memory in the last 20+ years. We're going to bring Optane. Rob's here today, you'll be having lunch with him. You can spend a lot of time if you really want to get into the details.
I'm impressed at how he's a product guy who's become a factory guy. In my mind, he's finally earning his pay. He's really I think he's got a great business strategy. What's important about this is this isn't a product that's 10 years out still, or five years out, or two years out. We're going to have products in market next year, on the Intel Optane technology. What this does is truly generate, as I said, a new class of memory that's really between NAND and DRAM, right? It's hundreds of times faster than NAND, much, much cheaper than DRAM. It allows opportunity to re-architect how memory and storage is thought of. It's important to not think of it as a replacement.
It's important to think about it as a re-architecture, a chance to change the way we think about memory and storage. You also saw us make an investment in our Dalian factory around memory. This is important. This is probably one of the first memory investments we've made in an Intel factory in quite a few years. I probably couldn't even tell you how far back it was, but I actually worked in one of our original memory factories when I got out of college. That's probably how far back it is. We said that we'd invest up to $5.5 billion. It is initially built out to do 3D NAND, so it's really bringing that into high volume and driving the cost down there. We've built it such that in the future, if we want to do 3D XPoint there, we can.
We'll do that just based on capacity and as the industry requires the capability. I've walked you through the virtuous cycle. I've walked you through the three growth engines of that virtual cycle. Hopefully, you see how they're connected, right? That faster memory and faster storage will really fuel the data center and the cloud. It also has great IoT applications as you start thinking about artificial intelligence and think about automated driving, autonomous driving. The ability to have all of that at a much faster rate, right next to your CPU is a very shift in the performance of those devices. I want to talk to you about, in the last few minutes here, our two investments that we're really building on that are technology-driven, that is the Altera acquisition and then our wireless investments.
Altera, we believe, is going to really feed and basically speed up this virtuous cycle. Think of it as an accelerator of this whole cycle of components. Every one of these devices, whether it's in the data center and it's a big data application because you've built out all of this cloud, and now you have all this data and you want to do something with it. That health study is a good example. I've got all this data on these employees now, and you want to be able to do something with 500 people. What do you do with that? You want to start to put big data applications against that and understand as their weight shifts, as their activity moves, how does their health shift, right? You want to understand all of that, and you really need to do a big data application.
What you find is that if you put an FPGA next to your CPU or, in fact, eventually embedded into your CPU, that you can take those algorithms and put them basically, instead of off the board in memory or in storage, you can put them right on to silicon. Performance improvements can be 25%-30% improvements. Just like any other improvement, you can either use that for lower battery or longer power, lower usage of power, or you can use it for higher performance at the same power. Okay?
When you talk about IoT, you put this into that, you put an FPGA onto a CPU for IoT, and you can take all of the autonomous driving algorithms that you're using, all of the sensor fusion that you're trying to take, all of the data from the sensors, plug those through the algorithm that then says, "Put the brakes on, steer left, do whatever you need to do." Those algorithms can now be in silicon and perform at a much higher performance rate than they can if the system has to go off silicon onto the board and back. This, we believe, will be an accelerator of this virtual cycle, both at the data center level and the IoT level. That's why we made this investment. Along with the fundamental business, is fed by Moore's Law. There are three fundamental reasons we did the Altera acquisition.
The base business is fed by Moore's Law. We believe in Moore's Law. We believe we have leadership in Moore's Law, and we're going to continue to lead. Second, it feeds IoT. I gave you the autonomous driving as an example. We can go through many other examples. Third is the data center. Once you have all that data out there, you want to be able to do big data applications. FPGAs are great examples or great usages of that. The last section I'm going to talk about is our investment in wireless. Hopefully, it's pretty obvious based on the conversation. If you just take a look at IoT, we believe those devices need to have every form of connectivity and the ability, in some cases, to move through that connectivity spectrum as they travel. When you start talking about robotics and drones and things like that.
These devices have to be connected. They have to stay connected. They need high-speed connections to the cloud in order to download data. We think that's important. Also, there's a transition coming. 5G is coming, and 5G is an opportunity for Intel to work with both the base carriers, the Ericssons, Huaweis, Nokias, to build the base stations and build the infrastructure of 5G. What we're finding is you need to be in the other end of 5G in order to be credible and to have a presence in the standards bodies and in the customers. We think 5G investments and IoT are just two examples. That, along with our ability to get into the mobile business as well. As we said, we're being very careful and cautious about that in order to do it in a profitable way.
We really believe that we are positioning Intel for the future of mobile by having these leadership positions. Okay, I've got 2 minutes and 25 seconds left. I'm right on time. I want to just wrap up our conversation and end with my commitment. I've tried to end each session of these investor meetings each year with a commitment that you should be able to walk away with and hold us accountable to. The four key takeaways and commitments I want you to walk away with is that we truly believe that this strategy of having a base business of the client that's stable, profitable, and innovative is core. That the virtuous cycle of these growth engines of IoT, the data center, and memory, all being connected through our wireless strategy, is truly going to be the growth engine.
You've seen that in the data center case, it's growing now, so it's a higher amount of our profitability and an increasing amount of our revenue as a corporation. We are seeing strength in client. We believe we put the right investments in place and have the right innovation. We've worked with our partners like Microsoft and the OEMs to bring great form factors, Win 10, Skylake. We have Kaby Lake next year. We believe the innovation engine in that space is strong, but we've not required that for our corporate strategy of growth. We've tested it to high single-digit declines and still believe we're a growth company in that model.
That you've seen this DCG, IoT, and memory virtuous cycle, and that it's important that we have good positions in all three, that when we think about them, we think about them as a larger platform, and that in almost every customer visit, it's never just a singular discussion. It ends up being a discussion on all three of those. Just my last commitment is around our leadership team. We believe that we continue to grow. You see that the leadership team of Intel today is quite a bit different. Roughly about 40% of my direct staff is new to the company from outside of Intel. We've really transformed Intel into what I believe is a modern-day, very aggressive, and transformative company, and you have our commitment to keep driving that transformation. As the industry and the world changes, we're going to continue to change as a company.
With that, I think I'm done. I now have 16 seconds. Oh, no, I'm 17. Oh, I'm 16 seconds over. I apologize. I'd like to invite my friend and the person who knows Moore's Law better than anybody other than Gordon himself, Bill Holt, on stage to talk to you.
Good morning. I'm not sure whether I should be pleased how many times people have said what I'm going to talk about today, there certainly were a number of them. Hopefully, I won't disappoint you too much now that you see what I'm really going to say. It's very nice to see Gordon's video again, it is sort of intimidating to follow an icon like that and talk about the same topic. Hopefully, I also won't embarrass him in that sense. Let me get started. Make sure I know what's what up here. I have four things I want to talk about today. I want to give you an update on 14 nanometer, and most importantly for me, how we're doing on continuing to deliver on the cost per transistor trend.
That is the fundamental economic law that Gordon was referring to, that Andy referred to. Without that, there really is no justification for continuing to advance technology. There's lots of other good things that come from it, the amazing improvement in performance. If that didn't come with a cost reduction, we wouldn't have been able to afford to do this. I want to show you that. Then I want to take a little different look at the economics of Moore's Law and take a little bit of thinking about what does it take to be able to afford to continue doing this? Clearly, there are less people doing this than there used to be, and that obviously relates to the affordability of continuing to do this. I want to look at that for a moment.
I will take some time on competitiveness, and then very briefly, I want to just give you some indication of some of the things that are coming down the pipeline, things that give us confidence that, in fact, there is a future in Moore's Law in the economic aspect of it and what we're going to be able to deliver over time. Okay. Last year, I showed you this chart on how the yield was progressing, and if you look carefully at the chart, we were predicting that early this year, we would have achieved equivalent yield with the 22-nanometer technology. We also noted that our 22-nanometer technology is probably the best ever we've had for yield. If I update that chart for today and stretch it out for the rest of this year, what you'll see is we didn't quite achieve those goals.
We have been showing steady progress, in fact, in improving the yield, but we have not yet closed the gap to 22-nanometer. That gap results in a number of things. It results in a tightness in supply, though we have been meeting all of our commitments in supply. Some of the recent reports about shortages are because people are actually ordering more products than what we had predicted, that obviously is a good thing, but it is a shortage. It also means that the products that we were producing are somewhat more expensive, Stacy will give you some data on that.
The good news is we have been seeing improvement, we have good data to suggest that we will in fact close this gap, that we do not see a sustained delta occurring that would in fact remove some of the value from continuing to do scaling. We see that while it's been harder than we thought, it's taken us longer than we would have liked, we do not see that long-term, there will be a difference between what we were able to achieve in the past and what we expect to achieve going forward. This is another chart I showed you last year. It's probably the most important chart as far as I'm concerned. The cost per transistor is the fundamental metric that we need to drive down and to continue to make each new technology, one, affordable, and also valuable to pursue.
This year, I want to add onto that chart some estimates of where we think our 7-nanometer technology will be. Last year, I hinted to this, I wouldn't actually put anything down, I'm not going to put an exact point down because there is still some degree of uncertainty in these various parameters. On the left-hand side, you see the increasing trend of wafer cost, this is cost per millimeter squared, but the same shape if it was wafer cost. This is coming due to some of the items that Brian mentioned, the technology is getting more complicated. We have not seen a fundamental improvement in our patterning capability, that has led to an explosion in complexity. The net result is that the cost per wafer is going up faster than it had traditionally.
As I said last year, our response to that is to improve the scaling faster than traditional. When Gordon mentioned that basically we're seeing a 2x increase in the number of transistors per generation, that's fundamentally this curve. What we've done is we've actually focused on that to a higher degree and used increased scaling to be able to offset that increased wafer cost. The net result then is the curve of the combination of these two, which is the cost per transistor. That's the fundamental building block of all of our devices. That's what drives the cost reduction on products. In fact, we saw very good results on 14-nanometer, we're projecting even better results on 10-nanometer. We can see enough of 7-nanometer now that we expect that it will in fact continue to stay below the curve.
Exactly where below it is some of the things we'll work out over the next 2 to 3 years. We have defined enough of the technology that we do know that we expect with high confidence to be with inside that oval on the right-hand side. Let me move on and talk about Moore's Law. Many of you have seen us talk about this. You've seen Andy talk about this, as Brian mentioned. I'm going to talk about it in a slightly different way today. Fundamentally, Moore's Law enables us to take the same functionality and make it in a smaller and smaller die. As I showed on the last chart, a cheaper and cheaper product. You can use that benefit to continue to use a product that's basically the same cost, but providing increasing a lot of functionality.
Those are basically the two choices. Over the last few years, we've had a model, Brian referred to it, that we've used it in the past to show you why it's better to go faster. We've built this model, and we have a fair amount of confidence in it. It allows us to take a look at what is it that we have to spend over a 10-year horizon for our manufacturing and development. We can then perturbate that model with varying parameters.
When we last did this in 2011, we showed you that if we were to look at what it's going to cost us in this model over the next 10 years, we'd spend about $104 billion in the combination of red, which is the developing of the new process, and the blue, which is the manufacturing cost of the factories and actually building the products. What we show you is that there's a substantial benefit to continuing to do these new technologies as opposed to remaining on the old technology and attempting to build the same products. This is a theoretical model. You obviously can't build the same products on a 10-year-old technology. You can calculate what the cost would be, but you wouldn't have a product that you could actually make unless you just made the same product.
If you continue to increase transistors, make them faster, if you were somehow able to do that, they would not fit in the package. They would not be able to fit in the power envelope. From an economic point of view, you can make this comparison, and you can say the benefit of, in fact, advancing technology for Intel in 2011 estimate was over $200 billion. Okay. This is the model that we're using. We've updated this because trends do change. In 2015, we went back and reran the model. What you see are things that we've talked about. The cost of development has gone up. The red bar is, in fact, larger. That's because of the same thing that's driving up the cost of wafers.
The complexity is exploding on us, the number of tools it takes, the number of wafers it takes, the amount of people it takes to develop the technology are, in fact, increasing. The amount of money you have to spend on that new technology development is, in fact, larger. You also see here a slight change in the case for not doing any new products. That comes from a slightly different projection on what the unit growth rate would be. You basically see the same fundamental benefit that comes from advancing technology. What it did tell us is that you can perturbate this model, and there are things that can actually make visible changes. We set out to look at what are the most important ways that you're going to, in fact, test this model.
The first one is, in fact, lower unit demand. As I said at the beginning, there are less people doing new technology. Why is that? Well, there are less people who can afford it. What is the fundamental driver of whether you can afford it or not is whether you have enough revenue to, in fact, justify the investment. If you perturbate unit demand, it's very likely that you can probably affect the robustness of this development model. The second one, as I already mentioned, higher development costs. If that red bar got sufficiently large, then it would swamp out the benefit that you would get on the manufacturing side. The third one is cost per transistor.
If we're not capable of delivering a sufficient advantage or a sufficient improvement in the cost per transistor, then in fact, the blue part of the new technology model will, in fact, grow, and it will again become approaching the size of the no technology improvement case. Let's go look at these. If we looked at the original 2011 model, we were projecting a given growth in demand. We began to change the parameters for the model, and we looked at what if it was flat. I'm not saying it's going to be flat. Obviously, Brian has showed you a lot of reasons for what the real projections will be, we looked at this, and there was no problem in terms of those two bars still being very different.
In fact, when we took the model to the limit to find out when would, in fact, the cost of developing a new technology and deploying it equal the cost of staying on the same technology in our theoretical model, you have to, in fact, have a decrease in unit demand of 25% per year. Now, the background of this graph is wafer starts, and you can see that this is somewhat of an absurd case because the number of wafers that we would be building out there in time is approaching zero on this scale. The conclusion from this is that there's a very large margin for us in terms of our ability to afford to actually generate enough volume to continue to benefit from advancing Moore's Law. The second one I also mentioned was increasing R&D cost.
In our original version in 2011, we looked at the historical rate, and the cost of new technology development was increasing at about 10% per generation. This was a metric that we actually used to try to control our spending, and we tried to keep within this window. In fact, that isn't the case, and if we look at what our projection was in 2015, and we look at the more recent history, it's closer to 30%. That's what I said. That's what made that red bar get bigger. On this scale, if you wanted to make the red bar basically eat up all of the benefit, you'd have to go to a rate of 190% per generation. I think the more significant thing is that if our revenue was projected by the model, you'd have to be spending $80 billion a year on technology development.
I think long before that, Brian would have found other people to go work on this problem. I think it's fair to say that the increasing cost of R&D, at least for Intel, there is an increasing cost in R&D, but it's not going to be the thing that stops us from pursuing Moore's law. Let's look at the last case, cost per transistor improvements. This is a stylized version of that earlier chart that I showed you, and you see the decreasing trend. Historically, this trend has been about 0.7x cost reduction per transistor per generation when you combine the increasing wafer cost and the die scaling. On this graph, I've shown some of the recent numbers just to give you a context that we think we will achieve on 14 and 10.
Again, we adjusted the model until there was no benefit, so the two bars were equal. On this scale, we'd have to be at a 0.86 cost per transistor improvement trend as opposed to the 0.7. Now, this one is the only one that actually is somewhat close, and this is still a fair distance away. This is a substantial amount of reduction. The good news is that this is the one that we most directly control. This is what we're paid to do. This is the one that doesn't depend on Intel's volume. It doesn't depend on how much it costs us to do this. It just depends on whether we have enough good ideas and enough motivation to, in fact, focus on and drive this trend.
While this could challenge the economic benefits, we don't see any indication in the foreseeable future that this will in fact be a problem. I think our conclusion from this is that while economics likely will be the biggest limiter, and it is the limiter that is, in fact, reducing the number of people actively pursuing Moore's law, based on our size and the benefit we have of our scale, none of these factors will be the ones that cause us to have any problems. Okay, so let me shift gears and look at competitiveness. I'm going to basically focus on scaling or density because that is the important part of cost per transistor that we actually, again, can control. Last year I showed you this chart. It was based on two very simple numbers.
We used this chart last year because data was increasingly becoming available both on what people were doing and what we were doing, on projections going forward. It's not a very sophisticated metric, it does generate the shape of the curve and give you a pretty good understanding of what you would expect for the entire technology. The year before, we took a slightly different tact, we made an attempt to try to estimate what would we think the full chip scaling would be for a roughly equivalent kind of product as we progress through the technologies. What you see is the shapes are very much the same. After the investor meeting, we actually got more information on some of the competitors, we adjusted the line slightly.
In fact, the shape is very similar to the previous one, it basically gives the same message. It's hard to get data for this graph because this requires you to know almost entirely what the technology is going to be or for someone to tell you what do they predict their scaling to be. What's happened recently now is we actually can look at products from Intel and from our most important competitors at the 14-nanometer generation. The easiest one to look at is to go look at the Apple products because we have a number of examples there of products that are quite similar. On here we're showing a normalized density for the Apple A8 at TSMC, the A9 at TSMC, then that same basic product at Samsung. Those are all available.
We've actually bought phones, taken them apart, we, like everybody else, have been looking at the die to see what it actually does. It gives us a good indication. Now, beyond just being able to look at two pitches, we can now look at many aspects of what the technology provides. This base number is basically simple to calculate. You just have a die size, you can count roughly how many transistors are there. If you make that same measurement on an Intel part, on our primary parts that we've been selling, our high-volume products, what you'll see is a number that actually looks exactly the opposite of what I just showed you. What it shows is that the raw density as measured by just area and transistor count actually looks worse.
Since I'm sure that you've seen various reports of these kinds of numbers, you should be asking yourself, why did you show us that previous chart? What is going on here? You said that you would be better, yet when you look at the raw data, it's not. It's worse. Well, we've said in the past that that name that we generate with technologies is not always the same, the same goes for the composition of die. I'm going to take a little diversion and talk about what are the various kinds of blocks that you typically will see in a product. What you find is that depending upon what the composition of the product is, you can make dramatic changes in what the transistor density or the transistors per millimeter are.
Remember, for an entire technology generation, we're talking about an improvement of maybe 2x. If you change from most of the die being memory to most of the die being logic, you can make a 3x or even 5x, depending on what cell type you choose. What it tells you is it's very important if you're going to do a legitimate comparison to understand, are these products actually similar? Are they aimed at the same market, and are they constructed from a similar kind of basic design blocks? If I go back and look at those same five products, what you will see is they are not really that much alike. Of course, the Apple parts are very similar, and the Intel parts are very similar, but they're addressing fundamentally different markets.
If you look at the details here, the one that is most dense, the red, which is the amount of on-chip cache memory, is almost three times higher, which adds to an almost 6x increase in those transistor counts as opposed to the area that's allocated towards doing logic. If we take this data, we now know the composition of the parts. We now know the relative density that we can get by actually looking at the layouts. We go back to that previous chart and adjust it for this. What you see is that, in fact, that improvement is present if you were to build a similar part. These are no longer the sub parts that they were. They are parts that have the same distribution of circuit types. I'm sorry if this is complicated, it's actually quite important.
If we now take this data and apply it back to that original graph, we can now actually update the points of 14 nanometers. Here's the graph that I showed you, if we now take the information from analyzing the actual released products, it has a slight movement, what you see is a number of factors that now we can validate with actual products. In fact, most people's technologies, as they moved to the 16/14 node, did not really provide any increase in density. That's evident in the raw data as well as in this chart. Also, you can see, just to validate the legitimacy here, is that the Samsung part is slightly smaller, has slightly higher density than the TSMC version of the same product.
When we take and normalize the parts or the technologies for similar composition, in fact, the data point for Intel 14 nanometer does show that we have a substantial lead. Recently, people have begun giving more projections on 10, we're able to update that data as well. Our projections for what other people would do and their more recent statements are very much aligned, they show that this increased density advantage will continue into the future. Let me move on then and look a little bit out towards the future. This is a busy chart. It's intended to be busy. The point of this chart is that there are lots of very interesting, exciting, and good ideas that are being researched, both at Intel and universities, at our fellow travelers.
The size of the chart or the number of things on the chart is what is important. This is not to imply that all of these will actually prove to be beneficial, the fact that in spite of the concerns about the difficulty in advancing Moore's Law, there are a lot of really good ideas. There's no question that you're going to have to make substantial innovation to continue. We've all agreed that the days of very simplistic scaling are long over, and the only way you can continue to drive economic benefit and all of the rest of the things that come with Moore's Law is, in fact, through significant innovation.
This shows that there are efforts of many different directions, either in increasing the function that you would get from a single device, simple scaling or scaling by using new materials to continue to reduce the size of things, or looking at innovative ways of doing integration. Probably the best for that idea that actually exists today is, in fact, 3D NAND. Stacking die in terms of die stacking doesn't advance Moore's Law. It certainly helps form factor, but it doesn't reduce cost. 3D NAND, where they've in fact found innovative ways to make many, many bits with a similar number of processing steps by stacking the bits up instead of across, does in fact provide the kind of cost benefit, and it basically produces a cheaper cost per transistor for that particular kind of memory.
Applying those kinds of ideas to other areas and looking at ways to in fact stack logic devices. At some of the things that will help with our basic equipment set in terms of more selective depositions or even eventually the kind of holy grail of directed self-assembly, which would help eliminate the needs. All of these are being studied. Which ones will actually come true? I wouldn't tell you if I knew, but I don't really know. The fact that the page is full and these ideas are all credible and they're worth pursuing tells us that the future is actually quite robust. If I summarize, on 14 nanometer, the yields are improving. We do not see a fundamental problem in the ability that scaled technologies will no longer be able to achieve the same yield levels.
We have a very high bar in our 22 nanometer technology, which is and continues to be the highest-yielding technology we've ever had. As you saw from the chart, while we are a little behind where we wanted to be, we are in fact closing that gap. We don't see any reason that we will not be able to achieve those same levels. Certainly, we do have some demand increases. Availability is not everything we might like. We have been consistently able to meet our commitments, and the shortages that do exist are generally due to the fact that people like these products. Brian said Skylake is probably our best introduction of a product ever. It is a very good product. It has very high demand.
If I look at the cost per transistor, it's clearly getting difficult, and it takes a lot of innovation and a lot of work. The progress really is promising, and the projection out in time also looks very good. If I look at the economics of Moore's Law for Intel, this is very solid. This is very expensive, it's very difficult. It takes commitment from the company to continue to invest. We have the ability to do that, and we project that, in fact, that investment will deliver a substantial benefit. When I look at the competition, basically, it's exactly what we thought three years ago. It's kind of interesting that it is, in fact, exactly what we thought. Everybody has done exactly what they said they would do.
That's an indication of everybody being pretty much on track, but it doesn't change the fact that we project and we have good evidence that we will have a substantial benefit in density that can be translated into cost or functionality. Finally, when I look out into the more distant future, and we've never said that we have much visibility beyond 10 years. 10 years is even getting kind of fuzzy these days. When you look at the kinds of things that you're going to need out in that time period, there's a lot of progress being made, and while we don't know exactly which ones are going to come, we do have a high degree of confidence that the right kinds of technologies will be available to continue to increase the fundamental benefit of Moore's Law. Actually, I'm done a little bit early. That's very unusual.
I limited my foils so that I wouldn't run over. I guess that means you get to go have your break four minutes and five seconds early. Thank you.
Ladies and gentlemen, we will now take a short break. Please be back in your seats at 10 o'clock. Once again, 10 o'clock.
Thanks, everyone. Thanks, Bill. All right. We'll take a quick 30-minute break now. I'd like to invite you to join us out in the lobby, and we can mingle there. I'll again remind you that we will be filing an 8-K, and we'll be issuing a press release momentarily, and the outlook that's in that press release and 8-K will appear up on the screen here behind me. To those of you out joining us on the webcast, we'll be back in about 30 minutes. Thanks, everyone.
If we're not going faster than the epidemic, the epidemic's going to win. We've got to be able to go faster than time. Death from Ebola is one of the most horrible and painful experiences that any human can endure. When the Ebola epidemic exploded in Africa in 2014, the doctors couldn't communicate quickly enough. They couldn't bring laptops or tablets into the tents because if they carried them back out again, they'd be bringing the disease out with them. All these guys could do was just write stuff on paper and then run to the edge of the quarantine tent and yell it out to somebody. A bunch of us got together and said, "How can we solve this? What can we do?" What we needed was a portable network, which was a series of tablets connected to a server wirelessly.
The only solution we found was the Intel Edison. Everything hinged on that. Each had to be covered by a protective case so it could be dunked in chlorine to kill the virus. Working in a crisis, it changes the way you work. It changes the way you think about problems. We made a system that would've taken a team two years. We did it in two months. Once we had a system like this in place, the doctors could spend more time with patients. There's nothing more personal than helping somebody live, than keeping a family together. Without a lot of money, with simply dedication and some time, you can make a difference, and that's really what making is all about.
Did you know Intel processors brought personal computers to the home?
Totally.
Intel made them more efficient so that you could fit all this into a laptop.
Tight.
Real tight. They helped bring Wi-Fi to everybody.
Whatever that is. Come on.
Intel's best processor ever is here, which helps PCs do things you never thought possible.
Like time travel?
Hey, these pants are timeless.
Introducing Intel's new 6th Generation Core processor. It's our best processor ever.
When I was 12 in 2013, a flyer came to our doorstep asking for donations for the blind. For a moment, I closed my eyes and I wondered, what would I do if I was blind? I thought about all the things that I would miss, like my friends' faces, a football game, reading a book. I asked my parents, essentially, how do blind people read? They were quite busy at the time, they said go Google it, and that's where I found out all about braille printers. The normal cost of a braille printer is $2,000. It just got me thinking there had to be a better option. That's what inspired me to try and build a new one. I started with a Lego robotic kit. Yeah, I just went for it. It was challenging because I had to break down models, build models.
It took some time. When I finished the Lego version, I made new designs, but I also needed a better brain. That doesn't even cost too much, which was super important. Just made everything work together perfectly. Then I saw the first words come out. That's what really motivated me. I felt like I could do something and help people's lives. I want to make more of these braille printers, we start a company. Within a year, we can start making hundreds and thousands of these, that's where I think we can start changing people's lives. Inventing is cool because you don't really need a factory. You don't have to be a certain age. You can be whatever you want to be. As long as you have the right tools to do it, the right people to seek out, there's nothing you can't do.
My name is James.
No. Wait, no. I'm not ready. I'm not ready. I'm not ready.
Intel's best processor is here, you can take this very real, very terrifying memory and edit it, share it, play it back in amazing 4K quality. That is terrifying.
Introducing Intel's new 6th-generation Core processor with amazing 4K quality. It's our best processor ever.
Ladies and gentlemen, please welcome to the stage, Stacy Smith.
Good morning, everybody. I want to thank you all for coming and add my welcome to you. I'm really pleased to have you here today. The management team really does look forward to the opportunity to sit down with you, talk about our business, and in particular, talk about longer term trends that we don't really get the chance to do during the quarterly earnings. I do need to start with some bad news. I know that you all greatly anticipate the annual haiku. There will be no guidance haiku this year. It's starting to feel a little bit dated, and you can see Brian and I aren't wearing ties today, so we're trying to be more new tech than old tech. I think the haiku dates back to the 1600s, so we're not going to do the haiku.
We're going to go right into the key messages here. Now, I will say though, as you can read the key messages while I talk, as we worked through the presentations, we did realize that we had a few really critical messages that we wanted to get across. As we were debating the lack of a haiku, we were thinking, is there something we could do to get that across? Maybe instead of a haiku, we thought hard about a guidance limerick. It's hard to rhyme with Intel. Most of the limericks you look up when you go to the web are not appropriate for work, so we didn't want to go that way. We kept thinking about it. We wanted something that could both educate, make our key messages, and inspire.
We thought, what about those motivational posters that employees all the world over love, that get them excited about teamwork and collaboration and things like that. What I thought I'd do is I thought I'd try to do my presentation in 5 motivational posters. This will tell you every message that I'm going to give you then for the 35 minutes after that. Feel free. You've seen the guidance. The stock reacted nicely, so feel free to just take off. You don't need to listen to me anymore because I'm not sure I'm going to help as I go deeper. I'm going to try to do this just with a few quick motivational posters. One critical theme that you heard from Andy is how we're prioritizing our investments.
We're trying to invest in areas that we think will generate long-term shareholder return, and that we continue to execute on our priorities around how we allocate cash to our shareholders. See, I see the photographer. PR has told me for this next one, I need to move out of the way so my picture is not shown with this one behind me. I'm going to move all the way over here since the photographers are there. Our second key message is, again, you heard from both Brian and Andy. We did not hit our expectations last year in terms of PC TAM growth. I'll talk a bit about what happened there. I think it's interesting, though, when you look underneath the surface, that even with that weakness in the PC TAM, we were able to keep the company approximately flat.
We'll be down 1% this year if we hit the midpoint of our guidance. We're seeing growth in the data center, growth in the Internet of Things business, growth in our memory business, that's offsetting that TAM weakness. That becomes really important as we think about the future of the company, and I will show you some napkin math at the end to illustrate that point. I also want to say, and you'll hear this from Kirk, we think we're extraordinarily well-positioned in our client business in terms of the kinds of products coming to the market, the grouping of features that we're doing, and the opportunity in the future. I'll now move back and where you can get my picture with this. The third critical message is really around how we're transforming the business. Brian talked about that.
I'll talk a bit more, but the transformation of the company is well underway. What becomes interesting is, as the other businesses grow, this becomes easier and easier. It becomes easier for us to generate growth with the portfolio that we have, independent of what happens in the PC TAM. The fourth critical message is about loneliness. It's tying to what you heard from Bill. Our competitive advantage is becoming increasingly rare. Pretty significantly, and almost all of that is being offset by strength and growth in the data center business. In particular, that business has become quite large, but also contribution from Internet of Things in the memory business, and actually, the software business is contributing some this year as well, and I'll talk about that. We are down, back to the Stormtrooper regrets foil.
We're down some from what we thought a year ago at this time. There's really two major reasons for that. First, the impact of the XP end of life last year resulted in a bigger pull-in of demand than we had anticipated. Secondly, is what we've seen in terms of macroeconomic growth rates. We talked about that over the course of the year. This is showing third-party macroeconomic forecast for 2015 a year ago versus what they are today. What you can see there is significantly weaker growth rates than what we thought, particularly in emerging markets and in particular in China, which is the world's largest PC market. Several parts of our business correlate to GDP. While we saw strength in lots of places, every one of the places where you see a strong correlation to GDP, we saw some level of weakness.
The business portion of the PC market, the enterprise portion of the data center, portions of the IoT business that correlate to GDP, more the industrial segments of the IoT market. I think this was a big driver of our results last year, not the only driver, but a big part of it. The first business I want to talk about today is the data center. For this business, what I'm going to talk about is how we're growing over time. I'm going to try to give you some insight into what we're seeing in terms of growth rates for the various segments of the business. Spoiler alert, the cloud segment is growing very fast. If you hadn't picked that up so far, you weren't paying attention.
I'm going to hit it at a pretty high level. Then Diane's going to come in and give you a lot of insight. I'm actually quite looking forward to her presentation. There's some great data in there. This chart, and this will be the format I use for both data center and for the client group. Just so you know what I'm showing you here, it shows three years of full-year revenue, so '12, '13, '14 on the left-hand side, and on the right, I'm showing year-to-date results for 2015. What you can see here is that this business is growing very fast. Last year, so in 2014, the revenue in this business was over $14 billion. That was 18% growth rate. Operating margin percent was 51%.
You can see that we're on track this year for another year of double-digit growth with a revenue projection of growth that's in the low double digits relative to 2014. Our year-to-date operating margin is 49%. Again, some pretty obvious observations here. This business has gotten to be quite large. It's growing fast, and it's very profitable. When you dissect the data center business, you see some interesting trends. The enterprise portion of the data center, which is today our largest business, has been relatively flat over the past three years. In fact, in 2015, it's actually down a bit from 2014. If you move over to the far right-hand side of the chart, you see our cloud business, because it's cloud CPU revenue. This is CPU revenue by these different businesses.
You can see that this portion of the business grew by over 40% last year and is on track to grow by over 40% again this year. You also see how large, because of that growth rate, how large this is now becoming relative to the rest of the businesses. There's some fundamental drivers that are driving this growth rate. This build-out of the worldwide cloud infrastructure. More and more devices are connecting. Increasingly, that's also the Internet of Things that's connecting to this cloud infrastructure. We're seeing richer content that's being stored and served out to the devices and consumed. We're seeing these viral-oriented consumer applications and services that are being deployed. That's all driving this build-out of the worldwide cloud infrastructure. Again, Diane will talk a lot more about that. In the middle of the chart, you see a collection of other businesses.
There's no rhyme or reason to this. I just wanted to separate out the enterprise server from the cloud server. In the middle, you see high-performance computing, workstation, networking, and storage. In total, when you add those businesses up, they're seeing a nice growth rate. It's 10%. What's interesting, though, is when you click down to some of the smaller segments that I think are going to emerge as being very important, you see very robust growth rates. For example, the networking business that's in that middle bar is growing at rates consistent with what we're seeing in the cloud. Some very fast growth rates in some of those, and I think they will emerge out in time. You heard about the opportunity we have in terms of low share, fast growth rates, great products.
I think those will emerge to be big opportunities for us as we go forward. The prior chart showed the relative growth rates of these different businesses. This chart shows the mix. This is going back to the beginning of the timeline on the prior chart, 2013. What you see here is the mix of our server CPU revenue. Enterprise, by far, is the largest portion. It's almost half of the business. I'd say we were very dependent on enterprise back then. Let's fast-forward then to our expectations for 2016. What you can see here is that that robust growth rate that we've been seeing in the cloud now makes that portion as big as enterprise. The other portion is a third.
It's a nice heuristic to think about enterprise server a third, cloud server a third, and the rest of the business, about a third of the business. What this means is that going forward, the growth rate in DCG is less reliant on the enterprise, which is good for us because it's a place that hasn't been growing over the last few years. Frankly, we have some plans to help that segment grow, but it's going to be a much slower growth rate than what we see in the rest of the business, and Diane will talk about that. Just one other comment. You can actually see that playing out this year. If you think about the data center for 2015, going back, we're seeing the enterprise segment actually down, and yet we're still able to grow in the low double digits.
You can kind of see how that's played out. If you go back three years, we would not have been able to generate that level of growth in a year where the enterprise was down. Moving forward to what we expect for 2016. We expect that we're going to see revenue growth in the mid-teens, operating profit growth in the low double digits. Remember, we talk about these long-term growth rates for the Data Center. It's not an annual steady march. It's going to be pluses and minuses. Think about some range around that 15% that we've been talking about. Our view right now for 2016 is that we're right in that range in terms of the growth rate.
Like last year, if you click down a level, we're expecting a relatively weak enterprise segment and continued robust growth rates in cloud and some of the other segments, like networking. You'll notice here operating profit is growing a little bit more slowly than revenue is. I'll come back to this when I talk about unit costs. The short answer is that we'll be ramping the 14-nanometer products for the Data Center Group in 2016, and while the cost curve for 14 nanometer is coming down nicely, Bill talked about the extraordinary costs we're seeing on 22 nanometers, so they're still higher than 22 nanometer, so we'll see a little bit of a cost increase in the Data Center business. I'm now moving on to the client segment.
As we've said, financially, this business is a little bit behind the expectations that we had for this year as a result of the PC TAM being weaker than we thought a year ago. Again, I think there's two key reasons for that. First is because of what we're seeing in terms of worldwide GDP. Second is because of what we're seeing in the overall XP refresh a year ago. Two thoughts, though, as we enter 2016. First, this business generates huge scale for us and a lot of cash flow. Second, we're very well positioned in this business for growth going forward, and I think we're showing increasingly an ability to get paid for our leadership products here, and I'll dive in on that a little bit because I think that becomes really important.
This is just showing how our view of the PC TAM changed over the course of the year. What you see on the left-hand side of this chart is what IDC and Gartner were showing for the PC TAM. You'll see a very long footnote on this slide. Remember, we have to make some adjustments to their forecast based on how they handle detachables and things like that. It gives you a relative sense of how the view changed. They were forecasting kind of slightly down to up a year ago. We were kind of in the middle of those forecasts. If you go to the right-hand side of this slide, you see where they are as of today. They're showing high single-digit declines in the PC segment. Again, we've been there for three or four months in terms of what we've been saying.
That's consistent with our view. That weakness is two things. One is we underestimated the impact of the PC refresh on last year's demand, second is the lower macroeconomic growth. This business is on track this year to be well over $32 billion in revenue. It's very profitable. It gives us enormous benefits of scale, I'll talk a bit more about that. Year-to-date operating margin for this business is $5.4 billion, you can see in the year-to-date compares that that is down pretty significantly year-on-year. This is due to the increase in unit costs as we've ramped 14 nanometers. While the Data Center Group will be doing the transition in 2016, the Client Computing Group did the transition for the most part in 2015, saw an increased unit cost associated with that.
That unit cost increase has been partially offset by the improvements that we've made in phone and tablet product profitability. One of the key strategies in this business has been to create brands and features that allow us to segment our product lines. I'd say this strategy is working nicely, even better than we thought at the beginning of the year, we're seeing an all-time high product mix. Just to give an example of this, would be the gaming PC. We have created an extreme edition version of our Core product line that enables a fantastic gaming experience. We've created SKUs and brands that allow people to identify that that's what they're buying and to actually desire it and walk into a store to buy it.
That's allowing us to get paid for the performance and the experience that gamers are happy to pay for and where we can add in this technology and get paid for our technology leadership. You see other examples throughout our product line. Turbo is a great example. vPro is a great example. Again, both of these are delivering strong value to the customers and places where we deliver that value, we can get paid for the technology leadership we have. Kirk's going to talk a lot more about the kind of future of features and segmentation in his talk. You can see the impact that this has had on the graph to the left, this translates directly into our financials.
For those of you that have followed us for a long time and you think about our business, you know that we have decades of history where the pricing in the PC market comes down year by year by year. We had a rule of thumb that was remarkably accurate that the average selling prices came down 4%-5% a year. You can see here how in kind of 2009 and 2010, that trend was disrupted. Since that point, you can see the shift in the curve, you can see that actually pricing over the last five years has been flat, maybe even up a little bit, depending on where you start the trend. This has had a huge impact on our profitability. This is a really important shift, I think you saw it as recently as last quarter results.
Really importantly here, this isn't that we're raising pricing. You're going to hear this from Diane, you're going to hear it from Kirk. We bring in more performance at the same price point, sometimes even lower price points. What we're seeing is a shift in mix, and that comes back to that ability to create brands and segmentation and bundles of features. We can create situations where people buy richer mix because we're giving them something that they want. That's a really important point here. All right. As you heard from Brian, what did he say, that I get to deliver all the good news? I remind him I was the guy that had to stand up and tell you about the contra-revenue program two years ago, which you all loved so much.
As you heard, we're on a path to exceed our mobile profitability goal that we set a year ago. My best estimate now is we'll be closer to $1 billion for the year. About a third of that improvement is coming from product margin improvements. It's both the fact that we're seeing less and less contra revenue that's on track to what we've been expecting. We're also seeing a less expensive product line that we're selling into the value segment of the tablet market, which would be SoFIA. A big chunk of it is coming from lowered investments. I'd say the one place where we didn't hit the goals that we had for the year here, although we made it up in other places, was we expected an LTE ramp. We didn't see an LTE ramp.
We saw some, but not nearly to the extent that we thought this year. As we go into 2016, we're forecasting another $800 million improvement for these products. The majority of it will be coming from product margin. I'll show you some investment shifts over time, but we think we're investing pretty much at the right level now for these products. The majority of that's going to come from product margin improvement. The way you're going to see that going forward is in the impact that it has on the overall Client Computing Group operating margin. I'll make that tie when I get to the final page for CCG. I do want to come back and recap a key component of our competitive advantage that we talked in depth about last year.
We talked about the scale that's required to invest in a broad range of critical IP blocks and the scale that's required to advance Moore's Law faster than the rest of the industry. We also talked about the fact that we think this is a competitive advantage for us and one that's becoming increasingly rare and more valuable as the industry consolidates around us. I think you've all seen 2015, that consolidation actually accelerated pretty dramatically. I think we're now virtually unique in that we have the ability to build our own factories and the ability to invest across this really broad range of leadership IP blocks that you've heard about today, and you heard a lot about last year. It's the client business that gives us that scale, and that's a really important point.
Andy talked in his preliminary remarks how we invest in IP for the client business. It gets used in other portions of the business. Increasingly, we're investing in low-power IP in other portions of the business that comes back and makes the client business better. That's a real virtuous cycle for us, but it's the client business that gives us the scale to do these things. The client business generates huge cash flows. It's the vast majority of our units, and it funds more than half of the shared IP across the company. Said a different way and even more pointedly, if it wasn't for the client business, our profitability in businesses like the Data Center and the Internet of Things business would be much, much less because of the cost they would have to bear to go and service the markets that they target.
Coming back to CCG 2016 expectations. From a revenue standpoint, we expect that revenue for this segment will be flat to low single-digit growth. Remember, 2016 for us is a 53-week year. I know that throws all your models into chaos. It is a confusing point, but it is the reality. Otherwise, we'd be closing the books in Thanksgiving, and we don't want to be doing that. We're expecting the overall PC market to be down slightly. Here again, I'd say we're probably a little bit more on the cautious edge versus the third parties. I think they're a little higher than us, although they'll be revising their forecasts also as we go in. We're expecting the PC market to be down some. We do expect improvements in our mobile product ASPs as we have that full-year effect of the contra-revenue reduction.
We'll continue to see some uplift in terms of overall ASPs at a platform level, but that's coming from tablets and phones for the most part. We expect significant improvement to our operating margins as a result of improving 14-nanometer costs and improving profitability on our phone and tablet products. I'll talk a bit more about that cost in a few slides. Now I'm going to move on and talk about three businesses together, which is the Internet of Things business, the memory business, and the software business. This just stacks the revenue for those three businesses. It gives you a three-year trend. What you see here is that each of these businesses now is well above $2 billion. In combination, it's a collection of businesses worth more than $7 billion in terms of the revenue contribution to the company.
They're on their way to $8 billion+. When we close Altera, and we add this to that collection of businesses, you'll see a collection of businesses that's starting to approach $10 billion in overall revenue contribution for the company. When you think about those, again, in aggregate, they become a relatively significant portion of our potential growth and our competitive advantage. I want to take a second and just talk about each of the businesses separately because it's not one business. I'll start on the far left-hand side of the slide, which is the Internet of Things business. You can see the Internet of Things business has been on this nice three-year growth trend.
Here for these businesses, we are including a 2015 forecast, but we cleverly didn't give you the guidelines, so you'll have to eyeball it or get out your laser pointers and figure it out. It is a business where we're increasing our investment. You can see kind of the flattening out of the operating profit trend. We're making some significant investments in this business because we believe there's some significant long-term opportunity for us to create value for our shareholders. The areas where we're focused in this business would be things like retail, transportation, industrial. You heard a little bit about that earlier, and Doug will be at lunch, and we'll share more if you like.
We expect that for 2016, the revenue is going to grow faster than it did in 2015, primarily as a result of macroeconomic improvements as well as the design wins that we've been working on over the course of this year. Moving to the middle of the chart, which is our software business. In 2015, we made several divestitures in this business, and we really refocused the business. I think Chris Young and the team have done a very good job on that. We're now cleanly focused on the threat defense life cycle with an emphasis of endpoint security at the core of that. On a GAAP basis, we're projecting a little bit of a revenue decline this year. It's in the range of 2% or 3%. Operating profit growth from a relatively low base, but up pretty significantly.
You can actually start to see the operating profit popping out on that. That's more than 100% operating profit growth on a GAAP basis. What's interesting here is that this is the only business inside of Intel that prices in local currency. If you look at this business on a constant currency comparative basis, what you see is we're seeing a revenue growth rate that's kind of 5% on a constant currency basis, you can start to see the impact of the focus that the leadership team has done, and a operating profit growth rate that would be much closer to 200% year-on-year, again, on a constant currency basis.
As we think about this business going forward, 2016, we think it's going to be another year of revenue growth, and another year of robust operating profit growth based on pruning the portfolio down to some places where we think we have a stronger competitive position, and the lowered investment levels that's come along with some of the divestitures. To the right is our memory business. You can see it's growing fast, and it's quite profitable. As you've heard, we've announced that we're making a significant investment in 3D memory technology called 3D XPoint. As we go into 2016, we expect continued robust revenue growth, but we also expect that our profitability will be down in 2016 as a result of two factors. One is just price competition in the legacy NAND business.
Secondly, the investments that we're making, the start-up costs associated with starting this up in China will be pretty significant. We'll continue to see revenue growth, we'll have some price competition in the legacy business, then we have the impact of the investments we're making in the 3D XPoint technology and the start-up of the memory factory in China. All that said, I talked to many of you on the break, this is a place where we're really excited about where we're positioned. If you think about it, we have this leadership over the rest of the industry with a really disruptive memory technology, and we strongly believe this is a place where we can create some long-term shareholder value.
We'll be able to grow the business that you see on this page, but as importantly, maybe even more importantly, we have the opportunity to connect this to our Data Center platform and connect it to our Client platform and deliver an experience that's a much enhanced experience when it connects to this kind of memory. This thing has benefits to the rest of Intel and a strong benefit just to the business itself. Moving on to the section where I talk about spending and CapEx and margins. I want to give you some insight into how we're shifting our investment portfolio. You can really see in this chart, this is a two-year view. It kind of shows the shift in investments starting at 2014 and going through to where we think we're going to end 2016.
You can see in this chart what I call how we're operationalizing our strategy. Company's strategy gets operationalized by where we invest, where we don't invest, and you can really see that here. We've made significant increases to our investment in the Data Center, in the Internet of Things businesses we've been talking about, technology development, which Bill walked you through, our multi-comms strategy on the way towards 5G, and then memory. You can also see on this chart the places where we're investing less, and you can see that we're investing less in the Client segments of the business, less in PC, and significantly less in phones and tablets. Then you can also see the impact that I talked about of the divestitures that we've had in the software business. You can really see these shifts. These are pretty big percentage changes.
Just to put it in perspective, the Data Center bar is well over $1 billion in shift, and it's all to scale, so you can kind of get a sense of how we're shifting dollars around for the portfolio. I think it's important to note that this is a place where we can continue to toggle things. The leadership team here, Brian and I, absolutely committed to making investments in the places where we think we get the best long-term return. Overall, our expectations for research and development and in MG&A next year are that we can bring down that spending as a percent of revenue by about 0.5%. That's what we're targeting for next year. Moving on to capital. You can see here the capital spending trend inclusive of the forecast for 2015. 2015 is down significantly from the prior year.
It's down significantly from what we expected at the beginning of the year. As we've said over the course of the year, there's really three main drivers of that. The first is this change in cadence of the timing of the 10 nanometer startup. That had the result of pushing out some of our 10 nanometer CapEx, so it's lower in 2015. Over the course of the year, we also did find a lot of efficiencies on 14 nanometer that came through. Lastly, as we said at the beginning of the year, our view of volume came down, which also impacted our capital spending a little bit. Putting on the forecast for 2016, we expect that CapEx, excluding what we're spending on memory, which I'll talk about in a minute, will be about $8.5 billion.
When you think of it, apples to apples, you see that increase. You can see on the chart it's all in capacity, which would make sense. It really is being driven by the 10 nanometer capacity additions. Everything else is kind of noise. The increase there is all 10 nanometer. In addition to that, we do expect to spend about $1.5 billion of CapEx on putting in capacity for memory in the China factory. That's that start of production there. I'm moving into unit costs. Here I'm showing you the cost trend along with the ASP trend. As in prior years, it's important to show those two things together because remember, mix can be a really important component of both. We've seen a richening mix, so that's a portion of what's going on in costs.
You can see on that cost trend that we saw an uptick in costs this year. We did expect that. We expected an uptick as a result of the increased cost associated with 14 nanometer. As we communicated over the course of 2015, that increase in cost was actually greater than we thought, and that ties to what Bill was showing you as he was showing you where our yield curve has been relative to what we thought at the beginning of the year. I'll click down on that a little bit more. I now want to give you some more specific insight into our unit costs by looking at the cost trends across the performance mainstream and value segments of the PC market. Remember, average cost can be really driven by shifts in mix.
I want to take out that mix impact to show you what's going on in the specific costs. This graph shows our costs through 2015. You'll see an overlay on that, the little dotted white line. You don't have to go back to last year's presentation and try to line them up. That shows you what we expected last year. That white line is the expectation for costs that we had a year ago for this year. What you see is that we did expect costs to be up this year, but in the mainstream and value segments, the costs are up quite a bit more than we thought.
If you listened to us over the course of the year, you heard us talking about that in terms of the 14 nanometer yields, and Bill showed you how we're rapidly catching up to that trend, and where we are today. It's interesting, the one place where costs are better than we expected is in the performance segment, and what's driving that is that the desktop products haven't yet transitioned over to 14 nanometer. We're kind of in the process of doing that as we speak. What you're seeing there is that impact of a really competitive 22 nanometer process that we talked about earlier. Now building this out to show you the forecast that we have for the average cost in 2016 across these segments.
What you can see here is that the mainstream and value costs come down pretty significantly over the course of the year or on average for the year. We're achieving a really good cost structure overall in those segments. In the performance segment, what you see there is the impact of that transition of the desktop products over to 14 nanometer. Remember, Bill showed you those 14 nanometer yield curves and how they continued to improve across 2015, and then how we expect them to continue to improve into 2016. Let me then take that and show you how we expect our costs to look as we end 2016. What you see here is not just the full year 2016, but I'm actually showing you where I expect to end the year by these different segments of the business.
What you see here is that as we exit 2016, we expect to see very competitive costs, really across the portfolio, but particularly look at those mainstream and value costs. We'll be achieving kind of an all-time low cost in those segments of the business. I want to tell you this fact. This is the power of Moore's Law. If you look at these trends over, what is that, an eight-year horizon, what you see is that over that eight years, we continue to bring more performance, more features. We're adding a lot of IP into our product line, a lot more graphics, those kinds of things, and yet we can just bring down those costs at that trend.
We're a little bit behind what we expected in terms of 14 nanometer yields, but we're catching up, and you look at where we're going to end the year, that's a really competitive cost. I'm shifting gears now over to overall gross margin. We've been in the high end of our gross margin range now for five out of the last six years, and we expect to be there again next year. Long-term perspective, this is due to our technology leadership, the product segmentation that we've done, and also importantly, a thoughtful approach to what businesses we're in, where we invest, and where we don't invest. You can see the technology leadership in terms of our data center results. You can see it in terms of the bending of the curve in the client ASP trend.
In terms of the investment decisions, you can really see, I think memory is a great example of it. If you go back to where we first increased the gross margin range from 50%-60%, to 55%-65%, one of the big drivers of that was that we were disinvesting in NOR flash. That allowed us to shift up the gross margin curve. I think you've seen over the time period since then, made very thoughtful investments in terms of what we're doing in NAND, how we're driving technology leadership, how we've structured the business to make it profitable, generating cash, and improving our competitive position. That has set us up now to the position to be able to make this bet on a disruptive memory technology. You can kind of see those investment decisions play out.
What it says is that we can enter new businesses, we can do it in a way where we're very profitable. All right. Clicking down a little bit just into next year's gross margin drivers. I wanted to give you a little bit of insight into that. We expect that platform ASPs will be up next year, you want to be a little careful there. It's driven by server mix, it's driven by the improvements in mobile product lines. The client ASP net of those two things we expect to be pretty flat. Server mix improvements in mobile, as a result of the reduction in contra revenue dollars. We expect that costs will be pretty neutral. You see two things going on there, the 14 nanometer costs come down, the mix of 14 nanometer will be much higher next year.
When you net those two things off, cost is pretty neutral. We'll see a little bit of a down associated with an increase in startup cost to gross margin. There's two elements to that. One is we expect to see increased startup costs on 10 nanometer. I know that the change in the timing for 10 nanometer has kind of thrown your models off a little bit, let me click down on that a second. As we went from two years to two and a half years on the transition from 14 nanometer to 10 nanometer, it had the impact of actually lowering our startup costs this year from what we had expected at the beginning of the year. We saw some good news this year associated with startup costs.
We'll see an increase in startup costs next year, it's a more muted increase than what you'd normally expect. Think of this impact of being about a point on gross margin, I think historically, you've seen bigger peaks in startup costs. I think it'll come down some in 2017. Shifted the timing, spreads it over a longer period of time. I think we have less of a pronounced peak and less of a pronounced valley associated with startup costs as a result of this two-and-a-half-year cadence. The other thing that's impacting startup costs is, as I said, the startup costs associated with starting up memory production in Dalian, China, which will impact us this year. Okay. Quickly I'm going to go through cash generation and how we view capital allocation.
This chart is one of my favorites. I'll go quickly through it. It says we generate a lot of cash, $20 billion of cash from operations this year. We're generating on the order of $12 billion of free cash flow, which will be an all-time high for us. We've also been on a journey to bring our net cash balances down over time. You can see on this chart, if you go back a handful of years, we consistently had net cash balances that were well above $20 billion. We've articulated we're trying to get down close to zero. We don't get exactly to zero, but you can see, we're down in kind of the single-digit billions from a net cash balance standpoint. I do want to give you some insight into how we think about this as we close the Altera transaction.
As we close Altera, we do expect the net cash balances will go slightly negative for a period of time. We expect that we'll get back based on the cash generation of our business and their business. We'll get back to approximately net cash zero by the end of 2016. Think of that as the baseline. From that baseline, we do have the ability, we have the cash available in the U.S. to do buybacks, if we choose to do so. It's not a forecast of buybacks, but from that baseline, we definitely do have the capability to do some buybacks. We've shown our cash generation, we've shown how we brought cash balances down. Just want to reiterate the priorities of what we do with that cash. It's first and foremost, invest in our business.
We think that's where we generate the most shareholder return. Second is our longstanding commitment to the dividend program. We think that's a way that we can return cash to you and generate return for you. We continue to target about 40% of our free cash flow to the dividend. Third way we return cash to our shareholders is via the share repurchase program. If you look at how that's played out over a 10-year period of time, what this shows is that we have returned about $100 billion over the past decade between share repurchases and our dividend program. When you add in the investments that we've made in R&D and CapEx, between returning cash to shareholders and the investments we've directly made in our business, this adds up to about a quarter of a trillion dollars over a 10-year time period.
I was very pleased to do this. This kind of stole my thunder that we have to do the 8-K in advance, but I'm super pleased to be the first one to get on stage and tell you that we've increased the dividend $0.08, which I think you've all seen. You can see from this chart our long-term commitment to this program. For reference, at yesterday's close, this takes our dividend yield to something just under 3%, I think is the math. I think it's a very good program, a longstanding commitment from us. We know that it's something valued by you. Okay, just putting it all together here. We're seeing revenue growth that's in the mid-single digits. Remember that 53rd workweek. That's going to throw off your models a little bit.
Gross margin at 62%, spending down a half a point, and capital spending about $8.5 billion excluding the spending on the memory factory, $10 billion including that, and an increase in the dividend. That's all I wanted to say on 2016. We'll share more and more specific guidance in the January earnings call. There's one other topic that I do want to address here. The nice thing about the investor meeting is you can take this longer-term perspective of the business, and I want to come back to a theme that Brian introduced in his talk. The transformation of the company that we're going through is becoming, I think, clear, and you can actually start to now see it in our financials. You can see the strong shift towards us becoming an enterprise company, the strong diversification of our business.
The implication of this on growth rates going forward is actually quite interesting. I think this graphic shows it nicely, and this is the last five years. We're a different company than we were five years ago. Client is still the largest segment, so I'm not taking anything away from that. The other businesses, led by the size of the data center, now are making up well over 40% of our overall revenue mix and a much higher proportion of our operating profit mix. These businesses are growing fast, and they're delivering growth for the company even in a time period. The overall company has been able to grow, even during a time period where we've seen very little growth in the overall client business. That becomes easier going forward because of how large the data center business has become.
I know you love my napkin math, so I want to end my presentation today with just a little bit of napkin math, and this is, I think, pretty important. The non-client businesses have now grown to the point that if they continue to grow, we can grow the company at a fast pace, even in situations where the PC market declines. Let me just explain what you see on the napkin math. This is just holding data center growth in the mid-teens. Remember, it's not a linear march, but we're just kind of saying, hold that growth rate. Various assumptions around the size of the PC market, and you can calculate an overall Intel revenue growth rate CAGR. I want to be clear, this napkin math is not a forecast. It's a CAGR. It's a heuristic to think about the business.
You can read the assumptions that we have on the slide. What it says is that the data center business has now gotten big enough that if we can continue to grow it at levels near our targets, we can generate good Intel growth rates when the PC market is down, and we can generate great Intel growth rates when the PC market is flat. Therein lies the tie-back to our strategy. We want to invest in and grow the data center, the Internet of Things business, the memory business, invest in things like Altera. We want to protect and hopefully even grow our client business. Very importantly, use the scale of the business to generate enduring competitive advantage, profits, cash flows, and long-term shareholder return.
This napkin math is not a forecast, as we go through this exercise, it gives us confidence in our strategy, and it gives us reinforcing of our confidence in our ability to grow. With that, I'm done, I think we're going into Q&A.
Thanks, Stacy.
I'd like to invite Brian and Bill to join us up here. We'll get some chairs set up. We'll have both Trey and Gary on either side of the auditorium with microphones. We would like you to speak into the mics when you ask the questions so that the folks on the bridge can listen to everything that you're saying here. I will also ask that we just try and do one question per person until we've had an opportunity to make it all the way around the auditorium. That'll just ensure fair play to everybody and equal access to everyone. Why don't we start right over here with Vivek.
Thank you. Vivek Arya from Bank of America. Thanks for a very informative analyst day. My question is on gross margins. I think, Bill, you mentioned that you are expecting the yields to catch up on the 14 nanometer side. How much of that assumption is built into your gross margin forecast? Longer term, one chart that you showed is that 14, 10, and seven nanometer cost per transistor is actually below the line. When I couple that with the fact that your mix is going to shift to the more profitable areas, tells me conceptually, your gross margin should be above the peak levels that you have gotten before. Where I'm still a little bit hesitant, to go and call that is just that getting the yields is becoming harder over time.
Just the near-term thing, what is the assumption underlying next year's gross margin on yield? Longer term, how do you think about the puts and takes of gross margins would be helpful. Thank you.
Sure.
Let me start, and then Stacy will correct me if I get parts of it wrong.
I was going to do it the other way. I was going to start and let you correct me.
All right. I like that.
His forecast is we have complete tie between his forecast and what's in the gross margin forecast. Those two things tie out.
One comment on the cost per transistor is cost per transistor. It's not cost per product. Stacy showed you the cost per product, and that's because there are multiple ways to use that benefit. You can use it as a pure cost reduction, but then you wouldn't build more compelling products. The more that you increase the transistor count, the more of that improvement you take as better products. Part of that is what drives a better mix, because if the products are in fact better, people will tend to buy the better ones. You can't take that curve directly. That curve is the fundamental building block. It does not say that in a given generation, the product will be cheaper. It just says that the building block of the transistor that you need to make it will be cheaper.
I think it all is self-consistent.
Yeah.
The only thing that was the miss is that we didn't learn as quickly on yield. Therefore, you see a little bit bigger blip. You look out at Stacy's cost chart, in previous generations, sometimes you see a blip up just depending on way everything lines up. This time it was bigger than we would've liked because we didn't learn quite as fast.
Exactly right. I just add one last thing, the key is, and the reason I want to show you the segmented costs, the key is to be bringing down the cost of each of those segments. That's where Moore's Law really benefits us. We can bring down the cost in a segment, and we can continue to do more and more features and drive more and more performance. Then the mix of the company, though, has shifted more towards the high end and a little away from the low end. That's why you're not necessarily seeing an average cost coming down as much. The cost of each of those segments is coming down consistent with what I think you'd think.
Great. Let's come back over to this side. Michael McConnell.
Yeah. Hi, Mike McConnell, Pacific Crest. Just a quick question on the mobile outlook for next year.
If I just think about you have a full year of SoFIA, which gives you a lower cost structure, and then the contra revenue going away for a full year, you had very nice results relative to the original expectations for this year. Why wouldn't the losses go down by even more than the $100 million? What's working against that?
Sure.
I think there's a couple things. One, there's still a fairly large investment in our modem, and you're trying to get up onto our yearly cadence, and as we're ramping the volume. The volumes, you're right, we're eliminating the contra. We've gotten that. We're continuing to improve our product mix, but we're still ramping our volume, and you've got a declining PC and we're going into the phones, both in full AP units and then modems as well at a fairly cautious rate. It's about volume.
That's exactly right. Part of it is the math. We're getting twice as much improvement on a product margin basis than we did in 2015. Remember, it was about a third of the improvement was product margin in 2015, two-thirds of that $800 million is product margin improvements in 2016. To your point, we're seeing a pretty significant improvement based on, you get a full year of no contra, we've got a full year of SoFIA. We had bigger spending reductions or investment reductions in those segments of the business from 2014 to 2015 than we do from 2015 to 2016. That's only one-third of the improvement in 2016.
Great. Let's come over to this side.
Thank you. Chris Caso from Susquehanna. With regard to the CapEx outlook for next year, the $8.5 billion for next year, it's still at a lower level than where you've been in the past, where you've seen the baseline in the past. Do you think this is a new baseline at around $8.5 billion, or is that more dependent on what's going on with the 10-nanometer investment? I guess, as you go into 2017, you obviously don't want to provide guidance at this point, but just in terms of when that investment hits, we have 10-nanometer and the memory investment sitting in 2017, at least give us some direction at that point.
Are you asking for us to increase CapEx and-
No. I'm just asking the question.
Yeah, we're not going to forecast 2017, but I do think what you're seeing is a reflection of the two-and-a-half-year cycle and aligning the CapEx. That's really what you saw this year, too, the drop in CapEx. I know sometimes people were frustrated that it kind of went down in increments. A lot of that was us first getting to the, yes, it's going to be two and a half years, then it was not wanting to over-correct, and there's a natural DNA within the organization to never overspend, so they tend to approach it from the top slowly. That's why you saw the iterative process this year. I think that you've got two competing trends here. Yes, if you stayed at two and a half, and if the cost of technology stayed relatively flat, you would have these kinds of numbers. There's two competing factors.
One, we are investing in memory, so that added to this year's. Two, Bill showed you that the cost per technology is increasing, and a certain amount of that is in CapEx. That would offset that as well. I'm cautious about saying this is what 2017 will be because all of those things will come into play in 2017. Great. Let's come right here to Stacy.
Hi, Stacy Rasgon with Bernstein. You showed obviously this year, enterprise I think was a little weaker than you would've thought it originally. They'd grown in 2014, it was down a little bit this year. Frankly, I think the anomaly was actually not this year, it was 2014 when it grew. I think that might've been the only time in the last five years where the enterprise piece actually did grow year-over-year. Cloud is obviously very strong, what gives you, I guess, any degree of confidence that the cloud growth is not coming at the expense of the enterprise growth? How can we get confidence that the trend that we've seen over the last few years in enterprise is not a structural trend rather than a cyclical trend?
I can start. I also don't want to take the thunder away from Diane, that she's got a ton of data. First thing I always remind people is that the cloud of today is on people and the cloud of tomorrow will be people plus things. The cloud growth is going to be there for the future. You're right, enterprise hasn't grown, we also showed that clouds are now about to overtake enterprise in size from us as a data center. As it becomes a larger percentage, that's why we look at the data center in general and say we can continue to make it to grow. It's the cloud, it's networking, it's storage, it's HPC, it's all of those. There is some component, I guess I want to bluntly answer your question.
There is a significant component of cloud growth that's taking from enterprise growth as people move to public and private and on-prem and off-prem clouds. It's still growing faster than the rate of either slowed growth or decline of the enterprise, and it's overtaking the enterprise.
Stacy, we'll come back to that in more detail when Diane speaks.
I don't want to take her data away. She's got great data.
Why don't we move over to this side?
Hi, it's Timothy Arcuri, Cowen. I had a question, really, two-part question. Number one, can you talk a little bit about the thought process on Dalian and why you decided to go your own, number one? Number two, maybe just bigger picture, can you get to where you need to be in memory? You're talking more about that as being a big piece of your growth strategy going forward. Can you get to where you need to be with the current JV structure where you share the IP? Thanks.
I would tell you that, yes, we believe we can get to where we want to be. I mean, the JV structure works very good, and it works on two levels. One, we jointly develop the technologies, and Micron brings a deep understanding of the silicon and processing and fab processing, that since we've let go of our memory fabs, we don't have as much of anymore. We bring the system architecture and what kinds of technology leadership we want to drive. 3D XPoint's a good example of that. The reason for doing Dalian, first, we had that factory, it's a logic factory today. It's not a factory we have to build. It's in existence. It was a good platform in which to move memory into. That was opportunistic.
Second, we see the opportunities in 3D XPoint and 3D NAND both, as consuming our existing capacity. When we looked at that cost structure, it made sense to go ahead and fill up an existing factory with this technology at a better cost now. That's what really drove this decision.
Let's hit Ross over here on this side.
Hi, Ross Seymore from Deutsche Bank. Stacy, this one's for you, and it's on the OpEx side. It's good to see you're getting a little bit of leverage next year. If we look at where your target is, I think longer term, 30% of sales in OpEx, and you're still going to be somewhere in the mid-30s. In the past, you've said you want to grow into that, and I get that, but if you look at where the dollars are going to rise year-over-year, can you talk a little bit about where you're putting the investments?
If you bring in the slide that you showed about how the DCG was going up and the other areas were going down, is there a point at which that kind of harvesting from one side of the company to the other has kind of reached a balance and we should stop seeing any switches and overall the OpEx will slow down?
I would say that He didn't like the OpEx question. We still plan to bring down OpEx as a % of revenue. As we've said over the last couple of years, we're at a point in time where making these investments are really critical, and we believe that we're going to drive significant shareholder return. You can see the kinds of increases we've made on things like the data center, Internet of Things and whatnot. The expectation is that as we make those investments, we have the opportunity to harvest that growth and bring down spending as a % of revenue. It'll come down some next year, but it won't be dramatic next year. The intent is still to do it. I think the other question was, where are the pluses and minuses for 2015? It wouldn't look different than the chart I showed you.
That chart was for 2016. That chart was 2014 to 2016. In 2016, the places where we're increasing investments would be Data Center, Internet of Things, memory, the technology development, and the places where we're decreasing are very consistent with what was on that chart as well.
Okay.
Gary, over here.
Thanks. Blayne Curtis at Barclays. I want to ask you on mobile, with the guidance, you're essentially halving the losses that you started off with. I think the next part probably would require revenue to make that a profitable business. Is that a possible target for you to actually make this a profitable business? You saw many companies pursue modems, and most have bowed out because I think it's a two-pronged problem. You need to get carrier approvals. The smaller OEMs just don't add up to a lot of revenue. To get to the big global SKUs, you need to already prove yourself on the small guys. Where are you in terms of getting traction with carriers and getting proof of concept of volumes such that you even can kind of elephant hunt the big SKUs? Could you actually make this business profitable?
Absolutely, we believe we can make this business profitable. If you take a look at our 7360, we showed you the 7260 is shipping today around the world and including the U.S. with the Asus Zenfone. You can walk into an AT&T store today and buy that. The 7360, which is our next-generation CAT 6 modem, is shipping to customers, and carrier validations today are ongoing. People are out testing the 7360, which is the next generation, for shipment next year in products as well. We've got Spreadtrum and Rockchip both starting to work on products. Spreadtrum's starting to really get into the design mode of that agreement. Rockchip, building on different levels of iterations of their products. We showed you with the Internet of Things. All of those things have to be connected.
That's really, I think about the number of connected devices there, both through modems and connectivity, will grow. Then 5G. The real transition that's going to occur is 5G. 5G is this integrated, full spectrum kind of comms product. We believe that from the base station out through the modem, we have a real opportunity to be one of the only companies that can provide and work with the carriers and the providers end-to-end. We do believe we can get the volume to make this a profitable business. We think we're well-positioned with it. We first had to get our products on a yearly cadence, because if you don't have that yearly cadence, they're not going to take you serious. We're there now, we believe, with the 7360. We've got the next sets of products already starting to come out on our benches as well.
Yes, we think we're well-positioned.
Thanks, Brian. Let's come back over here to John.
Thanks, Mark. John Pitzer with Credit Suisse. Stacy, maybe another way to ask Ross's OpEx question. As you talked about the revenue mix transformation of the company, you've kind of grouped IOTG memory and software into a bucket that's helped driven the revenue mix. If you look at the op margins of those businesses, IOTG is pretty close to corporate, but still kind of below, but the other two are well below. I'm kind of curious, as those businesses continue to outgrow the core business, how should we think about operating margins in those businesses separately? Do they ever get to kind of the corporate average and kind of what's the target there?
I think there'll be differences, but I think at least in the horizon that I'm thinking of, they won't be negative on the overall gross margin or operating margin of the company. Software is a good example. Typically, a software company has a higher spending level, but also a higher gross margin. I think we would acknowledge, we haven't been at the op margins that is appropriate for a software business. I think the team has done a great job of pruning, focusing. You can see the improvement we made this year, and I think there's more improvement to come there. It'll be at a lower % overall to the company's average, but it'll be increasing from here. The memory business is another great example.
It's clearly a lower operating margin than the overall business, but a place where it's a really good operating margin, opportunity of growth, where we're going to generate return, get a good return on invested capital. We look at it as a portfolio. Not every business has to get to a 60-plus % gross margin. We do try to manage the portfolio such that we can continue to have a very profitable collection of businesses and grow and generate shareholder value.
Great. All right, we'll take one last question over here on this side.
Great. Thanks, guys. Doug Freedman, Sterne Agee CRT. You've taken questions on gross margins, you've taken questions on OpEx, if I look at the real bottom line, it's all about cash flow. You did have a good year this year due to the CapEx cut, can you give investors any commitment to your long-term growth in cash flow per share? Should we be modeling a 5% per year? What type of rate do you think this enterprise should deliver, given that over the last 10 to 15 years, Moore's Law clearly is driving the company forward, yet the one thing missing, there hasn't really been any cash flow growth over this timeframe. When do we get to see the leverage that should deliver?
I would say, tying it back to Bill's cost per transistor curve. As we continue to drive that cost per transistor, and it won't match exactly in a year, but as we continue to drive that cost per transistor, you should see cash flows that are growing consistent with what's happening with revenue and operating profit of the company.
Great. I think we do have time maybe to sneak one more in over on this side. Have a mic? We'll take this one here.
Thanks. Brett Simpson at Arete. I had a question for Brian on foundry. Bill laid out a thesis in the presentation about cost per transistor. You've got a better cost per transistor than the two leading foundries, both at 14 and 10 nanometer. Can you maybe just lay out how you plan to exploit that in the foundry space? If we fast-forward to 10 nanometer, should we consider Intel as winning meaningful business in the foundry space at that node? Thank you.
We continue to win foundry customers. Again, most of them don't want to have a discussion of the business until they've got product in market. That business does continue to grow. We do believe that at 10 and then 10 nanometers and beyond, that the differentiation starts to become quite large and significant and an opportunity for us to continue to win customers. We have additional customers talking to us, a broader set today. Unfortunately, there's not a lot we can talk about because we're either in discussions or they're in a quiet period that we're just not allowed to talk about.
All right. With that, we'll wrap up the Q&A. Thank you to Brian, Stacy, and Bill, and thank you to all of you as well. I'll invite Diane Bryant to take the stage.
Hey. Can you hear me okay?
Quick.
Pardon me?
Yeah.
Got it. Thanks. Okay. Good afternoon. I'm hoping that I can answer all of your questions, and I'm going to start out with some key overarching statements, which is the underlying drivers of the data center business that we've been talking about for a while now, they remain the same, they remain strong. As BK talked about, there's this fundamental move to cloud computing as an architecture, that is pervasive.
As you know, the world is becoming more and more dependent upon technology, which drives the data center business. You have many, many, many more devices, whether they're consumer devices or things connecting, which invites more services, which invites the growth of the data center. There's this constant underlying growth driver that remains quite robust. The cloud computing as an architecture, as I said, it is pervasive. I know we like to talk about the public cloud service providers, but cloud computing is a fundamental architecture. You can move from mainframes to client-servers to now cloud computing with a significant pop in reach, the expansiveness at a very high efficiency rate. That is an architecture, a fundamental architecture that is fundamentally driving growth.
In addition to our traditional business of microprocessors, we have three new product lines that we'll be launching, and you'll see that that becomes a relatively significant portion of that 15% CAGR that we continue to hold to for the long term. We'll talk about each of these in more detail. Overall then, our 15% CAGR holds, as Sue says. You can see now we're going out to 2019. As Stacy noted, we do expect to land in the low single digits space for this year. Low double digits. Did I say single? Holy cow. Low double digits. Thank you for correcting me. That was good. The low double digits for the year. You can see then how the 15% CAGR, how we look at it by segment. These are the end user segments.
The end users are actually procuring the infrastructure, deploying it, maintaining it. It starts with the comm service provider space. As we keep saying, the network is moving to a cloud-ready network. It is being virtualized, is moving onto standard high-volume server software or server hardware, and that will continue to grow at over 20%. This year, we expect to land at over 50% growth in the comm service provider space, so significant growth. The cloud service provider market, as you heard Stacy say, this year will be over 40% growth, and we continue to say it will continue at a greater than 20% CAGR. The government, academia, and science, so I have to do a little bit of a speech here on this one. This is traditionally the consumers of high-performance computing have been government, academia, and science applications.
We've looked at it as an end user segment, and you can see that it'll continue to grow at about a 20% CAGR. However, this will be the last time that you will see high-performance computing or government, academia, and science broken out as an independent end user segment, and the reason is high-performance computing is becoming a pervasive workload. We see high-performance computing now deployed as a cloud service provider solution. You can see, you can go to Amazon, and you can spin up a top 500 supercomputer in less than 60 minutes. It is a workload that's being delivered and serviced through the cloud service provider market. We also see that with the amazing growth in data analytics solutions, those are solutions that gravitate towards high-performance computing.
You see much more deployments of HPC in a commercial by businesses in support of their data analytics solutions. High-performance computing is no longer an end user segment for us. It's really quite pervasive, we're going to start treating it as a workload, deployed across all of those other segments. On enterprise IT, as Stacy showed you, the server CPU revenue is flat to down, right? It's been flat to down. As we look forward, we expect enterprise IT server CPU revenue to be about flat. Pretty flat, constant out through the future. However, we do see growth in the enterprise IT segment as a whole because there's more that goes into enterprise IT than just server CPUs. You have storage, you have network, we have other components, Ethernet board systems, and then a growing range of products beyond that.
The enterprise IT segment, as stated here, we expect to grow at a less than 5% CAGR, okay? Our growth then, to that point, has historically been driven by CPUs, by processors. In 2015, 12% of our revenue was non-CPU, that's basically chipsets, Ethernet controllers, boards and systems, a small portion of non-CPU products. Going forward, though, we expect the CPU growth to be about 12% of that 15% CAGR, with 3% of our growth coming from new non-CPU products, we'll talk about that as we move forward. 22%, if you go out to 2019, 22% of our revenue will come from non-CPU components. The move to cloud computing, we want to spend a little time on this because cloud computing, as I said, it's an architecture.
It's an architecture that was first deployed by the public cloud service providers, clearly, it's an architecture that is pervasive, it's being deployed across all segments. If you look inside of each of these end user segments, we have a kind of lighter shaded area, that's the growth of cloud as an architecture being deployed within that end user segment. It's the on-demand, self-serve, fully automated attributes of cloud computing that are required as the dependency upon ICT continues to grow and as the number of devices and services pulling from that infrastructure continues to grow. On the right-hand side, it's how do we benefit from that? How do we benefit from cloud computing becoming a pervasive architecture across all end user segments?
First of all, we benefit through accelerated technology adoption, we've seen these patterns now over the past 10 years from the public cloud service provider market. When you're running a business that is dependent upon IT, you want to get the latest and greatest next-generation technology deployed as quickly as you can. That's that focus on the bottom line of your business, we see an accelerated adoption of our next-generation technology in the move to cloud computing. We also benefit from an accelerated refresh rate. Again, if you're dependent upon technology to run your business, making sure that infrastructure is up to date, getting you the very best performance per TCO possible is important. There is an accelerated refresh that comes with that move.
Obviously on the network side of the house, as BK noted, it's an area that we have a very small share today. The move to cloud computing is a growth in our total share of that market, a significant growth. Cloud computing also enables the delivery of new businesses. You see this all the time. You now have new connected services that you couldn't have deployed prior to cloud computing. Through that, we benefit from TAM expansion, the market growing overall, and we also benefit from an ASP uplift. We've seen this trend, and I'll talk about it a little bit more. That monetization of the IT service means that you have a focus now on performance and the performance per TCO. If we just look now at the public cloud service provider market.
A few points I want to make here. The main point is that this market is diversifying significantly. As you would expect, as the cloud computing architecture continues to mature, you have more and more entrants into cloud computing. As we would all say, a diverse market is a nice, healthy market. This is a good thing for us. The Super Seven, the Big Seven, the Super Seven, as we call them, they are still big. They're about half, or just over half of the total public cloud service provider market is the top seven guys, the big seven guys. They are big, and they are still growing, obviously. What's interesting is if you look at that next wave, and it's really, when we say the next 50, it's actually 52.
You look at the next wave of public cloud service providers, they are growing at an even faster rate. These are big names. They're not small companies. JD.com, the second-largest retailer out of China, e-commerce retailer, grew at 6x over this time frame. Uber, we all know Uber, grew at 10x over this time frame. Dropbox 8x, SoftLayer 7x. You have some really big cloud service providers that are below the Super Seven, but growing at a very nice clip. We also, maybe it's just us because we're hardware people, we think about the cloud service provider market, and we think about infrastructure as a service, the deployment of compute capacity. The other thing I want to point out here is infrastructure as a service is not the largest portion of the market, clearly. It's about 16% this year.
Platform as a service, PaaS, is less than that. It's at 10%. As you can see, the bulk of the public cloud service provider market is really SaaS, software as a service. Again, when you think about software as a service, by definition, it's going to drive a highly diversified market. Everything from Twitter to Oracle, if you attended Oracle OpenWorld, you heard about how they've moved all of their ported, rearchitected all of their applications to be cloud-based. You have Oracle, eBay, Workday, just a real breadth of different types of services, consumer services, and also enterprise services.
The last point I want to make here is that the growth of local data centers in support of local consumption is truly occurring, because often we think about the public cloud service provider market, and we think about the big North American companies. It is truly worldwide. It is truly diverse. You have lots of reasons for it, whether it's data sovereignty issues, control and security, whether it's just local cultural and language issues. You want to have local solutions, local services. An example, Flipkart is the largest e-commerce company now out of India. You have Naver, the fifth-largest search company out of South Korea. As I said, JD.com, now the second-largest e-commerce company worldwide out of China. There's great geographical diversification of this market as well. Okay? Great diversification in the public cloud service provider market.
To address, I think what the question was is, gee, isn't the lack of enterprise growth a result of the fact that enterprises are moving their applications out of their on-premise private clouds into the public cloud market? The data here on the left shows that if you looked at the entire public cloud service provider market and said, what types of services are they deploying? Two-thirds of the services that they're deploying are consumer-based services like Twitter and Facebook. One-third of the services are business. That number has grown. It used to be about 75/25. We think it'll stabilize out at about 60/40. The reason it is growing is what we've been seeing since, gosh, 2007, 2008, is that consumers are adopting technology before enterprise. It's the consumerization of IT, as we used to say.
You saw consumers adopting cloud services first, enterprise comes along and adopts at a later, slower pace. Of that enterprise piece then, that one-third, we've done our own assessment, and it's not perfect, but we've done a lot of research into what are those services, what is motivating the use of those services by an enterprise, and it's about 50/50. About 50% of those services are true cannibalization of enterprise IT purchases, and about 50% of it is TAM expansion, new services. Every CIO is going through their applications, and they're deciding which of those applications are non-differentiating for their business, don't have particular data sovereignty concerns or data regulatory issues, don't have high security issues, aren't mission-critical. If they don't meet those criteria, you're certainly better off pushing it to a public cloud service provider.
They're also then looking at the public cloud service provider offerings and saying, "Which of those services" That, by definition then, is true cannibalization. I used to run email on-premise. Email is non-differentiating. I'm going to move it off-premise to Office 365. That's direct cannibalization of the enterprise market. They're also then looking at those services and saying, "Which of those services being offered I could never offer to my employee base, my line of business, but I'm going to now take advantage of the fact that there is a public cloud service provider offering." Examples of this are video conferencing. Video conferencing is incredibly complex. I don't know how many of your companies have a good, solid video conferencing solution, but it's a very complex solution to deploy for IT. IT is saying, "Well, I couldn't have deployed video conferencing before.
Now I'll use Cisco Webex." That is TAM expansion for us because you're now deploying a service that you wouldn't have deployed prior. Another easy one is DocuSign. If we didn't have DocuSign, we would literally still be walking around contracts and manually signing them, and then the admin takes them to the next, and some lawyer puts them in the bottom drawer of his desk, and then we're not quite sure where it ends up after that. None of us would have thought to deploy an online secure document system, but DocuSign comes out, and it's a great new service for enterprise IT to adopt that they wouldn't have naturally adopt. That, again, is TAM expansion for us. That one-third of the market, 50/50 between cannibalization and TAM expansion.
On the right here, we see a very strong connection between what we do and what cloud computing values. Again, why we see this move to a cloud computing architecture as a good thing for us and why we see the nice growth. We're using here the Super Seven because they have been in business running cloud architectures longer than anyone. We look at them as a benchmark. Again, as I said, when your business is IT, you're going to invest to maximize the performance of that infrastructure and the performance per TCO. You're going to run it like a business. The Super Seven, their ASPs will grow for us this year by about 10%. That's twice the rest of the market.
That's because, as I think Stacy was the one that said, we see our ASPs going up because seven out of seven of them this year decided to buy up the CPU stack. Every generation, we give about 20% performance at a constant price. They're deciding that that 20% performance improvement gen over gen isn't enough. They want even greater performance, they're opting to buy further up the ASP stack. We see a nice continued increase in our ASPs into the cloud service provider market that reflects 80% of their volume. 80% of their volume this year actually went up the stack to a higher ASP. They're also, as I said, the first to adopt our latest technology. Five out of seven of them now have custom CPUs. We do very targeted solutions for them, giving them greater value, greater performance.
They obviously pay for that value. We'll have 50% of all of our CPUs shipped to the Super Seven will be custom CPUs this year. Half of the volume going to them is custom. Our first silicon photonics samples shipped last month, we're very happy to say, and they were the rapid consumers of those samples. They're first to deploy, first to get their hands on silicon photonics. They are the first ones that will be receiving our integrated FPGA and Xeon product that we're very excited about. We just announced yesterday we will be shipping that in Q1. Our first integrated products, it's obviously a multi-chip package. We're acquiring the die from Altera and then integrating it into an MCP and shipping that out early next year. They're very hungry for these solutions.
If you look at the cloud service providers, it's all about optimizing their particular workload, and if they can take their algorithm and accelerate it through an FPGA and get all the benefits of general purpose processing of Xeon, gives them a tremendous performance per TCO advantage. They continue to put pressure on us to give them the next-generation processors as fast as we possibly can, and they're even willing to take errata just to get it into their hands sooner. We already have been shipping 14-nanometer Broadwell to the top public cloud service providers as well as some of the HPC customers, even though our actual launch of Broadwell is early next year. They're the first to adopt. They're always driving us to get the next technology as soon as they possibly can. Enter into the private cloud, enterprise private cloud deployments.
The deployment of private cloud by enterprise IT is growing at about a 14% CAGR. It'll be about 12% of their total spend this year will be in the deployment of private clouds. When you survey, and this is an external survey, if you survey CIOs, enterprise IT decision-makers, and you ask them how they see their future in the use of public clouds or private clouds, you still see that overwhelmingly, they either leave it as a traditional virtualized environment, or they're going to deploy a private cloud. There's lots of reasons for that. As I said, business-critical applications, you want to have control and run them, security issues, regulatory issues. If you're a large enough enterprise, you can run that environment at a lower cost than what you will pay to have it as a service.
To be honest, there's also just a general fear of lock-in. They don't want to become locked into a given vendor, and unfortunately, the cloud market, there's still a lack of standardization. Once you pick a cloud service provider vendor, there is that concern of being locked in. All of those play into the fact that you still see them leaning towards the desire to deploy a private cloud solution. The other very interesting thing that we see at a quite nice rapid rate is that traditional enterprise companies are figuring out ways to use cloud computing to deploy all new business opportunities, so new revenue streams for their companies. If you look at GE has publicly stated that they're moving a lot of their traditional enterprise IT applications to AWS.
At the same time, they're deploying their own private cloud in support of a new business. They call it Predix, which is an analytics service for their industrial operations. You can get industrial data and analytics services for their aviation, for energy, healthcare, transportation. It's a service they provide on top of their standard product offerings. If you look, BMW actually states that they will increasingly become a cloud-focused company. That's what they will publicly state, so they have their connected driver services. They say they'll have 10 million connected cars by 2018, which means 100 million requests per day or a terabyte of data per day. They're using those connections to deploy new services to their customers, like intelligent emergency calls, which is actually interesting.
When your airbag deploys through all these connected sensors and insights into the vehicle, they know how many passengers were in the car, what the extent of the damage is, where the car actually is, they can get emergency services to you immediately. They're coming up with new ways, new services, new business revenues, given the fact that their cars are now connected. John Deere was founded in 1837. You talk about an old industry, and yet they now have deployed a new revenue stream based on connected services, connected cloud, and big data analytics. When you buy their equipment, you can also buy an add-on subscription that will allow you to get information on your production.
Using satellite data, combining it with data from the water sensors and chemicals, and actually from the equipment itself, they can give you insight on how to maximize the productivity of your fields. Honeywell, they have a new revenue stream as well around connected big data analytics. They're using data analytics services. They're actually monitoring first respondents. You can track the first respondents with wearables and sensors on the Honeywell equipment. They can track CO2 levels, heart rate. They can tell if a man is down, the body posture, they can get someone in to take care of them. They can track gestures, they can monitor signals, and through machine learning algorithms, it becomes more and more accurate. Another example of a historical business that is using cloud computing and big data analytics to create an all-new revenue stream.
As these are being deployed, they are predominantly being deployed as private cloud solutions. Again, that dependency, if this is your business and you're running your business, you want to have control over it. This is the fun one. We talk about the transformation of the network. This is where it is a lot of opportunity for us to grow into what is a very large business. We talk about the network itself being transformed from fixed-function proprietary boxes running on either ASICs or ASSPs, moving on to standard high-volume Intel architecture-based servers. Last year, it was a $17 billion silicon market. If you compare that is the same size as the server CPU business. If you think about it, in 1997, we were 0% of the server CPU business.
We're now 96% of the server CPU business, and that business is an $18 billion business. We have a comparable size business, $17 billion. You can see how it was made up in 2014 between FPGAs, which Altera has a very nice share of today. ASICs and ASSPs, ASICs, the TEMs, a lot of the TEMs have their own ASIC solutions or the ASSP market, merchant silicon market, and then processors at the bottom, which we had about $1.2 billion of that CPU business last year. This move of the network is driven by the desire to get the economies of scale of standard high-volume solutions, get the CapEx and the OpEx savings, of course, but to also for the network providers to be able to deploy new services and new businesses and actually monetize their networks through cloud computing.
You can see our share in the middle there. We were at 5%, going to 7%, we'll probably rest of the year right around 9.5%. As I said, we'll grow over 50% this year. The market is growing less than 5%, that means we're growing at 10x the market. That isn't a forecast, that's a historical. We're not going to likely sustain a 10x the market, but it's been growing very nicely as you see the carrier service providers or the comm service providers moving over.
Just for you, just for this audience, and I do mean that, for the first time ever, the large carriers that you see there gave us permission to tell you that not only are they moving to a virtualized cloud-ready network, which they have said many times over and over, today, for the first time, they said we could tell you that it's also to Intel architecture. AT&T, Verizon, SK Telecom, and Vodafone all gave us permission to say, instead of being vague and saying they're moving to a virtualized network, and it could be anyone, they're willing to say they're moving to a virtualized network, and it is Intel architecture. You see the big opportunity.
You can see the savings they're already seeing in their deployment of virtualized environments, great benefits, and the opportunity to monetize that through self-service networks, through new features like consumer services, like immersive augmented reality, through new enterprise services like mission-critical services. Autonomous vehicles is a great one. If you're going to have a driverless car, you want to make sure you have a very robust network connection, very high reliability. You have things like that. You need cloud computing, you can slice up the network and guarantee a certain quality of service. There's all kinds of business opportunities for them as they move to a cloud environment.
The last is we do have three new products, I know this is a busy slide, I want to hit on each of them because in each one of these, we have a very strong, clear, differentiated position, a strong competitive position, some underlying reason why we believe that we will win in these markets and be leaders in these markets. The first one is silicon photonics, our competitive advantage here is Mr. Holt, our technology, our manufacturing capability. We are able uniquely to integrate the laser material with the integrated circuits of the modulator and the receiver. This is all manufactured on a single piece of silicon together. That integration does all the things that we know integration does. It gives you lower cost, smaller form factor, lower power, all those benefits, and it's a fundamental differentiator and value proposition we have.
As I said, the cloud service providers is the first place we're targeting. They have insatiable demand for network bandwidth, and this is a way of getting that bandwidth, very high rates at very long distance, thanks to the integration. It's a very big opportunity for us. As I said, we're shipping samples already, and we're going into production early next year. Oh, dear. There we go. Okay. The second one I want to hit on is Omni-Path. This is the high-performance fabric for high-performance computing as an initial target. You'll see the TAM there is the TAM just for the high-performance computing market. We do believe over time it will expand into the cloud service provider market as well. This is really the competitive solution to next generation InfiniBand EDR, so 100G high performance fabric.
Our fundamental advantage here is an architectural advantage. We have architected Omni-Path to take advantage of the Xeon and Xeon Phi processors. It's an unload solution, so we get a fundamental better performance, better latency, and that drives actually a lower cost savings as well, system power. We have a much more optimized solution than the InfiniBand EDR, which, as you know, the dominant provider of that is Mellanox. This value proposition here is for a discrete PCI Express Omni-Path controller solution. It just gets better when we then integrate it into a multi-chip package at the end of next year, and then it just gets better again when we integrate it onto a single piece of silicon with Xeon and Xeon Phi. That value proposition there is just the value proposition of the architecture.
It will continue to improve. The last one here to talk about, you heard BK and Stacy talk about the wonders of 3D XPoint. I have Rob Crooke to thank for this technology. As a technology, it is a game changer. It allows us to get a much larger memory footprint with big data growth as it is, you can imagine. We can get 4x the memory footprint in a two-socket Xeon. We have a fundamental cost advantage versus DRAM, so the cost is up to half the cost. That's not a price statement. We'll figure out how to price, and we know the DRAM pricing market changes quite dynamically, but we have a fundamental underlying cost advantage. We're targeting this with our next generation platform, the Skylake platform. We'll start sampling to customers early next year and be ready for the Skylake platform.
You can see there, the memory market is obviously a very large market, so that's the TAM of the entire memory market. It's a great opportunity for us in a space that we haven't been before. All three of these product lines, we've made relatively modest assumptions about our initial entry into the space, but we do plan to have a leadership position in all three of them based on our fundamental competitive advantage. To conclude, the underlying drivers of the data center growth are very sound, and our strategy remains unchanged. Cloud computing as an architecture is pervasive, and I think we'll look back and we'll compare it to the impact that PCs had on the extended reach and impact as a technology had back in the '90s. Our 15% CAGR still holds.
As I said, 12% of that coming from microprocessors across server, storage, and network. Then 3% of that coming from non-CPU products, including the new three product lines that we'll be announcing. With that, I will conclude, and I will introduce to you Kirk Skaugen, the Senior Vice President and General Manager of the Client Group.
Okay, well, good morning. I'm excited to talk about some of the advancements we've had in the Client Computing Group. Remember, January 1st, we merged our Mobile Computing Group with the PC Client Group into this new entity representing PC, tablet, phone, and all of our wired and wireless connectivity. There's three things I wanted to talk about today. One is we absolutely are seeing PC stabilization, and we think that one of the big takeaways I hope you get is that there's unprecedented innovation here coming out of both existing partners and, believe it or not, new entrants into the PC space, specifically in the areas of two-in-one. The second is Intel has always been about intelligent integration.
You're going to see an integration roadmap from us where we're taking discrete componentry, very much like Diane just talked about in fabrics, and we're integrating those on to our integrated SOCs across the board from PC down to phone and tablet. A better part of segmentation. We are seeing tremendous opportunity to sell up with a very simple good, better, best strategy, and that's helping in a declining TAM to improve our average selling prices. In the tablet space, we're confident that we're growing at market while we're improving our profitability. I'll talk about the forecast for next year. Again, we believe we'll be growing at market for tablet next year. We're expanding our wireless investments, both in 3G and LTE, both in discrete modems as well as in integrated SOCs like SoFIA, as well as in the Wi-Fi and WiGig space as well.
We're going to kind of drill into the loss reduction that Stacy talked about. Let's talk about PC stabilization first. We're in a very unique opportunity, right? If you thought about that growth of the iPad, there's no doubt that the PC life cycles extended out through that period. What we're now seeing is a saturation in tablets. Tablets declining likely double digits this year. What we're now seeing is there's more than a billion PCs in the installed base that are more than three years old, and maybe more importantly, 600 million PCs in the installed base that are four to five years old. When we do the surveys, we're interviewing thousands of people on a monthly basis, and what we're hearing loud and clear, very consistently, is notebook buyers want to refresh their notebook. Desktop buyers want to refresh their desktops.
You've seen third-party surveys from Gartner now, for the first time ever, say that more people that have a tablet want to refresh their tablet with a PC than people who have a PC that say they're going to refresh with a tablet. That's a fundamental shift that we've seen over the last year or so. You see articles here saying there's never been a better time to buy a PC. We believe that. I'll share with you why on sixth-generation Core and Windows 10. Just most recently, in the last couple days, CNBC saying the PC is suddenly alive again based on this innovation. There's three strategies I want to talk to you today about why we think PCs are stabilizing. One is the form factor innovation we're seeing.
Very much like high-performance computing and cloud and networking were very different workloads and even form factors in the server space. We're seeing the traditional notebook and traditional tower shrink while these other form factors are growing rapidly. The second is we've expanded our brands this year. We've added levels into Intel Atom, X3, X5, X7, to differentiate SoFIA and Cherry Trail. We've also added levels because of the dynamic range of Moore's Law on our leadership Core M products, targeting twice the performance of Intel Atom or Arm in a fanless envelope, M3, M5, M7. This very simple sellup philosophy is working really well for us, and I'll drill into that.
What are we going to do around these or experiences to rid ourselves of the keyboard and mouse that we've been using for several decades as we move to something I'll call ambient computing, where computing's all around you, not just in a one-to-one fashion. We truly believe, sitting here today, that with the announcement of sixth-generation Core, this is our best processor ever. Not only do we announce a refresh on sixth-gen Core, but we also announced for the first time putting Xeon into a mobile workstation. We now plan to have over 150 designs in the holiday. This is a result of over a four-year collaboration with Windows 10. I don't think Intel and Microsoft have ever worked as closely as we have on Windows 10 and sixth-generation Core.
When you look at that aging install base, and you look at about half a billion PCs that are now out there in that four to five-year range, you're literally able to upgrade now with a performance that's 150% faster, with graphics that are 30 times the level that you had on a five-year-old PC, three times longer battery life, really no compromise versus a tablet, 10 years, and probably a PC that's half as thick and half the weight. Working really closely with the OS and hardware architecture, delivering our most secure platform ever. In fact, in the last year, we've seen a significant increase in the amount of enterprises that are going to pull in their Windows 10 deployments. As they learn more and more about Microsoft's plans there, we're seeing a pretty significant pull-in of Windows 10 deployments for enterprise.
As Brian mentioned, Kaby Lake, just want to verify, on track for 2016, Cannon Lake, our first 10 nanometer, on track for 2017. This is a great point in time, and it's only going to get better as we look out to 2016 and 2017. If you look at that 5-year-old PC, traditionally it would've been a traditional notebook or a pretty large desktop, and you just look at the breadth of innovation that's been brought out today, whether it's 2-in-1s, purpose-built gaming notebooks, mobile workstations, and on the desktop side as well. I'll show you everything from a Compute Stick up through purpose-built gaming towers. It's easier just to show you why we have confidence that now is the time to refresh.
If you believe more than 90% of the people in notebook want to refresh their notebook with a notebook, this is traditionally a mainstream kind of notebook out in the 2010 timeframe. Dual drive, optical disc, lots of old legacy connectors. This is a Lenovo product. If you compare that now to a best-in-class Core i7, you're now talking about something that's 1.7 pounds, just nine millimeters thin, and this is a full Core i7 class performance. Of course, in my opinion, there's no reason anyone will need a traditional clamshell. Everything will convert over time. This now starting at $699. Real mainstream price points. If you're a tablet first user, the same kind of thing. Just very simple detaches now. You get the best of a notebook, a full mechanical keyboard, and Lenovo, HP, Dell, and others coming out with great detachables as well.
That's, I think, what's happening as we move to a world of 2-in-1s. Now, this is the tough demo here. Imagine you're a professional Adobe user. This is a dual Xeon. It's about 65 pounds. It's 8-core Xeon. This is what mobility meant to the Xeon market 3 years ago. That's literally the best-in-class 2-socket Xeon workstation. What we just announced from Dell is higher performance now in one-tenth the weight. This is a new Xeon workstation, higher performance than that machine just 3 years ago and now in a notebook form factor. Fully certified with all the major applications that you would expect, like Adobe, Pro/ENGINEER, et cetera. Huge innovation. You see the edge-to-edge display, just a stunning 4K display here as well.
The Xeon is obviously allowing us to sell up from a traditional Core i3, i5, i7, full vPro, full management and security. Great innovation here. What does that mean, and does this thing really matter? Why do we get confidence that with these growing products, we can help stabilize the market? This is just a 2-in-1 example. A few years ago, we said Ultrabooks would drive this way, the best of a notebook and the best of a tablet in a single device. The yellow line is U.S. retail or North America retail without the 2-in-1 category. As you look at that, you would have had decline in the notebook year-over-year.
As you can see now, when you add two-in-ones into the notebook category, we've had pretty consecutive growth year-over-year, and this is one of the most important retail markets. With Windows 10 and now two-in-one awareness, when you detach that two-in-one, Windows 10 operates as a tablet. When you put it together, it gives you that same look and feel of Windows 7 with a tiled infrastructure and a mouse and keyboard feel. Great innovation there. We've seen over 40% growth in two-in-ones 2014 to 2015. We're actually projecting now that we'll have over 50% growth as we look forward to next year. We're doing several thousand exit interviews every month in retail. Here's what it tells us. If you are a traditional notebook buyer, you're refreshing out of want versus need.
You see these new mechanicals, you say, "Hey, this is interesting. I want this." It's not just because my hard drive broke or my screen cracked or something. We're seeing an 8 to 12-month pull-in of the TAM from somebody who was going to buy a notebook to someone who buys something like that Lenovo Yoga. In the two-in-one detachable, we ask people, "What did you walk in the store with?" 40% of the people that are buying two-in-ones would have bought a premium tablet. We really are seeing that conversion, not just in pulling in the total market for traditional notebook buyers, but also a sellover from the tablet market. Of course, we have new entrants. This is the Surface Book. If you've been to the store, you know they're doing incredibly well, and they've been sold out for a while.
Again, a new entrance, a pretty simple detach. A gorgeous keyboard, every bit a notebook, but a tremendous stylus experience and tablet as well. I do expect because two-in-ones are now the second fastest-growing part of mobile after phablets, that we'll have more entrants that have never been in the PC market announcing their presence in 2016, and I'm kind of excited about that. In addition, you've got CTE or the China Tech Ecosystem. What's happened there is we had about, as we said, 40 million tablets next year. About 44% of those came out of China. As the tablet market has decreased, they're doing two things. One is the Intel presence we've had there, they're going down into the phablet category with Intel. On the high end to get more margin, they're building these products like two-in-ones.
You see people like Yifang now in Walmart at $149 selling two-in-one detachables down at $149. Over 110 designs now coming from China as well. On the desktop side, same kind of thing. In gaming, 26% growth now in gaming systems. You've seen HP reenter with their Omen brand, the Predator systems from Acer. Yet, one of our largest desktop customers is still the white box channel. About 1.2 billion gamers out there, $115 billion market today. PC gaming crossed over console gaming just a few years ago. It's never going back in terms of software. About one in 10 people in the world consider them an active PC gamer. That means that they're literally buying software for a PC game once a month. That's a pretty amazing number, over 700 million people. This is iBUYPOWER.
We added K SKUs unlocked, super fast, Core i7 records here. You start seeing purpose-built gaming notebooks as well, very aggressive IDs. People are taking these on the road. We have Intel Extreme Masters coming to San Jose this weekend for the World Championships. We're literally putting, in some cases, 20,000-40,000 people in a stadium to watch PC gamers compete for the World Championships. To differentiate and just to give you an example of sell up within gaming, you take that same ID and you now put it in a tablet. This is an Atom x7 tablet, Dolby Surround Sound four speakers, haptics, so it can actually shakes in your hand when you play the game. This, in many cases, runs applications that have more than 100 million users that won't even run on Arm and will now run on an Atom x7.
It's a great example of selling up from the base level of tablet to something that's much more competitive with Intel. On the mini category, we're also seeing 30% growth in minis. We are able now to shrink a PC form factor down to something as simple as this. If you see this at maybe a hotel lobby or something, this is also going behind things like smart signs, going into Doug Davis' business in social vending machines, point-of-sale terminals, kiosks, et cetera. Now in terms of selling up, you can actually put this now into Core M, so you can actually get a fan-less Compute Stick where this literally just plugs into the back of a smart TV or a TV that makes the TV smart just by plugging this into the back of an HDMI port. Just don't think small has to be cheap.
This is a full Core M3, M5 vPro-based system with all the security and manageability that we sell for ASPs well over $100 just for the CPU itself. Lastly, in all-in-ones, great growth here in what was a traditional, pretty mature category. With all-in-ones, this is just an example of one of the world's first curved all-in-ones, it's a 34-inch HP. If you basically look in there, I can just log in with my face now because it is integrated RealSense camera. When I talk about no passwords, we now believe that your face with a RealSense 3D camera can eliminate passwords forever, basically starting this year. 12-month growth in all-in-ones, eight-month pull-in. People are buying these out of want versus need. Over about half of the people now buying a desktop out of want because it's more attractive looking than out of need.
That's important because four or five years ago, the number 1 and number 2 reason why people bought desktops was something was broke, or it didn't power on, or my fan was broken, those kinds of things. User pain points. We've talked about this for a while, and what's exciting is we're now taking technology readiness and moving it into ramp. Average person right now is carrying around six cables per person to power all their devices, charge cables, USB cables. Average person has between eight and 20 passwords. They have to change them roughly every 90 days. Average person enters a password up to 28 times a day, and the average passwords are the word password and 1234567. Our job is to go eliminate that. Obviously, we've been using a keyboard and a mouse for decades. It's not very ergonomic, and it's a challenge.
Where are we on our path? We're on our vision. It's a multi-year vision to eliminate all wires from computing. With wireless display, we'll ship over 60 million notebooks this year with wireless display capability, with adapters now down as low as $39.99. Moving it into business with professional WiDi and by even putting PCs into the conference room where you can take a small Compute Stick, and anywhere you're on VPN in the world, you can connect wirelessly to your conference room. In wireless data, we acquired Lantiq this year. There's going to be 880 million broadband-connected households by 2018. Intel is now number one in the world in cable gateways. With integrated Atom SoCs, we have over 50% market share. Now with DSL, LTE, and fiber capability, we can expand our gateway presence.
This just happens to be one from Haier in China. We're selling over 40 million gateway chips a year, which is pretty phenomenal, and I think it's a hidden gem that we have here as the number one or number two gateway provider in the world. Obviously, 3G and LTE are fundamental. We'll talk about that. Then in wireless charging, we were excited a few weeks ago to announce the merger of the Power Matters Alliance and A4WP into a single magnetic resonance standard. Our vision is to charge wearables through PCs under a common standard, and that's what AirFuel is all about. People like Starbucks and AT&T have now joined Marriott, Intel, Hilton, and others.
If you just go next door to the Marriott, you'll see that the first proof of concepts are happening where you can sit in the lobby now, charge your phone through the table using that technology. Can you go back a second? Sorry. On the no password side, we are putting in Windows 10 and True Key. It's the combination now where you log in with your face, just very simply like I did on this RealSense system. We have a common security spec with Microsoft, and then once you're logged in, Intel Security and True Key take over all the rest of your passwords for all your sites. This is all in production now.
The registration process is going to get much easier as we go to the next versions of Windows, but all the fundamental building blocks are there in a highly secure fashion to eliminate passwords. Then on natural user interfaces, we're expanding with Cortana, and we have this concept of ambient computing. If any of you use Amazon Echo is probably the best example today, where you literally just sit in the kitchen, and from 10 feet away from your device, you can just ask it the stock price, the weather, order things in your Prime account. Over time, we think this is going to be true across all computers, where voice will just be a natural interface. With RealSense now developing over 100 applications for the 3D cameras for both user-facing and world-facing. Next is segmentation, integration, and IP reuse.
There's several things we've done on PC segmentation to try to hold the average selling prices here. The first is around cost. If you go back three or four years ago, our Celeron and Pentium products, which is where a large majority of PC growth was all built on big cores, on our Core architecture. Now we've moved 80% of our mix on Celeron and Pentium over to the Intel Atom microarchitecture base. Much more cost-effective product line. What that's meant is our Chrome share now has grown from in the low 30s to now over 80% of Chrome share. You've actually seen tablets in K-12 education shrink year-on-year, while Chrome, our estimates, could be up as high as 50% year-on-year in education, with Intel the leader there.
As Stacy and Bill showed you, we're going to get an advantage this year on CCG, where we moved over to 14 nanometer last year, now we're going to see the benefits of Moore's Law on 14 nanometer across our product line this year. On integration, we have a history of intelligent integration. If you look at audio, Iris graphics, touch controllers, image processing units, sensor hubs, and security, all of these, when we integrate, we add a little bit to our average selling price. We give a little back to our OEM, we give a little bit back to the end user. Everybody wins in this process. This is something that Intel's been doing for several years, I'll show you a roadmap there in a second. Lastly, around sell up, vPro, pretty much all-time records, right?
Every time there's a cyber attack on a large enterprise, I can tell you, the next thing that happens is we get a call and we get a 100,000-unit order. Doesn't matter whether it's a point-of-sale terminal that was attacked or a kiosk or just an internal enterprise. vPro is the world's most secure managed processor, we're selling at all-time records there. We talked about the Xeon being a sell-up opportunity, within Core M now with the 6th-gen Core, we're launching M3, M5, M7s. We can sell up in retail just like we have with a record Core mix in the traditional i-series space. What that's meant is over the last several years in a tough TAM environment, our ASP is up 4.5%. Our share against predominantly AMD is up 10%, our Core mix is above 70% at an all-time high.
Remember, that's while we grew most of that share at the low end in Celeron. When we say Core mix is at an all-time high and we raised ASPs, it's while we were aggressively winning that low-end business around Windows Bing and Chrome. That's kind of a double thing there. What I wanted to talk about today is sort of what you should expect over the next several years. Again, intelligent integration. If you just look at Thunderbolt, we announced that Thunderbolt and USB-C are now under a common connector. You can now put power, data, and display all over a single cable into your PC. Six times the number of designs are coming from Broadwell to Skylake. What I'm here today to say is over the next several years, we will integrate Thunderbolt very much like Diane talked about integrating Omni-Path architecture.
Wi-Fi, obviously in 2003 with Centrino, we were a huge Wi-Fi player. Over the last several years, we've grown to more than 40% of PC Wi-Fi, now number one, profitable, and still growing. We'll integrate Wi-Fi over time. WiGig, you saw competitors buy other companies to compete with us. As far as I know, we have now 100% of all the WiGig solutions for docking, both in the notebook as well as in the docks. You can come into work, don't have to have those proprietary docks. Brand-new business for us. That'll be up 150% year-on-year on WiGig as we go from 5th-gen to 6th-gen Core. In Iris, a growing number of Iris wins. You saw Surface Pro join Apple and others of using our Iris technology. We're now over 70% share in graphics and media, the largest graphics company in the world.
We'll have twice as number of designs on Iris and Iris Pro as discrete continues to get squeezed into a smaller and smaller portion on the high end. Thanks again to Rob, Intel Optane coming into PCs, enthusiast first, and then as we go into 2017 into mobile. Those first products will be ramping into the marketplace in the higher-end PC segment in 2016. Lots of opportunity here as we integrate to grow our average selling prices. Okay, last on the mobile side. There's certainly four strategies we're driving in mobile. The first is just ecosystem leadership. When we have products in there, we're aggressively participating in standards. We want all this to happen in a standards-based way, right? We were a leader in USB and PCI Express. As we look forward to 5G is in terms of in-home interconnects.
Obviously, we want your appliances and your home energy management systems and your phones, tablets, and PCs all to be working together in a standards-based way. That's why we joined the open and formed the Open Interconnect Consortium. We are running, in many cases, a board member position on things like the AirFuel Alliance around wireless charging, et cetera. The second thing you've seen us do is aggressively fill out our IP portfolio, because if everything in the world is going to be smart and connected, we have to have all of the wireless and wired connectivity options. Recently, we acquired CDMA assets from VIA that let us get into networks like Verizon and China Mobile. You're seeing us aggressively pursue IP development in WiGig, LTE, NFC, 5G, 3G, everything. On the right side, it's about competing profitably in mobile devices.
The Zenfone, I think, has been a great example of us getting on the AT&T network, $99, Moorefield and 7260. We've got the MeMO Pad 7. You can buy it at AT&T and Walmart. Surface 3 came out. Not only did we win Arm over to Intel, but the LTE option of Surface 3 is using our modem. In CTE now, we have competing in mobile more than 10 reference designs, generating almost 100 different SoFIA SKUs around our 3G and then future LTE solutions. You can get TrekStor, Teclast, Yifang. These are people that you may not know, but what's happening is they were basically under these no-name brands. Now their quality is up to a point where you're going to start seeing some of the large retailers in the United States put that retailer's brand on there because the quality level's up enough.
Brian talked about our relationships with Rockchip and Spreadtrum. Primarily in the phone, we can expand our presence there through Rockchip and Spreadtrum and the proliferations that they bring into the marketplace. Our 3GR or Atom x3 parts with Rockchip are in the market, and you can see just a lot of the examples here, where the presence that we had with China in tablets have now expanded not just down into 2-in-1s, but up into the phone and phablet space as well. As we said, we're going to improve our profitability there again another $800 million. We are investing in a 5G future. We believe if everything is going to be connected and smart, that you have to be a leader in 3G, moving over to 4G and then ultimately in 5G.
We plan to participate in proof of concepts at the Korea Olympics in 2018. You've seen us with public announcements with SK Telecom, 2020 Olympics in Japan, relationships we have publicly now with Docomo, and then you heard about the carrier designs as well. Stacy set our expectations as just to flat to low double-digit growth in revenue. Growth in the low double digits. My job is to do better than that. I think macroeconomics is the one thing we still got to watch, especially relative to the PC business. I think we've been conservative in our forecast here. In summary, stabilizing PC business. We think the innovation has been unprecedented. A lot of that sellover from tablets, but also a refresh from traditional clamshells over to 2-in-1s.
We think that segmentation, just that good, better, best is expanding just from Core, but now to Atom x3, x5, x7 and Core M with M3, M5, M7. Integration shared with you some future over the next several years. The integration to help us ASP will continue. Our forecast for next year is we'll grow at market in the tablets again. Probably a declining market again, as more people go over to a 2-in-1. Expand our wireless portfolio further with both discrete modem and Wi-Fi WiGig and others, and then we'll continue to reduce the losses. With that, I will bring Diane up, and we'll have a quick Q&A.
Thank you.
All right. We'll use the same approach we did last time. If I could just remind you to limit yourself to one question, that way we can get to as many folks as possible. We'll start here on the left side of the room with Ambrish.
Thank you, Mark. Diane, thanks. Thanks to Kirk also, My question is only for Diane. Sorry, don't want to put you on the spot. Overall, your performance on DCG has been phenomenal, You're also very kind to walk us through all the details every time we see you. My question is on the competitive front. What is your expectation as you look out over the next few years? The Arm camp has been, I would say, disappointing to not there so far.
As you look out and you build your expectation for growth, that's my first question. Second one is a little bit more blocking and tackling. When you put out NFV up there, what's the update on the software support? When you say that they're going to be based on IA, what is the timing that we should expect products to roll out? Thank you.
We'll go ahead and take both of those, Ambrish, If I could just remind everyone to limit yourself to one question, please.
They're complicated answers. I'll try to give the answer. On NFV/SDN, there is a big portion of what we are doing, and Intel is very good at this, if you've noticed for a while, is we build ecosystems around next generation technology. We have a Network Builders program. We also have the open NFV program that we're contributing to. We are building an ecosystem with the industry, both on the software side as well as on the hardware side, to support the move to NFV and SDN. As you can see there, we have the benefit of very large service providers pulling that technology through. There is lots of investments being made by all of us across the industry in order to make the transition happen. You said, and when do we expect to see it?
You see today, and I think those stats at the bottom there, you see today deployments of cloud-based infrastructure happening at all of the carriers in production today. You guys know, the telco industry moves slow. It's very complex. It's everything from the edge to the core, back to the data center. You see each of them picking certain portions of the network and deploying a cloud solution, trial, pilot, then trial, then production, and then scale. They're all in different phases of that across the network. It is a multiyear build-out for certain. You can see our growth. As that's happening, we're gaining the share of that market. It'll be a multiyear transformation for sure. I mean, you're talking about a lot of infrastructure that needs to be upgraded.
I know I said I would be short, and now I'm not, am I? I will say, too, the move to 5G is very compelling for them because they look at 5G and say, "I have got to have a virtualized cloud-ready network by the time 5G hits because of the massive change in the amount of demand and new services that have to be delivered." Now is the time for them to make that move in readiness for 5G, which is 2018 to 2020. The first question, what was it?
Arm.
No, you said competition. You said Arm, yeah. There's two main competitors today. You have IBM through OpenPOWER, which is an enterprise-class architecture coming in at the top, trying to win share by opening up their architecture. You have many, many Arm players. Some come in, some drop off. There's always about 16 that are active saying that they're going after the data center business. You might have seen Arm holding their last earnings day. They actually upped their projections for both the percent of the server market. They now say they'll have 25% of the server market by 2020, and they upped their projections on the network, saying that they'll have 45% of the network market by 2020. We take all competition very seriously. If Andy Grove taught me one thing, it's be paranoid. Andy Bryant reinforces it every day.
We take all the competition seriously, but I think the network statement of them having 45% of the network, I hope, is reinforced by the fact that I got four of our biggest customers to say that when they're moving, they're going to move on Intel, because Intel is the only architecture that is virtualized today. There is no virtualized Arm cloud, Arm solution. That network, we'll certainly be in battle with them for the years as this transformation happens, but I think when you have very large carriers like that reinforcing Intel as the destination, I hope that helps. On the server side, they made a statement that they have less than 1% of the server market today. We actually counted all of the Arm-based servers in the world, and it's actually less than one-half of one-tenth of a %.
You could say approximately zero, or you could say less than one, I guess they're about the same. It's a high bar. It's hard to get in. Where we see it, and I don't want to underestimate it, is it is a tool for our customers to use in pricing negotiations. They want us to know that there's always an option for them if they want to port their code and make that investment, they could do it. That was very long, wasn't it? Sorry.
No, it was great.
Okay.
Thank you. Why don't we come over here to Chris?
Hey. I think my question will fit nicely into that one. When you're talking about pricing and ASPs and DCG's seeing massive growth, partly because of cloud growth, partly because of ASP increases. There are potentially some growing risks here. As cloud becomes larger, you have customer concentration issues here, and also these customers are large enough that they could potentially explore x86 alternatives here. I guess around ASPs, how do you view yourselves as getting ASPs right in terms of the value that you're giving them? Of that 15% CAGR, let's say, for DCG, how much do you expect to come from ASP increases going forward?
Yeah. I hope I made the case that the market is actually diversifying, not consolidating. I go back to the enterprise days when the market really was consolidated. We had 3 customers, HP, Dell, and IBM. That was a consolidated market. Now the market is actually very diversified by comparison. Those cloud service providers turn to a wide array of OEMs to deliver their systems for them, so we actually have a very broad offering. To all of our customers, we offer a full range of SKUs, obviously. We have the basic SKUs, the standard SKUs, the advanced SKUs. They are choosing to buy up the stack. 80% of all of our volume has actually moved up the stack from 2010 to 2015, and as I said, 80% of the Super Seven have moved up the stack 2014 to 2015.
To your point of why wouldn't they just use an alternative x86 architecture, which would be AMD, is basically what you're saying. As I said, these folks, they are not competing with us, they're competing with their competitors. Cloud service providers compete with cloud service providers. The infrastructure is their business. They have got to get the very best performance at the very lowest total cost of ownership to run their business. If they don't use the best technology available, they're going to fall inferior to their competitors, whoever their competitor is. Our job at Intel, and thanks to Moore's Law and thanks to everything we're doing through innovation, is to always make sure that we have the best technology solution. Things like 3D XPoint, silicon photonics. We're always innovating in support of their business. In the end, that's how you win the business.
You got to have the best technology.
Great. Let's come back over to this side. Chris.
Thanks, guys. I'm not going to ask the 36-part question. Don't worry, I'm going to stick to the Tim Luke rule and just ask one. Kirk, you're looking lonely up there. I'm going to ask you a question on wireless.
Fair.
You guys talked about taking the cost out by $800 million this year. You're going to exceed that. You're saying you're going to take out another $800 million next year. It's no longer in the high-growth priority box. Can we assume that every year from now on, you'll be taking at least another $800 million out, or what would it take to really take those costs down? I think you're still going to be losing like $two and a half billion in wireless next year.
Would you like us to break that down into 2017, 2018, and 2019? Chris?
I'm not going to forecast beyond 2016, I think we are pleased, and we're micromanaging this business, meeting every single week to be able to meet and exceed the targets we set this year. We have confidence that the $800 million that we've put out for next year is risk adjusted. That's how we went and forecasted it. Beyond that, we won't forecast, I think we want to grow revenue and improve costs. By putting the PC group and the mobility group together, we got about 20% efficiency in the organization structure, our customers think that we're operating with higher velocity, with faster decision-making. I think it was the right thing to do. We got that benefit this year on top of that, we won't forecast beyond 2016.
We need to grow revenue in wireless, get new customers, and there's number 1 thing I've learned since picking up this business. Number 1 thing is in wireless, people want choice. We're being pulled into more deals than we can probably handle right now, which is a good place to be.
Thanks, Chris. Let's go to Ian right here.
Thanks. Question for Diane. How should we think of gross margins in your group as 3D NAND and silicon photonics ramp? Silicon photonics, obviously, Arista and Cisco tried very hard to make that succeed. Why are you optimistic? It had some manufacturing issues, and people were afraid of Cisco's sole source.
Yes. We do, absolutely. Our operating margins have been right around 50%. We have it right around 50%. Absolutely, when our product line portfolio diversifies and we bring in more non-CPU products, that operating margin will come down slightly over time, over the horizon we're talking about. As through scale, our spend today is about 30% of revenue. Our spend will actually decline, and the operating margins will pull up a bit after that as we get the benefits of scale. You're absolutely right. The non-CPU products have a lower gross margin than our CPU product line. You'll see it dip from the 50% to the mid 40%, and then it'll pull back up again towards the end of the planning horizon. Why do I think we're going to be successful with silicon photonics? I have to be careful.
Some of the telco equipment manufacturers that you would think would be investing in their own solutions are actually receiving samples of our silicon photonics. There's tremendous interest in our silicon photonics product. Our fundamental manufacturing capability, we're able to put the 3-5 laser material onto the silicon wafer. No one else can do that because no one else has the availability of silicon process technology at your fingertips. We have a fundamental advantage both in reach and in density that you just don't see in the industry. Our main competitors in silicon photonics would be Luxtera and Mellanox, and we actually believe we will be the first to market with 100 gig, and we'll be a superior solution, is our expectation.
Let me come over here to David.
Thanks so much. Can you give us some idea for your smartphone revenues next year? You've got cost coming down in part because smartphones are ramping. Will that be standalone modems or primarily SoFIA? Roughly what the mix will be between standalone and the integrated SoFIA products.
I think we're not going to talk about specific customer design wins, but I think both will grow. Right now on our discrete modem, we have over 30 unique designs ramping. We talked about some of them like Surface 3 and a lot of the Asus designs. I think we can grow the number of designs on discrete, and then SoFIA will grow not just in what we call 3G and 3GR, which are the two versions that are out there now across form factors, but we'll have the SoFIA LTE solution out, and we're committed to put that into our factory, like we said, in 2016. We're on track for that as well. Both SoFIA LTE and the discrete are out aggressively sampling and putting through the customer networks for production in 2016.
Thank you, David. Why don't we go over here to Joe?
My question's for Kirk also. How do you think about pricing in the client business, particularly in the low end? If you think tablets are receding as a threat, if you think AMD is receding as a threat, can you see actually price improvements in the value segment, and then maybe the whole stack sees the benefit from that?
Well, I think we're seeing great growth in the low end, whether it was the low-end Windows SKUs or Chrome. Education, we just had a multi-hour education review this year or this last week, and Chrome is growing almost 50% year-on-year in K through 12 education, even though it's stayed relatively under 5% in retail. Education is doing really well, while premium tablets and value tablets have actually shrunk year-on-year. We can put innovation everywhere. Whether it's a RealSense camera or anything. We think we'll start seeing innovation. Our issue, I think, was we stopped investing in netbook, and as a result, that category collapsed. If you look at this is a gorgeous machine at $149 at Walmart, we can sell up now.
I think one of the reasons people wanted the m3, m5, m7, and the Intel Atom x3, x5, x7 is we've actually do a really good job training millions of retail salespeople every year. If they understand the simple value proposition of why a Cherry Trail is better than a SoFIA or why an m5 is better than an m3, we do a really good job of selling up. Not only do we have record Core mix, but we have record i5, i7 as well because everybody in the world, whether you're a 16-year-old on two weeks at holiday at Best Buy, you know what Turbo is. You can explain Turbo, and it's i5 and above. We see a large sell up once you get people into Core, up into the five and seven.
We think that same thing will happen in the value space as well now that we have these differentiated brands.
Great. I think we have time for two more questions before we break. Go over here.
Thanks. It's Matt Ramsay from Canaccord. Kirk, you talked about some good stats in your business as the product mix diversifies. 26% in gaming, 30% in minis, all-in-ones, Chrome. Units, you're talking about being slightly down for the PC market. Maybe you could talk about particular regions or product segments where you're concerned and how we should monitor those. Thanks.
Yeah. Diane and I, and then myself ran servers for a long time. There's some analogies, where no one is going to argue that tower servers are going down, or that tower desktops are going down. If you go just look at the Marriott next door. We're replacing an old PC with a new PC behind the counter, but all-in-ones are showing up at the concierge desk and then at the airline ticketing across the hallway, and there's five new all-in-ones if you just go a quarter mile down the street. We call that density. Same thing with all-in-ones, where traditionally you would have replaced your desktop tower in your office at home with a new tower. Now you're starting to see an all-in-one show up in the kitchen, and you're asking it for the stock price in the morning because it has voice recognition with Cortana.
What I mean by that is, I haven't convinced myself, that it's just like Diane was saying, can the cloud grow even if enterprise shrinks? Even if desktop shrinks? We have the opportunity to put more PCs or compute devices in the home, and we have an example like in retail or hospitality, like hotels, to not just replace the PCs where they were, but expand them into new places. That's why if you think optimistically about the TAM in the future, I think we could put those things a lot more.
The PC-like embedded reuse of these NUCs, I think this is fantastic for Doug's business because you're going to find these behind a lot of smart signs, the new Coke machines and Pepsi machines, believe it or not, where you choose your own syrup and everything, have this kind of device in there because they want to have the out-of-band manageability of vPro, whether it's a kiosk or a point-of-sale terminal. We're building it not just for our growth, but I think in a lot of cases, they can get reused into embedded as well. Doug, I just don't think of it as a one-to-one replacement, where you have one tower and it gets replaced by one all-in-one. I think all-in-ones and minis and Compute Sticks could expand pretty rapidly.
If you think about it, every TV in the world, why do you want to be limited to the apps on a smart TV? If you're going to keep that TV for 7-10 years, why do you want to be constrained by the application processor in that TV for your smartness? All you have to do is plug in a Compute Stick, and you could refresh that on a completely different cadence than your TV screen.
Thank you, Matt. Let's take our last question over here with Betsy.
Thanks. Betsy Van Hees with Wedbush Securities. Kirk, during Stacy's presentation, in the memory section, he talked about not only DCG and that we're seeing a great memory growth with enterprise, but he also mentioned the clients as well, and that we start to see that as driving long-term shareholder value. Can we start to see Intel bundling SSDs, 3D NAND SSDs, or 3D XPoint, or how should we be looking at that, and when can we expect that?
Yeah. On the SSD front, we are seeing more and more people move over. I think once you've seen SSD performance, especially in enterprise, which is more than 50% of our business still, nobody ever has come back, I think, after using an SSD, wanting to go back to a hard disk drive. The problem in consumer was you have to convince someone to go from a 500 GB hard drive or TB hard drive to some lower density. In Optane, we'll have products in the market in gaming and enthusiast in 2016, and then we'll expand it through our mobile portfolio, I think, pretty aggressively, and it's a great opportunity for us as we go into 2017.
Rob, did you want to add anything there? All good? Great. All right. That concludes, actually, our formal presentations for the day. I'd like to thank everyone who joined us via the webcast. For those of you who are with us here in the room, I'd like to invite you to join us in the cafe for lunch. If you don't have your table assignment, we'd be happy to help you. You can find it typically either on the back of your sign-in badge or via the app. If that doesn't work, again, ask any one of us, and we'll be able to direct you to the right place. Thank you.