During the course of this meeting, we may make projections or other forward-looking statements regarding future events or the future financial performance of the company and the industry. We wish to caution you that such statements are predictions, and that actual events or results may differ materially. We refer you to the documents the company files on a consolidated basis from time to time with the Securities and Exchange Commission, specifically the company's most recent Form 10-K and Form 10-Q.
These documents contain and identify important factors that could cause the actual results for the company on a consolidated basis to differ materially from those contained in our projections or forward-looking statements. These certain factors can be found in the investor relations section of Micron's website. Although we believe that the expectations reflected in the forward-looking statements are reasonable, we cannot guarantee future results, levels of activity, performance, or achievement. We are under no duty to update any of the forward-looking statements after the date of the presentation to conform these statements to actual results.
Okay, thank you everyone for coming to Micron's 2015 Summer Analyst Day, both those of you that are here in person as well as the webcast. I think it will be a great day for everyone. Our CEO Mark Durcan will kick it off, followed by Mark Adams, our President, Scott DeBoer, our VP of R&D, and then Ernie Maddock, our Chief Financial Officer. Just so everybody knows, we are going to save Q&A for the end. You may want to take notes as we go through in terms of what questions you want to ask. At the very end, we'll bring all the speakers back up for Q&A. With that, I'll turn it over to Mr. Mark Durcan.
Well, thank you, Ivan. Good morning, everyone, and thank you for coming. It's a great pleasure to see so many old familiar faces in the room today. I think we have a pretty good presentation put together for you. It should answer a lot of the questions you have about the company and our focus on a go-forward basis. As Ivan just mentioned, in the event we miss anything that's important to you or that we need to cover, we've set aside plenty of time at the end for questions and answers. As we move through the agenda, we'll try and present material in a free-flow fashion and then come back at the end and answer those questions. See if I can advance the slide here. Clicker's not working. Here we go.
I'll kick off with a few views on where we are from an industry environment perspective, Given that, how Micron plans to focus and execute over the next 18 months to two years. Spend a little bit of time in terms of key strategic imperatives for us and where we really focus on executing. Mark Adams will then join with a lot more detail around how are we interacting with our customers, what are the markets that we're going after and the products that we're going to bring to market, and how we plan to grow the business over the next couple of years. Scott DeBoer will talk about the technology roadmap, the innovation we're driving at Micron, and how that fuels not only the product roadmap, but our ability to build in new value for the customers.
Finally, hopefully, many of you already know Ernie Maddock from his previous jobs and career. Those of you that didn't know him previously, hopefully have met him at this point. He will come up last and talk about the financial strategy of the company, a little bit more about capital allocation, and how we plan on providing information to our investors on a go-forward basis, as well as how we plan on managing the company. Again, I'll come back at the end to talk about Q&A. I thought it would be good to start off today with a view of the last couple of cycles, the last three cycles in the semiconductor industry.
It's not because I want to convince you where we might be in any particular cycle or what the magnitude of any cycle might be, but just to kind of show you the trend, because we are navigating, I think what many people think are challenging circumstances at the moment. There is a lot that's positive when we actually look at the data in terms of what's happening in the memory industry. If we go back before the financial crisis in 2008, you can see what was a pretty severe semiconductor cycle, down roughly 25% negative growth at the bottom of the cycle, and an even more severe memory cycle with a strong kickback following the end of the crisis.
This cycle is actually probably shorter, attenuated in terms of its period, given the interruption of the financial crisis and what was going on in the global economy, but probably more typical in terms of what a semiconductor cycle looks like if you go back to the previous cycles prior to that. If you look at what's happened since then, L-shaped cycle that followed the financial crisis. What I'm plotting here is peak quarter-over-quarter growth to the quarter before the next peak. I apologize here. There we go. This will work a little bit better. What we're looking at here is peak quarter-over-quarter growth in the semiconductor industry in the dotted lines, and the solid lines showing what's going on in the memory industry.
You can see that next cycle had a big snapback from memory, obviously a very strong cycle at the beginning, then a long, what we think of really as sort of an L-shaped recovery. Not the big upside
At the end of the cycle, not the big downsides that we've seen in some previous cycles. When you look at what's happened since then, you see this continued protracted L recovery, really. A little bit of a minimal peak there in the memory market, up 30% quarter-over-quarter at the peak in the middle of 2013. Really then ongoing, really a ripple sort of performance. I think what you're seeing here, the reason I show you this is not to say where are we in the cycle, but to show you that really now we've had a long period of relative muted volatility in the memory market and a long period of relative outperformance for the memory business.
When we think about what is going on in the memory business today, yes, we're seeing price declines given the historic underperformance of the PC segment over the last a number of quarters. I think really caught a number of us by surprise, but not a cataclysmic cycle. This is a slide now that I think is familiar to many of you, but I think it is supported by what I just showed you, which is that the memory industry really has changed. It's in the data, it's in the cycle, and there's real reasons for it, and you can actually see the detail when you look at the detail of those cycle slides. First of all, consolidated suppliers. We all have sufficient scale.
We're really kind of in an environment where the suppliers in the marketplace appear to be investing for return on the invested capital. If you look at the kind of the ripples that you saw in that last cycle, you don't see these big swings in terms of wafer supply in the marketplace. What you see is the ramifications of small perturbations in the demand cycles. We've got low supply growth, limited new wafer capacity coming online. I'll show you some more data on that here in a little bit. The slowing technology migrations that come with the slowing of Moore's Law, really driving a situation where the incumbents in the market have to intentionally add wafers in order to grow supply faster than demand.
Finally, what I think is maybe the least appreciated change that's gone on in the memory industry, which is the rapid diversification of the end markets we serve and the differentiation of the products going into those end markets. Really creating a situation where memory suppliers today have a lot more opportunity to differentiate their products, a lot more opportunity to add value for end customers, whether it's through component-differentiated functionality or through system-level performance. That broadening of the markets and the ability to target those markets, I think over time, will lead to a memory industry that looks a lot more like the dotted lines in those previous slides or much more like the semiconductor industry overall, but with the enhanced growth that you've seen over the last couple of years. Here we go. What does that look like for DRAM today?
When we think about what's going on in DRAM supply, our view now is that 2015 bit growth will come in right around 25%. We're not through the year yet, so that could still move up or down a small amount, but I think the die is pretty well cast. That's about what we're going to see. On a go-forward basis, as we look at what do we think is happening in the marketplace for 2016, we think it's going to be about the same. We're projecting 24%, and that we believe, while a little bit too much in 2015, is probably just fine given any sort of normal market on a go-forward basis. On the right side of the graph here, what you see is this diversification of end demand, as I mentioned before.
We've got continued shrinking of the most commoditized client segment of the market, continued growth in mobile, which provides opportunity to differentiate and add system value, differentiated form factors. In enterprise, where we can really drive value-added performance and differentiation. Finally, in the embedded markets, not growing as a % of the total, but still growing with the market in areas like automotive, medical, industrial, and providing large opportunities for value-added products as well. Switching to NAND, it's a very similar picture. The numbers are coming from a higher high and saturating at about 35% on a go-forward basis. Although, really, in order to maintain that, we're going to need additional wafers into the industry going forward.
One of the dynamics that's different in the NAND market today than in the DRAM market is we have this additional effect, which is not only the slowing of technology migration, but for many of the competitors in the industry, we're starting to see a saturation in terms of the conversion of MLC to TLC. As that happens, it will take bit growth out of the marketplace. With Samsung running maybe roughly 70% TLC today, their ability to drive more bits just by shifting applications from MLC to TLC will diminish. That should slow the growth of supply in the marketplace. Similarly for SanDisk at 50%, et cetera. For Micron, we're further back on that curve. Micron today does have TLC components into the marketplace, so we have more headroom to grow that piece of our business on a go-forward basis.
Again, on the right-hand side, looking at the diversification, you see a continued shrinking of the consumer and removable section of the market, but very robust growth in both mobile and SSDs. The nice thing about the NAND market is the elasticity of this demand. As we think about some of the new technologies that are coming to market, we believe there will be the opportunity to grow the market as we grow supply, given some of the technology shifts that are coming to enable potentially lower price points. Given all of that, and what we've seen historically, which is steady growth on real terms of the memory industry, whether it's DRAM or NAND, we feel pretty good about continuing to invest through these cycles.
As we think about what does Micron need to do to participate in the memory market, we believe that there is continued growth ahead. We believe that the products we are providing to our customers will be able to add differentiated value. We believe we'll be able to put them into differentiated markets. Over time, generate a real return on the invested capital that we're bringing to the marketplace. We'll talk more about what some of those investments are as we go through the presentation today. One thing that's changed dramatically is how that supply is coming to the market and how it is meeting end market demand.
If you look now at, in the blue line, the growth of memory as a percent of total semiconductor market, you can see that, yes, it's cyclical also and that it's continued to grow through the years, but that it's grown for different reasons. If you go back 10 years into the mid-2000s, you can see that growth was being driven by a dramatic surge in wafer capacity. That was new entrants at the time, and a transition with incumbents bringing new 300 millimeter capacity in the marketplace, driving a rapid expansion in wafer supply in the memory industry. We have a very different dynamic today with consolidated suppliers. What happens when we have that dramatic increase in wafer supply is that we didn't get paid for the extra bits we were providing to the industry. You can see these two curves are actually anti-correlated.
If you flash forward now to what's going on in 2014, 2015, since we've had the significant consolidation we've had, is you don't see that growth anymore. The result is, in fact, you see it shrinking as new technologies take net wafers out of the market on a relative basis. What you see there is continued outperformance of memory relative to the semiconductor industry as we get paid for the incremental bits we're providing. Overall, we believe that this trend is more likely to continue rather than less likely to continue and gives us additional confidence in the investments we need to make. What are we focused on for 2015 and 2016? It's pretty simple. It's about investment in technology to drive new products and manufacturing efficiency.
It's about continuing to invest in our capability to deliver advanced products, new technologies, and system-level functionality to drive value-added solutions. It's about growing our customer relationships and our partnerships. If you think about Micron today, we are really the most global memory manufacturer by far. We have great relationships with customers all over the world because we don't compete with our customers when we're there with technical resources close to them. We really have the ability to drive differentiated solutions and deeper relationships to drive these value-added solutions into the future. In 2015, 2016, it's about execution. We're going to execute on our 20nm DRAM ramp. You'll hear more from Scott and Mark about that. We've got 16nm TLC NAND in the market today.
On a component level, we need to ramp our solutions, our TLC SSDs based on that technology over the next couple of quarters and beyond. We need to make sure we execute on what's a really exciting new technology and differentiate it from our competitors with our floating-gate 3D NAND solutions. We need to continue to create these new technologies like the 3D XPoint that we introduced a couple of weeks ago, Scott will talk about some of the new ones we have in the pipeline that we believe will also position Micron very well for the future. We need to continue to grow our system-level capability, and you'll hear more about that from Mark Adams. Finally, this piece about investing in our customer relationships and our partnerships is very, very real. We have great partnerships. Many of you know about some of them.
Some of them are not so visible yet to you. These partnerships, I think, are critical to growing in the complicated markets and in the markets that require the customer intimacy, we believe, for the future. I'm going to just stop right here, turn it over to Mark Adams, but just leave you with the thought process that notwithstanding the challenging times right now, we have a very strong plan. We have a lot of room to grow and improve, and we're very focused on doing that. You're going to hear a lot more about that over the next couple of hours, and then I'll come back to take your questions. Let me turn it over to Mark Adams.
Thanks, Mark. We probably should have done this up front, but for those of you in the room, we're very apologetic for the setting that we're in today. Sorry about the views and the windows, and hopefully not too much of a distraction. As Mark talked about, it's a very exciting time for our company when we think about the long-term opportunities. I wanted to start with a little bit of a setup for how we see the market today. No shock to anybody that we're dealing with some softness, obviously, in the PC market. That was a catalyst for some of the turbulence in the DRAM business. When we think about broader market conditions, we think there are some really good pockets of end demand, and there's this PC-driven dynamic in primarily DRAM today. We're watching, obviously, the broader markets.
It's also an important point to know that other markets are doing quite well. Mobile, networking, automotive, the server market, still doing pretty well, and that's an important point to distinguish today as we think about our business. On the NAND side, there's still tremendous growth. We've said all along that NAND, over the long term, because of this growth opportunity, probably is a little bit more turbulent. Notwithstanding that, it's exciting growth. We think we've demonstrated that already with the success we've had in mobile NAND, as well as our growth in SSD. Yes, we've seen softness in PC, and it's certainly had an impact on PC pricing. The message is that there are other aspects of our business that are doing quite well. I talked a little bit with some of you last night to think about this.
We all feel at Micron a sense of urgency in a cycle. You saw Mark's slide, in a cycle. While I can't share with you anything about current quarter results or any of that stuff, just go back 90 days to where we reported margins that you saw. Well, if that's the new low, if that's the bottom of the trough, that's the new memory business. You say, "Okay, great, Mark. Why?" I would suggest to you that the diversification aspect that Mark talked about in terms of end markets is the really primary driver for why the memory business is different today and will be more different in the future. I'd like to talk to that a little bit.
If you went back to this meeting five years ago, we may have shown a slide that showed an aspirational state of where we wanted end market demand to be. We may have shown you that all these markets that are now material today were going to be five years ago. Well, matter of fact, that's our business today. Networking is a big part of our business. Automotive is a big part of our business. Mobile is a huge part of our business. Servers, data center, cloud, huge part of our business. We're no longer a PC memory, PC DRAM company. This diversification on the DRAM side is mirrored on the NAND side. Even though NAND is a relatively newer technology segment, the products in NAND continue to diversify as well. 10 years ago, five years ago, it was mostly MP3, USB, photography.
Today, solid state, automotive, mobile, tremendous growth and continued diversification. Here's the interesting thing. We forget. We talk about DRAM, we say, "Well, DRAM's only going to grow 25%." I'm not an economics expert, 25% growth is 25% growth. I imagine there's some people in the PC business who would think 25% growth pretty freaking good. As I think about our business, meet DRAM and NAND, and the growth will continue to diversify, that is why we're so bullish on a long-term opportunity for Micron. What I thought I would do, because this chart is nice and it starts the discussion, what I thought I would do is I'd pick a couple, few cases of how this shift has happened at Micron and how it's impacted our business.
By the way, to go back and substantiate why margins that we reported 90 days ago, I think, somewhere in the 30% range. Why it's different. Let me show you why it's different rather than try to convince you with one slide. I don't know if there's been a technology that's shifted our lifestyles more as consumers than mobile phones. Okay. Let's just start with the fact that I get less kid time because my kids are walking around with these devices playing games basically 24 hours a day. Then think about all the video, all the photos. I came in and people were taking pictures of the Golden Gate Bridge this morning, this beautiful view. Then what do they do to them. They upload these pictures through social media.
Remote access to other applications, being able to check your home security system from a device like that. How we communicate. I know you all travel, right. Going back home, communicating over applications that allow you to video chat with somebody. The level of applications that's driving these devices to be so critical in our everyday life, go back in terms of product development and our engagement with our customers. None of that happens without memory. It's not a commodity memory that they just go buy on the street. In mobile, our mix and our portfolio is a huge asset. NAND, DRAM, how we integrate those products into packaged solutions, different power consumptions, different performance capabilities, all working with products that are designed for one customer, one customer's requirements, one customer needs.
As we think about the mobile case study, for us, we're not just selling a commodity product to a group of customers. Each one of our customers is defining how Micron and how we can work together and collaborate to provide unique solutions to them so that they can direct their products to a certain customer set. This segmentation and this product development initiative is why our mobile business unit is the top-performing business unit at Micron. We couldn't say that three years ago. As a matter of fact, it was our worst-performing business unit three years ago. Today, our mobile business, because of this type of engagement model, is providing tremendous opportunity. Again, portfolio, mix, capabilities beyond silicon and firmware controllers, packaging provide us an amazing opportunity. Here's the thing with mobile. We're just starting.
These devices will become more and more integrated into how we live our lives on a daily basis. Another example, probably one that's more changing in our lives today. I hesitate to ask the question, I won't, but I took the Tesla plunge. I live here in the Bay, it's hard to miss these things around. As a matter of fact, when I pulled up downstairs yesterday, they said, "Would you like us to charge your car while you're at your meeting?" Of course, I used my smartphone to make sure they charged my car, I don't know if they're going to bill me. If they're going to bill me, hopefully Juergen will approve the expense statement, that's another story. You think about the automobile experience. Okay, first thing, my kid gets in and goes, "Holy. Look at that screen.
The display on this thing is bigger than any PC I've ever had in my life. Then the applications set in front of you. First of all, there's diagnostics. There's also energy consumption models. There's all this information. There's navigation. There's media. Of course, in this specific model, okay, I can go surf the Internet. Shouldn't be doing that while I'm driving, but the option's there. They're smart enough to turn off the video function. The automotive experience today is incredible, and that's way on the early curve. Companies like Tesla and other companies are innovating in self-driving cars and software that you're downloading to let you know that if you've changed I've never done this, but if you don't put your blinker on, it'll come back and tell you to get back over in the other lane.
The information and the entertainment value of what's going on inside of automotive, the safety, all the sensors, it's incredible. None of it happens without memory. You can't do it. Same thing for us. We have field labs. Field labs. Remember the DRAM company from 15 years ago? Field labs to add customers that define products, that work with customers, not on, "Hey, ship me next quarter. Five years down the road, I'd like this function. How can you help me get there?" That's powerful. That's the shift in our business. When we think about the portfolio there, it's amazing. Again, products for one customer that we define together, integrated. Automotive SSD, think about that. Packaged material. A NAND and a DRAM package with a specific controller with firmware customized for that customer. Again, a very unique opportunity for us, and we do this well.
It's one of the reasons why we're the number 1 supplier of memory products to the automotive industry. Perhaps one that's a little bit less consumer-obvious, but I would argue is the most radical in terms of change right now, is the value for us in terms of cloud in the enterprise. Surely, the phone's cool, the cars are cool, and that's really fun, and we do that for our lives, and we transport ourselves, and we talk, and we game, and all that stuff, but the infrastructure to make that happen is probably in a more dynamic change than in the prior 2. First, when you think about compute, the capabilities today for memory to play a role in changing how companies leverage their compute architecture is unprecedented in terms of computing.
You think about the opportunity for us to react to things like big data, in-memory database applications, pattern recognition, to name a few opportunities for us in this segment in terms of computing. We now are designing new architectures, and we're designing new products that, again, are not just silicon, but architecturally changing the way people do compute and use compute power. When you think about even a product we're shipping today, Hybrid Memory Cube, where we've redefined how close we can get to the processor, and the packaging and the overall architecture provide the throughput that memory can help make compute more efficient and more powerful. That's a powerful story for us.
Going forward, beyond our leverage and our strength in this business, innovation around packaging, innovation around software and architectural improvements to get us closer and drive performance and overall throughput in working with the CPU and the processing and power is a critical opportunity for us to leverage because we think that's going to change the way computing's done in the future. Beyond that, when you think about storage, you've kind of got two issues going here. You've got people just massively storing everything. I talked earlier about our behaviors, right? Uploading all these pictures through social media. You go to Google or you go to YouTube, I'm just amazed. You just look up anything, there's a video on it stored somewhere. This massive amount of storage bit increase in our industry, it's calling for new ways, new methodologies to store things.
The backdrop of all this is the advancement in solid-state technology, that is, I'll stop short of saying fully replacing mechanical, but certainly making a dramatic impact on how people are thinking about storage architectures for the future. The growth of the solid-state market, both in client and enterprise, while early, is just staggering. Not only are there going to be more bits to be stored, there are going to be more bits to be stored on our technology. You can't do it. You can't do these changes. You can't change computer architectures. You can't change storage capabilities without memory. Of course, a couple of weeks ago, we have announced a revolutionary product in 3D XPoint.
Beyond our core technology today, we think 3D XPoint can be a game changer, not just in today's applications at the high end of our spectrum of what we market to, but we think it actually redefines what's possible. We think for things like in medical, the speed and the performance and the capabilities of 3D XPoint will enable us to contribute to how people are able to diagnose certain diseases, how quickly we can get people cures. When you think about how we're looking at our business, we're actually defining our products for the end markets. I think that's a big shift that you're going to continue to see in the memory business, that it's not just one size fits all in memory.
We're defining technology for the long term that solves these problems that people are having today, and we're excited about that opportunity, the role that Micron can play in that. You can see the big shift in our business. The business is expanding, and it's also expanding to a different model, not just the scale of it, but it's expanding to a different model. The legacy memory business that we've participated in for years, kind of this demand fulfillment model, it's becoming less and less a part of our business. What we're doing now is investing in resources and capabilities to develop value-added products with a whole host of different customers, major OEMs, hyperscale customers, different channel partners. Mark referenced earlier, strategic partners. How can we do things differently? How can we innovate together? We're having conversations with the end markets.
You could imagine government and security being an issue. I just gave an example about medical. Gaming. What's for consumer entertainment? The whole host of things that we look at, the conversation is starting with the end market opportunity, what do we need to do? How can we innovate? How can we work together? What are the type of products you need to create and solve your requirements? That's a different discussion than 10 years ago. Let's shrink the die, and let's sell it to PC guys and put it in a hub and hope they call. That's a different model. We're investing in this business to provide these value-add solutions, and by the way, it's not just chip level. It's much more than that. It's a complex business that's beyond just silicon.
It's how we stack products, how we package products, the firmware, the software that we add. It's all different because of different requirements. In phones, power is a big issue. In automotive, power is a big issue, but industrial temp. They have to operate in certain weather conditions, just have to be done. Networking provides another type of opportunity. This is a different business, a fundamentally different business, and that's how we're reacting to it, and that's how we think about investing in it. One of the ways we measure things and how we're doing on this curve, and we'll continue to measure this way, is how are we doing relative to our competition in utilizing our capacity, our capabilities in placing our product in these high-value segments?
It's nice to put these slides together and talk about tomorrow, what have you, but we need to be making sure that we're getting somewhere on this path, and we are. One of the ways we measure this is how much are people paying us for a broad portfolio of products, and how good are we at putting these products in the end markets that are attractive to us? We're pretty confident that we're leading the pack here. We get more for our products by putting them in these high-value-add applications, both in NAND and DRAM. We've talked about it before. Automotive is a very big, strong segment for us. We do very well in networking. We do very well in servers. Even in mobile, the continued growth and leadership there.
When you think about mobile, for example, if you take out the fact that one of our competitors consumes a lot internally, we're the largest provider of mobile memory solutions. The innovation and the capability to put our product into these higher value segments is our focus. We do sell commodity products still. I'm not trying to suggest otherwise. The mix of what we're doing as a company and our mission at Micron is to take that capacity and put it in the best areas, best high-return areas, the areas that actually we develop core properties is in to drive value. Mark talked about investments and the message that we are going to continue to invest in the cycle because we believe so strongly in our future and the long-term success of the business.
This is really, when you think about it, this is really the first time that Micron could get up here and say, we have the financial wherewithal to do so, and we're going to do so. In a capital-intensive business, returns don't happen overnight. That's just the nature. They're big-ticket items, and there's a process that happens, right? There's research and development being done, and that research and development then gets handed over to manufacturing, and equipment's put in the fabs. Of course, you're taking out older equipment at times. You're qualifying the product for your own internal qualification. You've got to bring customer qualifications in the picture. It takes some time. That's the nature of the business. We talked, actually, Mark talked a year ago at our event, and he said 2015, the slide was, is an execution year.
Of course, every year is an execution, but his point was that we have a lot going on and a lot of investments to make to position us for long-term success. The fact of the matter is, we have line of sight on these investments. The point of this chart is, yeah, 20 nanometer has been a big focus for us, and we made a lot of progress, and Scott's going to get up and talk a little bit about that progress. Even product mix, low power, DDR3 to DDR4, it has an impact on our ability to grow as we transition to the end state. On the NAND side, obviously, we've been very focused on our 3D NAND development and production. Beyond that, emerging technologies.
We just announced a pretty significant revolutionary technology just two or three weeks ago, and that's stuff we're doing in the background of everyday execution. With all that investment, our message to you is we need to invest. This opportunity doesn't come for free, and we're going to invest through the cycle to position us for continued long-term success. We want you to know that we're starting to see it. We're starting to see it in our business, and we have line of sight in 2016 that you're going to see more of it because of the investments we're making in 2014, 2015. That's the nature of the business we're in. What does that mean specifically? NAND, and DRAM. These new technologies show up. We have capacity to do things.
The mix of products and how the memory business is going to behave, something like 3D XPoint will be a material part of our business. In two years, that's going to be a major segment for us in terms of how we use our capacity. As we think about our business, as I said to you earlier, I believe that's the new memory business, memory for specific applications that helps drive differentiation and innovation. It sounds on where we are from a bit growth standpoint, because I know that there's some of the traditionalists in the room who would like to know, okay, sure, what does that mean to us?
The bottom line is, with all the shifts in technology, all the mix issues going on and what have you, 2015 was a year that our cycle lined up to a little bit lower than the industry in terms of DRAM. We expect to be back in line and grow at least with the industry in 2016, second half of 2016, everything in our fiscal year. We expect to be back in line because we see that. Those investments I've mentioned are going to be putting us in a position to be back there. In addition to tremendous growth in NAND, we fundamentally believe in the NAND business. We know it's going to be more turbulent than DRAM, we're in this for the long haul. We're investing for the long haul. What I'd like to leave you with today is a few things.
We believe in this business. There are cycles, as I mentioned earlier. We're not calling the bottom of a cycle. That's not our objective today. It feels like a cycle that we're in. We're in a cycle that in 2008, if you had told me, hey, in the range of 30% gross margin in your business, you're going to be in the middle of a cycle, I would have taken it then. That's the new memory business. We think we're in great position to go forward. We're going to continue to invest and pursue high-value segments because I think we've developed a core competency, and we've demonstrated that in how we measure it to deliver on that and to add value in our products.
Of course, we're going to manage our business on a daily basis to make sure we're making the right short-term decisions, there's risk and rewards, there's ROIs and what have you. Of course, we do that. That's our job. We can't lose sight of this great opportunity for us and Micron to play a role in the changing transformation of this business. We're going to invest to grow. As I mentioned, capital-intensive businesses, the returns don't happen overnight. We've invested a lot in 2015. There are investments we need to continue to make to take advantage of this market opportunity, we're starting to see some of that show up, you'll see throughout our fiscal year 2016, you'll start to see more benefits of that investment and as we grow. With that all, we expect to drive relative operating profit expansion given market conditions.
What I mean by that is we think we're going to do better than our competitors, given our plan, that's our aspiration to grow. I'd like to stop now and ask Scott DeBoer to come up. Scott's going to give you a little more detail on the technology direction and where we are on some of the recent efforts.
Thanks, Mark. Let's see if this works. Oops. Today, I'm going to spend most of my time focused on our core technology, really the current status, some of the next generation pieces in terms of our technical solutions to support our major customer requirements across the memory space. When we look at high-performance memory, our DRAM scaling, I've talked about here many times before, headwinds in terms of process complexity, in terms of physical scaling limitations, and those are all true. At the same time, we have line of sight to three technology nodes we're working on right now in R&D. I think I'll go through in some detail our 20nm status right now. We're really pleased with where that technology is. It's on plan. Team's executing well, really, I think, starting to hit on all cylinders coming out of the post-acquisition time.
We're bullish on what you're going to see out of us in the DRAM technology area over the next several years. On the NAND scaling side, we see a fundamental path to scale vertical NAND technology for the next decade in manufacturing. We continue to view NAND technology as the lowest cost per bit long-term option for solid-state storage. We are very confident in our technology execution path and our manufacturing ramp, and I'll go through that in quite a bit of detail today. In addition to our main core product technology areas right now, both Marks referenced the 3D XPoint technology that we announced in San Francisco a couple of weeks ago.
We have some other ones behind that, and I'll talk a little bit about innovation at Micron in general and why we have confidence to bring something as new and revolutionary as that along. Starting off with our DRAM technology. Fundamentally, we're in the middle of our 20nm ramp right now. The technology is qualified and ramping in both Hiroshima and in Inotera right now. The technology status in both of those two facilities is actually relatively the same right now. Good progress in both those fabs and really executing to plan right now. I've got a couple of graphs on this particular slide to give you an indication of where our 20nm technology plan is right now.
At the top, showing that we expect our bit output on 20nm and below technology to be more than half of our bits produced really by the end of this calendar year, and you can see the plot by fiscal year on this slide here. We're on track to do that right now. Now, how the two fabs do that transition is somewhat product dependent, depending on which products we're displacing. The net of all of our bits produced will be greater than 50% by the end of the calendar year. On the bottom graph on this slide is just a representation of where this technology is right now relative to our yield ramp through the similar phase at our 25nm node. I think this is representative of a couple things.
Post the merger of the technology teams, we spent quite a bit of time understanding what we needed to do in terms of manufacturing capability, recognizing some of those headwinds in terms of process complexity and the difficulty in ramping technology that faces everyone in the DRAM industry. We put a lot of effort into understanding how we're going to continue to improve in terms of early technology introduction and ramp capability. I think the data right now for our 20nm relative to 25nm is a representation of progress from this DRAM technology team over the past year. Our 1x technology node is also now under development in Hiroshima and making good progress there. We started that technology node out in Boise and now have the collective team working on it in Hiroshima and making good progress.
Our 1x nanometer technology really is going to be starting a ramp in the second half of calendar 2016. It's really a volume story for us in 2017. One of the things I wanted to also talk about today relative to our DRAM technology efforts are, as many of you know, we have a collective team of people we put together out of Elpida and Micron post-acquisition. I think that team is really coming on. We have a much stronger resource base than we've had in many years on our DRAM technology, and we've put a great deal of focus on that. The innovation we're starting to see coming out of that collective team is, I think, very impressive, and I think you'll continue to see a growing contribution in our DRAM business from the momentum that that team is picking up over time.
Moving over to the 3D NAND space. Just as a reminder, our leading high-density products on 3D NAND are a 32-layer, 256Gb MLC product and a 384Gb TLC product. This technology is still the highest density technology of any announced or currently in production technology, whether it's planar or 3D, in the NAND space, especially the 384Gb TLC product. This technology is, we think, positioned for the right entry point into 3D NAND technology into volume manufacturing. We think it comes in at a good cost point, and we're very pleased with where the progress is on this technology. Like DRAM, and maybe even more so, from the grounds up on this technology, we recognized it as extremely difficult and potentially disruptive technology that is probably the biggest shift in the memory business in decades in how memory is actually constructed, fundamentally different.
Recognizing those challenges, I think we put extra focus into really understanding where the technology needed to be from a manufacturing point, as we put it in position to ramp. I think we've had some actually pretty amazing results in terms of early success on our ramp capability for 3D NAND. A couple of years ago, if I would've projected where I thought we would be relative to our planar NAND performance as we're starting the technology ramp, we have really exceeded any expectations relative to how this technology has come on. I think it's a combination of that early diligence in the development process, and some of the technical choices we made relative to our 3D NAND technology, that have enabled our yield ramp actually on 3D NAND technology to be ahead of what we've historically done on our planar nodes.
We're very proud of that right now. We're in the process of ramping up TLC product for mobile, for storage, and I think have a very solid path to continue a density leadership position, cost leadership position relative to our 3D NAND technology as it ramps up. As a result of the early success on our 3D NAND, we're actually now looking at pulling in our prior expectation that we had discussed. Previously, we said we were going to have our bits on 3D NAND exceed our bits on planar NAND in calendar year 2017. We're actually now projecting that it will be before the end of calendar year 2016 that we'll be able to have more bits produced on 3D NAND than on planar NAND. Overall, very positive feeling, great execution by the team on this technology.
The next big thing for us on 3D NAND are really around continuing to execute this into maturity in manufacturing in terms of yield and ramp levels. Moving on, we have good results on our next generation right now that's in development phase, and is partially running in our Singapore manufacturing fab right now also. Really getting volume up on our 3D NAND technology. The picture on this particular slide, showing really they're working day and night right now to get this manufacturing fab to ramp 3D NAND up in Singapore. The overall team is very focused on getting first wafers out by the summer of 2016. A lot of effort there. One thing I wanted to do today is talk a little bit more, we haven't shared much detail on our technology, intentionally, since we think it is a competitive advantage.
Wanted to talk a little bit more about some of the basics of Micron's technology relative to at least the currently announced technology from our competitors, and talk a little bit about why we did this. The main reason is, at a fundamental level, we think that our technology is going to allow this rapid pace in terms of ramping the technology maturity and provide some advantage there.
When we started out looking, and I promise I won't get too technical here, when we started out with this technology, one of the basic concepts for us was take the fundamental cell technology out of the critical path by picking a technology that allowed us to fall back on decades of reliability improvement and physical understanding, and pick that physical cell as the basis, and face the other challenges in terms of this new complicated memory architecture with a known base of a solid cell technology. It's kind of a core choice. There's other ways to do it, and those ways will have other challenges and other advantages. In the end, we think the combination of this cell technology and the manufacturing capability focus we put early in the development cycle are really going to be advantages to us as we go into this volume ramp.
We're feeling very good about this. Now I'm going to change base here a little bit and talk about innovation at Micron in general, and how that's leading to 3D XPoint, and some of the other new technology areas that we are thinking are very important going into the future. One of the things that I wanted to do today is really step back and say why do we have confidence that we can take something like 3D XPoint that is very revolutionary, and drive that into mainstream memory technology? What are some of the things that give us confidence that we can go execute on something that is so dramatic as that?
Looking back across innovation at Micron over multiple years, really, we've faced several different points in time where we've put major focus on programs that actually weren't obvious that they would work. I think when we started 3D XPoint, it was absolutely in that situation. Looking back over time, many of the innovations at Micron actually are in that same kind of category. Something like 6F squared architecture in DRAM, which ultimately, of course, became the mainstream cost reduction path for the DRAM business across our competitors. When we started working on that, it wasn't obvious, and it was a pretty major change for the DRAM industry.
Picking a couple others in that same category Pitch multiplication, which is a lithography patterning methodology that we brought in on our 50-nanometer NAND, really was how we got into the NAND business, was with that core technology and bringing it in first. The rest, of course, the NAND industry followed that, and it's also gone across the whole semiconductor industry now. Those kind of innovations, and execution on those, has given us confidence that when we face something like 3D XPoint or even vertical NAND, that we have the confidence to push it forward. If we look at our recent technology announcements, we talked about in March, our 3D NAND, and I gave you a summary on that already. Big technology announcement since the winter analyst meeting.
We go to the announcement from a couple of weeks ago on 3D XPoint, even bigger, game-changing technology, multiple years of work culminating in a technology that I think is very unique and also very difficult to duplicate. We've given very little technical detail related to 3D XPoint, so I wanted to expand on it a bit just so you have a little more of an idea about what this technology is and also why we think it is such a big breakthrough to have put it together. If you look at the basic pieces of 3D XPoint, it's the architecture, the integration, it's the memory cell itself, and it's the switch. Each of those three things have been major sources of industry and academic research over multiple years. If you take something like the memory cell itself, there is an endless possibility of elements.
This is one of the things we use internally sometimes to talk about how many different ways you can build an RRAM memory cell. It's almost every element on the periodic table that either a startup, a university researcher, or someone in the industry has worked on. We usually cross off only the ones that are radioactive or tend to float away before you can get a hold of them and stick them in a material set. But other than those, almost every element on the periodic table is capable of building an RRAM memory cell. It's a great job security for material scientists, and it's a source of almost endless startup companies and different things that can bring out just that RRAM cell. Similar kind of situation for the switch. Then when you look at the integration of this, there's many ways you can integrate this technology.
Even, I think yesterday or the day before, there was a seminar at the Flash Memory Summit talking about the cost structure of this is probably awful because you have to have multiple quadruple patterning levels or multiple EUV with double patterning. If you actually required that, this memory wouldn't be very interesting. That's agreed. But the novelty of this technology is in how you integrate it, in how you pick these materials, complicated materials, and put them together, all of them, along with an integration concept, and you make something work, actually work, in production that has these properties. XPoint memory is a decades-old idea. It's an aspiration of the industry for years. That's not the new idea. How to build something, putting unique materials together, and integrate it in a way that provides these properties at a cost-effective point, that's the newness to this.
We're very proud of this technology, obviously. When we look at some discussion of where does this technology go, we think it has actually a broad range of uses, and the different big business categories may make use of different aspects of the technology. So if we look at something like a mobile space, the things that are important there may specifically be, how do I get faster non-volatile memory? Or how do I make use in the mobile space of the fact that it is non-volatile and I don't have to refresh it, I don't have an active refresh in standby. In mobile, you take advantage of one subset of the properties of this new memory.
When you look at client, there are obvious, and Mark discussed some of these previously, but there's some obvious ones around putting a high-performance cache of some kind of memory in front of the cheapest memory you have possible in order to build a higher performance SSD that makes use of TLC or QLC NAND. On the enterprise side, there's an obvious application taking advantage of almost all of the properties of the memory, with the endurance, the speed, and the pure bandwidth. So I think there's some obvious applications across big spaces. As we mentioned in San Francisco a few weeks ago, one of the things we're even more excited about is now that this kind of memory has become available, where do the system people take it, and what are the new applications that this expands into?
Beyond this, we still have focus on multiple other new memory technologies. Obviously, 3D XPoint is a multi-year project. It is now the bar for memory technology, new memory technology, and we have to look further out in the future, and it'll be a ways now, before something actually comes along that has new properties that provide an additional advantage. Our focus, especially on new memory technology relative to other materials for resistive memories or spin torque, which is on this slide, are more around the two extremes. We're interested in developing memory for the lowest cost and something that can compete with NAND long term, which is complicated, and as I mentioned at the beginning, in the highest end, which is the highest memory performance possible, with the adder of non-volatility.
Our long-term research now is focused on the extremes, and we have a significant effort in the background still on that. Certainly, 3D XPoint is not the end, but it's a major milestone for us, and we're excited about the future options too. To finish up, just touch on current priorities for the next year and where things are, and restate a couple things. On DRAM, really excited about the positive momentum. 20 nanometers on track right now, and we're confident actually that you'll continue to see better and better things out of the technology organization working on DRAM on Micron over the next couple of years. On 3D NAND, ahead of schedule. Really pleased with the progress and pulling in the crossover on the technology between 3D and 2D to the end of calendar year 2016.
3D XPoint over the next year is really going to be about maturing the technology, working closely with customers, and looking at applications to ramp into in calendar year 2016 in an early stage. Finally, we didn't talk about it much today, but we're still very focused on package technology and how we drive differentiation, with 3D packaging, and custom mobile packaging, to add value to our different customer applications. Those are our big focuses for the next year, and below is the roadmap of the different nodes and new memory nodes. Thank you, and I'll turn it over to Ernie.
Thank you. Good morning, everyone. Thanks for-- Oh, sorry. Wrong direction. Thanks for joining us today. My talk is going to cover three basic areas. First, give you a quick recap of some of our activities in the capital management area, our dilution management, our areas of investment. Move into something that you're all keenly interested in, which is a discussion of how we view 2016 gross CapEx, but also how we look at CapEx on a normalized basis, because we think that some of the changes that everyone has talked about today are causing us to look at CapEx in a bit of a different way as we go forward.
Finally, also talk about how some of the changes to the business are going to impact the guidance that we provide you on an ongoing basis that we think will more fully describe the business and also, frankly, make things a little bit simpler for you. Diving into each of these areas, this is the capital allocation strategy of the company. No big surprise here. Our first priorities are always investing in the business, whether that be for the ongoing development of the core businesses of the company, manufacturing efficiency, technology advancement, but also for the development of value-added solutions, and ways of addressing some of these markets that Mark Adams spoke about earlier today, which require ever more custom sort of application solutions, and closeness with the customer.
Finally, and really importantly over the last couple of years, a renewed focus on a return of capital to shareholders, and managing the dilution associated with the company's capital structure. We'll talk specifically more about that here in the next couple of slides. This does a nice job. The time period here is 2011 through 2015 fiscal Q3, so a five-year period. The company's invested about $5.2 billion cumulatively in R&D activities and $11.6 or so in CapEx. All of these designed to continue the advancement of our core technologies, as well as developing those new solutions like 3D XPoint, things like SSDs, things like non-volatile Hybrid Memory Cubes. Those are all part of what that investment is buying for us and buying for you. We've also spent about $3.8 billion managing dilution and returning cash to shareholders.
$200 million of that is in direct share repurchase, and the other $3.6 billion has been our activities around the convert exchange that we've had. The growth of the company over this time has also really allowed us to maintain the strength and flexibility in our balance sheet, and I'd like to spend a minute on the next slide sort of demonstrating that to you conclusively. For purposes of clarity, this slide actually has fiscal Q4 of our 2015 year included at either the midpoint of our guidance or where we don't guide, some Street estimates. No doubt, through a combination of organic and acquisitive activity, the company's revenue has grown, it's doubled, in fact, so CAGR of about 16%. That really is sort of the start of the story.
What that revenue growth has enabled is a growth in cash from operations at about 50% higher than the CAGR of the revenue growth. You can see that cash from ops has increased over that same time period from about 28% of overall revenue up to about 32%. Also, it's enabled, that scale, us to reduce our overall CapEx intensity from about mid-30% of our overall revenue down into the 23% range. The most profound difference you see here is in free cash flow, going from a slight negative back in that 2011 timeframe to a positive north of $1 billion, as we think about 2015. Clearly, scale has benefits, and those benefits are multiplied as we sort of work our way down through the P&L. During this time, we also established this return of capital framework.
This is not new, it's exactly the same as you have seen in the past. We have established a targeted capital structure. I'll talk more about that in an upcoming slide. We have talked about our dilution management activities, and we have a capital return policy that we now review on an annualized basis and make adjustments for as the business requirements dictate that. I would call your attention to one specific number on this slide, which is our activities through 2015 at about 79% of our free cash flow. That is a Q3 number. We will be providing you an update on that here as we exit the presentation. To preempt a little bit, the number looks a lot more like our prior year number, maybe even a little bit higher. Dilution management, it's worth spending a minute on this topic.
There has been tremendous progress made in this area. Net net, this includes the benefit of the cap call, it's all in what we've done with respect to the convert retirements, the share repurchases, and the benefit of the cap calls. Over a pretty broad range of prices, we have essentially reduced the outstanding share count by about 141 million shares and relatively 12%. If you really go back, one of the things I'm always interested in doing is going back and looking at history. It's the lowest net share count of the company in a couple of years. It's really been effective in helping manage the dilution and ensuring that the company does its part in returning capital to shareholders. Quick summary of the past financial performance. Revenues expanded, scale helps. Scale matters a lot, in fact.
We are now serving a diversified market with segments and customers that are far beyond anything that we've imagined in the past. That has allowed us to strengthen our balance sheet. It has allowed us to make the necessary investments in R&D and capital, will allow us to do that on a going-forward basis, as we're going to talk about here in just a minute. Enabled us to make significant progress relative to capital return and managing the dilution associated with being a public company. This is exactly the same chart that you saw a little earlier today. Investing in this sort of productive capacity output, this bit growth, is what largely informs our capital plan for fiscal year 2016.
However, there are some other things in fiscal 2016 that are pretty unique, occur maybe once a decade or a little bit more or a little bit less, depending on things, and that is a fab expansion in Singapore that you saw on Scott's slide. We also plan to do a facility investment on the Boise campus to expand the R&D fab. In addition, you've heard us talk about now 3D XPoint for a bit of time here this morning. There's an investment in the CapEx plan for next year related to that unlike some of our normal technology progressions, has a bit of a longer gestation period. It's a little different than doing the next generation of NAND, the next generation of DRAM.
This is a brand-new technology, albeit based on some ideas that are maybe a lot older than that, but it takes a little bit longer gestation time. There are those unique elements combined with the company's, if you will, ongoing plan. As we look at that gives us a CapEx number for next year on a growth basis, somewhere in the mid to upper $5 billion range. I would ask you to bear with me, because just as there have been some changes in the business, there's some changes into how we think about total CapEx. We also will go into some more detail of that here in just a minute. This is a historical breakdown that you have seen before from us. Starting at the bottom, about 20%-30% of our overall spend will be on DRAM.
That is for 20 nanometer and beyond, ramping of that technology, pretty straightforward. The non-volatile memory category, and notice it's called non-volatile now and not just NAND. The reason for that is that we've actually seen a movement of that XPoint CapEx from that top category, which is where you saw it last year, into that non-volatile category. Things get born up there and grow into their adolescence in the middle of that category or down below. It's real tangible proof that things actually move down between the categories here. That's going to be somewhere between 40%-50% of our overall CapEx. Remember I said there were some extraordinary items associated with the facilities investments and associated with 3D XPoint. Roughly 25%-40% of that non-volatile category relates to investments in those areas.
The remainder of that, which would be somewhere in the range of 60%-75%, is for things like the advancement of 3D NAND, TLC, and subsequent sort of normal technology progress that we would expect on an ongoing basis. Finally, at the top in that technology and product enablement category, another 20%-30% of our overall CapEx. That is where you see the R&D facility in Boise as a part of that CapEx, and then some of the things that Scott talked about in terms of leading-edge activity that he and his team are working on, whether it be for emerging memory, controller capabilities, or other productization activities. I've now used words like net and normalized and how different 2016 looks compared to the past, and I'm going to spend some time walking through that with you.
Starting with our mid $5 to upper $5 billion range, we have an extraordinary level of investment by third parties in our CapEx profile for next year. We have somewhere between a half a billion to as much as almost a billion dollars coming to us directly in support of this CapEx activity. As we look at sort of a Micron level of CapEx, it's somewhere in the realm of that $5 billion range. As we think about these once a decade investments in things like facilities, and think about how to help you think about our CapEx on a going forward basis, we come up with this normalized CapEx level, which is somewhere in that $4 billion range. Make no mistake about it, next year is going to be higher.
It's going to be about $5 billion plus all those other contributions we expect. On an ongoing basis, we would not expect that level of spending to need to continue in order to support the ongoing ramp of our business. Hopefully that adds some clarity, gives you some perspective, because the world we're looking at on a going forward basis, frankly, is a little different than the way it's been in the past, given some of the partnerships and the increasingly diverse nature of the world in which we're living. That actually also has a big impact on how we guide.
I've been through my first earnings call prep, the company has historically provided a lot of data on a cost per bit basis, which was completely relevant back in the days when what we were doing was providing commodity bits, and our bits looked like someone else's bits and it was an easy compare. Increasingly, what we are providing when we're providing these average cost per bits is a broad average across a range of product portfolio decisions. It's a set of decisions about what short-term output we make versus building the longer-term strength of our business. It also reflects, increasingly, you have to factor into that average that we provide, the various partner arrangements that we have.
In essence, we boil all that complexity of our business down into a cost per bit metric, and the reality is, it's sort of like averaging the weight of an elephant and a mouse. I can give you that precise number because I can put both of them on a scale and tell you what the average is. At the end of the day, I'm not really describing either one of those things. As we think about how to best inform you about our business on a going forward basis, we're going to move to a much more simplified and streamlined set of guidance. We expect on a going forward basis to give you total company non-GAAP revenue, gross margin, operating expense, operating income, and EPS. Now, I would tell you that that's a guidance statement.
As we describe our business on our quarterly call, so as we're describing the quarter just completed, we're obviously going to put more color around that. We're going to give you significantly the same information that we provide today in a historical context, because we do think that's helpful in letting you know how we're making progress in our business and what's going on. On a going forward basis, rather than going through this rather torturous exercise of taking all of that complexity and boiling it down into some cost per bit metric for a guidance basis, we're just going to give you at the end of the day what you're interested in understanding from us, which is what are we going to deliver in terms of earnings, and what are we delivering in terms of overall revenue growth as well as the margin profile of the company.
I told you a little earlier, we're going to come back and talk about our capital management framework. Here it is. Again, there's no big changes there. Still executing in the context of that. We have a hurdle for our return on investment to be greater than our cost of capital. Currently, our cost of capital is right around 10%. We continue to target a cash balance that's about 12 months, essentially, of our operating expenses, plus the short-term debt maturities and interest. We're continuing to operate through this leverage ratio of GAAP debt to EBITDA of less than one and a half, and CapEx to sales somewhere in the 20%-25% range. No changes there. We will continue to update that as it's appropriate. Key takeaways. Driving long-term returns with the ability to allocate and focus our capital decisions appropriately.
We're going to maintain a strong balance sheet. We're a growing business with lots of opportunity, our job as finance folks, among other things, is to make sure the company has the wherewithal to support and to deliver that growth. We are aligning our communications to be reflective and appropriate of the company that we are today, which is increasingly diverse, increasingly customized in terms of the products and solutions that we're delivering to our customers. We'll make sure that we both inform you of what we're doing, but at the same time try to be as simple as possible in helping you predict our financial performance. As I mentioned earlier, we'd provide you a little bit of an update to our activity so far in fiscal quarter four.
We've actually spent about $630 million so far in fiscal quarter four with respect to dilution management and capital return activities. We'll provide a further update on that obviously at the quarter end, but in case we're keeping score, that takes that 79% from the prior slide and moves it a little north of 100%, and we'll get the final tally here when we do our upcoming call. With that, I'm going to turn it back over to Mark for his closing comments.
All right. Thanks, Ernie. I'm going to wrap up very quickly here and then invite everyone back up to take questions for a little while. Obviously, in summary, we like this business. We think it's a much better business than it used to be. We think it's worth investing in. We like the cards we're playing with. We like our technology position, we like our global nature and the customer and partner relationships we have, we like some of the new products we're bringing to market. We feel good about where the company sits, notwithstanding some of the short-term market challenges that we're encountering. We're going to continue to focus on diversification, we're going to continue to focus on customer relationships, we're going to continue to deliver innovative new products to the marketplace. We've got some clear execution objectives for the next 18 months.
They revolve around the 20 nanometer deployment. As we mentioned, by fiscal Q3, you'll see a majority of our bits at 20 nanometer or below. We've got some good technologies coming behind that. You should see an acceleration relative to what we talked about previously, relative to our 3D NAND. You will see, as Ernie said, the realization of some of the technology developments that we've engaged in over the last couple of years, starting to bring those new products into the marketplace in 2016. We continue to generate strong operating cash flow. We expect to make good progress on our manufacturing efficiency and continue to close the gap relative to our deployed technology in the marketplace. We think over time, all of that makes Micron a great place to invest over the long haul, and we hope you continue to do that.
Let me stop and invite everybody up, Mark, Scott, Ernie, maybe we're going to pass mics for questions. There's one question I got in the back of the room that I think I would want to just address first, because I know everyone in the room is probably interested in. It has to do with, what are some of the cost headwinds that Micron has encountered? Bring a little bit more clarity to the cost headwinds we've encountered here over the last six months, in terms of our relative performance on a cost per bit basis. I think many people expected as we moved through the year, we would generate larger cost reductions than we've actually done. There's a lot of complexity to our business, as Ernie just said. There are a lot of moving pieces around product mix and technology mix, et cetera.
Let me start with the Inotera relationship. I think that's a significant one that maybe people haven't fully digested. When we look at our DRAM product portfolio, about 40% of that's coming from Inotera. Historically, and through the rest of this calendar year, we purchase that product from Inotera to market minus, and that minus is EBITDA-dependent. As the market gets weaker, the discount to market for Micron decreases, and then we sell that product out into the marketplace at a later point in time. There's a headwind there associated with the Inotera supply on 40% of our output that is significant.
When you layer on top of that the mix adjustments we've made, these are adjustments we make because we believe that they're in the long-term best interest of the company, to be in these broader segments and to be in these segments that will, over time, deliver more value. As we move product from a high-margin, commoditized PC segment into the mobile segment, there's a headwind associated with that move, just by virtue of the fact that those ASPs were lower in that time frame. There's a similar headwind when you think about what happens as you move from DDR3 production to DDR4 production, in ensuring that you have the right sockets for future development. There's another headwind that you encounter as you move the mobile products from LPDDR3 into LPDDR4, because the die sizes are bigger and the yields are lower on these new technologies.
Those are some of the big components. What I would tell you is that as we look to the future, as Mark and Scott talked about, we have significant ability to grow our bit density in manufacturing. That over time, that's going to drive cost per bit reductions. Now, they won't be as big as historically you would expect from that amount of bit growth per wafer because we're coming from a low depreciation base. You will see it in a significant way as we move through 2016, and it will have profound ramifications on the overall profitability and cash flow of the company. Why don't you guys all come up here and let's just open it up to the room. In the back here.
Yeah, thanks a lot, and thanks for the presentations. Can you expand a little bit more about the new memory, the 3D XPoint, and the Intel relationship? How is that going to exactly play out? Also, is there a timeline there? Because you talk about it's a game-changing technology, but are you very dependent on Intel here? Are you driving the horse here? Any expansion on that.
Yeah. Let me say a few things, and then if Mark or Scott or Ernie want to jump in, that'll be great, too. Intel is a very good memory partner for Micron. As these guys talked about, anytime you bring something that's radically new to the marketplace and the sockets have to develop and the new architectures have to develop, deployment is not overnight in terms of market acceptance. Having a partner like Intel actually is very advantageous for something like this because they are key enablers of those new architectures. The relationship with Intel is strong. It will be strong into the future. We will collaborate with them in terms of enabling their supply of product in the marketplace, and we will compete with them in the market, much as we do in NAND.
We sell them NAND today out of our manufacturing facilities, they own some of the NAND, and some we sell to them on a supply agreement basis, based on long-term agreement. We will have a similar relationship with this 3D XPoint memory, but we will also be developing our own sockets for our own applications in areas that are maybe of less interest to Intel and in some cases, in competition with Intel. In terms of how quickly will this market grow and how quickly will it become significant, I think that is hard to know exactly today, but it could easily be, in the 2018 timeframe, could easily be of the same order of magnitude as our DRAM businesses in that timeframe. Maybe not the same size, maybe half the size in 2018, but it will be a significant additive revenue stream to Micron at the time.
Does anybody want to add anything to that?
Just as a follow-up to it. Since close to $1 billion of fiscal year 2016 in CapEx is going to come from the partner, how should we think about the NAND margin profile, especially the wafers coming out of Singapore, going to your Intel partnership? Because based on existing setup, you ship based on the cost.
Yeah. We sell them some product at cost because they own that capacity in the joint venture. Then we sell them some product based on commercially negotiated supply agreements. Those supply agreements are more like a long-term customer relationship. They're more of a longer-term forward sale of that product. We anticipate we will have those in place for our partner.
On an annual basis, isn't that margin accretive to your NAND business because more wafers coming out of Singapore? Should we assume that there won't be a change in margin profile for this specific customer?
The margin profile will depend on the market supply and demand. We've said many times, very hard for us to predict over a short timeframe, exactly what the NAND margin is going to be. We certainly wouldn't try and forecast that. We think over the long haul, it'll be very rewarding business for us. Relative to the product that we might sell to a partner on a long-term contract relationship, obviously, we're going to think hard about what is the trade-off between risk and return, and strike commercial relationships that make sense to us.
One follow-up for Ernie. Thanks for all the color on CapEx, but also there's a lot of confusion on the equipment portion. If you were to look at FY 2016 versus FY 2015, how should we think about the equipment portion of the CapEx? As a follow-up to that, how does the D&A change which would have an impact on margin?
If you look at the equipment portion, you take out some of the commentary I made about facilities, I think you'll end up getting to a point where you're saying, it's not that different, right? We're not going to give you a specific number around that, but certainly, we've quantified before, in general terms, the cost of some of these new facilities expansion, and we've also talked a little bit about how there is an inclusion for 3D XPoint there. In terms of depreciation amortization, obviously, it's going to trend a bit up over time. Depends specifically on the timing of some of these capital investments, which we're still continuing to work, and we'll likely provide more color on that here on our upcoming call.
Quick question on the 3D NAND flash. You're using floating gate. I think your competitors are using charge trap technology. Does that give you an advantage in terms of ramp timing? How does the performance compare specifically to writing speed relative to 3D NAND from the competition?
You want to take that one, Scott?
Sure. Relative to the timing of the ramp, I think back to where I was discussing previously, we think by basing our fundamental cell technology on a known entity, and taking that out of the critical path, we think that is a piece that helps us ramp faster, and we think that's playing out in terms of our yield performance so far. I think that's a net positive for us in terms of that. In terms of product performance, we think that largely, the technology is going to look similar in the end applications, and there's going to be some cases where maybe there's some advantage for floating gate, and some cases where there's some advantage for charge trap. The net of the whole thing is that the technology won't look that much different once it's in a solid-state drive.
Let me expand on that just a little bit. A lot of the performance characteristics are going to be dependent on component architecture decisions and system architecture decisions that we make as opposed to the underlying cell technology. Yes, you can imagine in some circumstances you might have a write speed decrement for a floating gate cell, when you get to the end customer, a lot of it's going to depend on how did you architect the chip, and how did you architect the system in terms of the performance that they will actually see from these devices? We strongly believe that we're going to see a lot less hiccups, and our customers will see a lot less hiccups as we ramp this product into volume manufacturing because of the much greater understanding of all the fundamental yield mechanisms and reliability effects that customers will encounter over time.
Thanks a lot. Actually, I had two. First of all, a question on DDR4. ASP, the DDR4, DDR3 gap right now, there's maybe a 15% premium currently for DDR4. Of course, your costs are higher than that's why that's a pinch right now on margins. Can you talk about the risk that the premium that DDR4 gets in the market goes away by the time that you're able to make up the cost delta between DDR3 and DDR4?
Premiums always go away over time because you get to the end of the life, and there's a scarcity. The pricing ends up being supply-and-demand driven. We would anticipate that over time, you may start to see a premium for DDR3, right? It's just a matter of which of these products are in oversupply and undersupply. You are correct that in the marketplace today, DDR4 is a headwind for us. We think that is very quickly behind us. We ramp to 20 nanometer node.
Okay, just one more. You talked about DRAM bit growth category being 20%-30% between FY 2015 and FY 2017. There's a slide that sort of suggested that it's going to be well above that in FY 2016, and then it's going to be well below that in FY 2017. Does that imply something like 35%-40% in FY 2016 and then something like maybe 10%-15% in FY 2017?
No, we tried to give you a curve that would directionally show you what that looks like. Actually, there's an acceleration as we move through time. Mark.
Hi, thanks for the indication. Maybe given the headwind we are seeing in the DRAM pricing, could you remind us how to think about the Inotera benefit, which rolls out calendar 2016?
Yeah. Okay. Obviously, it's a complicated set of parameters that drive exactly what that relationship looks like, because you have market minus moving to margin sharing. You have mixed changes going on, you have technology changes going on, all of those have an impact. Having said that, when we talked to you back in, I think it was February, we talked in terms of what would the change be when that new agreement goes into place and when it starts to ripple through our financials. We said, it's in the high single-digit kind of margin improvement on the Inotera output as that change happens. We still believe it'll be in the mid to high single-digit range, notwithstanding what's gone on with ASPs in the meantime, because obviously, we've had margin headwinds under the existing agreement in that timeframe.
The delta benefit, assuming that there's not some ongoing collapse in pricing, we still think that's a good placeholder for the delta benefit. If we see more price erosion than we're currently anticipating, it'll be a little bit less. If we see flat pricing, it'll be a little bit more.
Yeah. Just a follow-up, too, on the ROIC on the NAND business. You kind of talked about 40%-50% CapEx on NAND. The margins in NAND are still quite limited. If you look at ROIC, it's really low, right? When do you think Micron gets paid for all the investments you're doing in the NAND industry?
We don't know. Very hard to predict that quarter to quarter. I think there are indications that as we move through 2016, we may have very robust demand for NAND. Trying to time exactly what that is like relative to supply, as you know, is difficult, and I'm not going to take a stab at it today.
You can try and answer.
Fundamentally--
Yeah.
One comment, really. Just fundamentally, though, Monika, Mark's response notwithstanding, we still believe that the things we've been communicating to you are on track to drive competitive, relative better performance. Things like mobile, things like solid-state development and controllers, things like 3D, and things like TLC and planar. We think the combination, on a relative basis, notwithstanding hard-to-predict pricing and all that stuff, we're still very bullish on our performance and where we're going in NAND.
Thanks very much. David Longwell, Saga. Two things. First, 3D NAND. Floating gate versus charge trap. You explained the benefits of your choice. Can you give us some idea about why some of the other companies chose to go with charge trap? Are there some perceived benefits? Is there something easier about doing that?
Yeah. Scott can comment more. I think fundamentally, it's a little bit easier to build a charge trap down a long sidewall than it is a slightly more complicated structure that we have with floating gate. I think you may see over time others evolve as they gain more confidence in their ability to execute, or as they realize that there's an existence proof out there.
Okay, great. On the 3D XPoint, how many layers of metal will your initial process have? Can you give us some idea about why it starts out as more expensive than NAND? Is it because you've got more metal layers? Is it a bigger cell? Is it more expensive to deposit exotic materials?
Well, no. We're not going to tell you how many metal layers we have, we're not going to tell you how many tiers.
Although we might have told you how many tiers.
Yeah, we're running 2.
Two tiers.
We did. Okay.
Were you going to. Yeah, sorry. Go ahead.
Yeah, it will be much cheaper than NAND, without a doubt. When you start a new technology, and you have new equipment, and it's slightly underloaded, or it's underloaded as you're working through yield issues and commencing a ramp, new technology introductions are always a cost headwind. The technology, as currently defined, will be much cheaper than DRAM. Your question is more expensive than NAND? Yeah, sorry. Well, it's going to be more expensive than NAND for quite a while because 3D NAND is very cheap. Eventually, as 3D NAND scaling slows and as 3D XPoint continues to scale, or as we add incremental levels of cell per physical location, we can foresee that would cross over.
The only thing I would add to that is, all correct, the only other thing is NAND has the advantage of either MLC, TLC, or QLC, so number of bits per cell. Most our RAMs and our 3D XPoint don't start out with that capability. We're still looking at it long term, but there's some pretty fundamental challenges to different kinds of resistive RAM materials to actually making them have multiple bits per cell.
Great. Two questions. One, you answered most of the questions about the 3D XPoint, but long term, the number of layers you can add in 3D integration, is it comparable long term, you think, to flash, or is there another mechanism that limits it, like yield or perhaps temperature cycling, something like that?
It actually, in this case, comes down to the physical structure. NAND scaling is all about tiers going forward and how many vertical tiers you make. Those are 32, 48, 64, whatever number, big numbers of tiers, and the whole string is up. On 3D XPoint, it's fundamentally still a planar technology with another deck on top of it. There's actually two different scaling paths, which makes this, it doesn't compete with NAND in terms of that kind of tier scaling. You'll have 2 or 4 tiers or maybe more, but you can also shrink it like a DRAM. You actually have two fundamental scaling paths on 3D XPoint, and your opportunity to keep scaling down a DRAM-like path is, there's not a fundamental limit to that.
Thank you. The other question is on capacity and market share. In DRAM, looks like some of the big competitors are adding or at least maintaining wafer capacity. As you scale and you lose wafers, are you going to try to keep market share? Or is the idea to use the 3D XPoint to maybe eat into that market and maintain total market share, versus trying to maintain the percentage against the other two competitors?
Yeah, we're going to maintain flexibility on all of that stuff. Our decisions will, at the end of the day, they're going to be based on our long-term view of what the ROI sees, right? We don't need to be any particular market share point in any particular technology. We do need, as Ernie talked about, we do need in aggregate to have enough scale to be an efficient manufacturing technology developer and to be able to service customers that we care about in those markets.
You've talked in the last couple of these meetings about things you could do if supply overshot. You talked about could we take fab loading down a little bit? Could we build inventory up during those periods? I understand why you haven't done those things. You haven't had that much bit growth anyway. Can you talk about your thinking along those lines going forward, and are there still criteria that you would look at where you might run fab utilization lower than 100%, or where you might let inventory build up?
Well, a decision on inventory is somewhat easier than a decision on capacity coming offline. A decision on inventory is really based on what do you view the market looking forward over some relatively abbreviated time as you would move that back out into the market? A decision on removing capacity also depends on that forward-looking market view. It's a little bit more complicated because it also depends on what you think the competitor response to that is, and how quickly do you think competitors will realize what you've done and respond. From our perspective, we are totally comfortable not running our fabs flat out all the time.
At the same time, we want to maintain some reasonable market share, and we don't want to be unnecessarily burdening our forward-looking ability to deliver product to the customers, nor our cost structure to the benefit of our competitors. The decision to take capacity offline, I think, has to be more predicated on, is there something structurally wrong over the long haul with the supply and demand balance in the industry? Is that something that Micron should do something about?
You want to? Melissa?
We have a question here.
Right here.
Okay. Mike.
As we look at 2017, how should we think about the SSD portion of your mix in the NAND, going towards in terms you talked about value creation and all of them? How do you see the SSD portion evolving, and do you have all the holes filled in terms of your product portfolio to address the market?
You want to talk?
Sure, I'll take that. Needless to say, continues to be very bullish on our performance in SSDs. As you project out into 2017, with the byproducts of the portfolio and the technology, additional investments in controller development firmware. I hate to put too tight a range on it, but I think in somewhere in the 30%-40% of our capacity at least, we see it in the SSD business. Mind you that we categorize things like mobile differently and the embedded automotive type applications. SSDs will be a material part of our business and the capabilities that we're investing in is, again, similar to the CapEx discussion. The capabilities we're investing in terms of controller development, for example, whether it be like a PCIe type controller initiative.
Those investments that we're making in 2015 and 2016, we think that puts us in a real good competitive position and [valuable position] for 2017.
Let me do a quick follow-up. Not all SSDs are created equal, right? We have consumer SSDs, client SSDs, data center SSDs, enterprise SSDs. We will target our output wherever we think we have the best relative competitive advantage and can generate the best return.
Okay, just a quick follow-up. This is a question for Scott. Just wanted to understand on if, given the DRAM end of life, whenever it happens, do you think that the replacement could be a non-volatile memory? How ready is Micron if this transition happens, and can you give some flavor for it, and how ready is Micron with that, and what kind of cadence should we expect? We saw that with 3D NAND, people have been talking about it for five years.
Sure
the end of life for DRAM. Thank you.
Sure. As I mentioned, we actually have a longer line of sight right now on DRAM technology, probably even than we did a year ago at this time, through more focus on future nodes right now. I don't think the end of DRAM scaling is impending over the next five years. I think we're going to be strongly positioned, again, through our focus on new memory technology, when it comes time to have something ready for the end of DRAM scaling. We're putting significant effort into that. I think there'll be opportunities for us to talk more about that over the next couple of years. I do think that it's most likely that something that expands beyond DRAM winds up being a non-volatile memory. Primarily, because it brings so many other advantages along with it.
The majority of our focus is on non-volatility and high performance on something that can replace DRAM long term.
Thank you. Just two questions here. On the NAND side, as you exit fiscal 2016, what do you see the mix of 16 nanometer TLC and 3D NAND? The second question for Ernie. As you look at the guide, you talked about giving a gross margin and operating income guide. Do you think you'll guide for the full year, too, what the gross margin brackets should be or the OpEx?
That's an easy one. We don't plan on doing that at this time.
Do you want to take the-
Missed the first one. Sorry.
On the NAND side, how do you see the mix of 16 nanometer TLC and 3D NAND?
Well, somewhat for competitive reasons, we're not going to throw out a percent. Obviously, we've talked about our ability to now go to market with 16 nanometer TLC, you have to think, the conversation earlier about the broader decisions we're making on capacity. At each of the factories, we have capabilities, and we have end markets. For example, there are markets that are very attractive to us that are still on 25 nanometer NAND MLC. We don't want to, for competitive reasons, tell people our recipe. Needless to say that 16 nanometer TLC is starting from a relatively low base to now we're pretty excited to go to market in markets like channel SSDs, where it's pretty competitive.
We think what's interesting about that is, that gives us a little bit of margin uplift in that category, that segment, because we've been out fighting with MLC products up until today. Suffice to say that we're going to take every opportunity to drive that capacity where the returns make sense, but we're also looking at a broader mix of products and how that return aligns for those solutions we're selling into.
One other dynamic to keep in mind is that the 3D NAND, as it rolls out, a significant portion of that will be TLC. We will start at MLC, but very quickly that will transition.
Thank you.
There's clearly an impression out of Taiwan that Inotera is behind their curve as they've made their presentations. You're saying that you expect to have 50% of your bits out in Q4 on the 20 nanometer. Will that be primarily from-- I mean, the impression that's been left, maybe it's incorrect, but will that be primarily out of Hiroshima and the MMT facility, or how do you see that balance going?
Inotera is not behind where they expected to be. They are about on track. We would expect the crossover would occur in fiscal Q3 of 2016, in terms of Micron overall bits produced, including the effects of Inotera.
Okay. On the 20nm, as you talk about your scaling into 1x and 1Y, has your architecture, your materials choice for your cap structure, et cetera, do you have a path where you feel that you'll be able to close the gap between where you are now with Samsung?
We're definitely confident in the material path and architecture for the technology rolling out. As I mentioned, we've put a substantial amount of resources into DRAM relative to historical, the team's really starting to execute. I do believe we'll
Narrow the performance relative to the industry leaders. I think our DRAM story will continue to get better over the next couple of years as the R&D investment rolls into the manufacturing.
Okay. On your point about revisioning the way CapEx and the latency for a CapEx to revenue. If you look at some people may have viewed that picking up Elpida in the end of July of 2013, you got a free $4 billion CapEx, if you will, you put another $8 billion in subsequently. You have a $12 billion over a compressed two-year period on a revenue basis, roughly equivalent to where you were in August of 2013. Could you give us your sense of what kind of latency changes between where that CapEx would have been spent and where you would expect revenue dollars, leave out the margins and the market prices, but the revenue dollars to come from that CapEx investment?
I'm not sure I'm following all your underlying assumptions. In fiscal 2013, I think we spent about $2.9 billion, and in fiscal 2014, we're going to spend about $3.8 billion, $3.9 billion, something like that.
Yep.
The number, I think, is closer to $7 billion in aggregate since the Elpida acquisition closed. I don't believe that there's a sea change in the latency relative to capital deployment to fit out for established technologies. I think there is, obviously, a different return on investment building a new shell or investing in a radically new memory technology like 3D XPoint. I think maybe the underlying missing piece here is some of the headwinds associated with remix of tool sets in Elpida. When you bring in a mismatched tool set, you try and collapse a technology path, there's maybe a little latency associated with that. I'm trying to follow here. Let me leave it there. I had another point I was going to make.
Thank you. Over there.
Yeah. First question for Ernie. Your CapEx guide, you're pulling out your strategic partner investments. You're taking that out of the normalized CapEx. We also heard today that you have a lot of partnership activity that's not announced yet. You've got new memory technologies coming in 2017, perhaps beyond. Why wouldn't that component be more of a recurring CapEx versus discrete in 2016?
Which portion of this? Strategic partner investments or third party investments?
You're pulling out strategic partnerships from normalized CapEx. I'm just wondering how recurring that is given what's in the pipeline.
It's possible that it will recur on a year-on-year basis, and if so, we'll talk to you about that and let you know what it is. I don't think it changes the fundamental message, which is that on a normalized basis, we think Micron's business is in that $4 billion range. Next, we have some extraordinary items, some of which are funded by third parties.
Okay, great.
We have time for one more question.
Thanks for taking my question. Ernie, I had a question for you on sort of startup costs going into next year, specifically for 3D NAND in Singapore. Given the new tool set that's associated with 3D, and it's somewhat comparable to what we saw with 20nm DRAM. Wondering if you put some boundaries around it or some color on how we should look at that for next year.
Startup cost is a pretty broad category. I think the investments in general are, from a pure raw startup phase, somewhere in the range of low $100 million for a 1,000 wafer starts a week. We obviously have a shell facility that is more an expansion area as opposed to directly attributable to the startup cost. I'm happy to help you out a little bit more, but those are sort of the two boundary conditions I can give you right now.
To follow up, maybe in terms of margin impact or incremental D&A.
Well, as I said earlier, we're still in the process of figuring out the specific timing of all those investments for next year, and we'll be talking a little bit more about that and the margin impacts, potentially as we get on our call here coming up. All right. Let's just close by thanking you all once again. We're glad to share a lot of good information with you today. We're bullish on the future, and we're very focused on executing for our shareholders. Thanks for coming. Thanks.