Okay, thanks everybody for coming today. We are really glad to have such a great crowd for Micron's 2014 Winter Analyst Day. We are going to get things started right away here. Those of you on the webcast, we'll get the webcast started in one minute, too. We're going to kick things off with Mark Durcan first. He'll come up here in one minute. Just to let you know, as usual, we try to structure the presentation to address what we feel like are the key questions, key topics that we're getting from the investment community today. Hopefully we'll accomplish that. We will save plenty of time for Q&A. After each section, Mark Durcan will do an intro. He will do his Q&A at the end of the presentation, just so you're aware of that format. With that, I'll welcome Mark to the stage.
During the course of this meeting-
Can everybody hear me? Yeah.
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 achievements.
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, great. Now I get to talk.
During the course of this meeting, we may make projections or other.
All right, now I really get to talk. I guess not even the CEO gets to talk over the safe harbor language. Welcome, everyone. Thank you all for coming. I think we've got a good presentation here today, which as Ivan mentioned, should be responsive to most of the questions that are out there. Of course, our goal is to answer all the questions. As we go through today, you'll get an opportunity after each of the officers presents to talk to them, and then I'll come back up at the end, obviously, and try and clean up any residual questions or items you guys want to talk about. I'm just going to start off with a little bit of information about how I see the industry today and what Micron's focused on going forward.
We'll have Ron talk a little bit about how we're planning our financial strategy, Mark Adams on business and operations, Scott DeBoer on the technology position of the company and what we're focused on there, then some information from Brian Shirley and Mike Rayfield on our products and markets. Without further ado, let me just jump in. 2013, of course, was a very busy year and a very productive year. We accomplished a lot, and I don't want to take you back and run you through all of that, other than to say we did a lot of things that helped set the stage for where the industry is today.
We did a lot of things internally to help rationalize our business and streamline it so that as we move forward, we can really focus on what we need to do, which is memory systems, and our customers going forward. We also celebrated our 35th birthday. I've been here for about 30 years of that now. It'll be 30 years this summer. I've seen a lot of change as we've moved through time. A lot of things are different today. I think everyone wants to know why is it different today? What is it that is structurally changed in the industry that makes it so that we can all have confidence on a go-forward basis? We really are in an environment that is much better in terms of the types of returns shareholders in Micron can expect.
Let me start with, first of all, it's different. You can see that the structure is different. There are really only 4 manufacturers or 5 suppliers that supply 95% of the world's memory bits today. Of those 5 suppliers, 4 manufacturers, only 2 have the full spectrum of memory products that customers want, and only 3 supply DRAM at all. Certainly a consolidated industry that has a different dynamic. When you think about what is different about the memory business, this is really just one of really 3 legs of a stool, I think, however, that are making a difference.
One is that in an environment with a consolidated supply base, all with critical mass here, with enough scale to support their business on a go-forward basis, it appears that it is less likely that companies will, at least Micron, will be interested in adding capacity for pure margin share gain. Micron today, at 22%, 23% of the world memory market, has more scale than it's ever had in its history. We have adequate scale to fund all our technology development on a go-forward basis. What Micron here is focused about, or on, is making sure that we take the fruits of the business and reinvest them in a way that moving forward, we can continue to provide the solutions that the customers want. We don't need additional scale on a go-forward basis.
The second leg of the stool revolves around customers and what's going on from a customer perspective. There's been a significant diversification in end markets, and I'll come back and talk about that, and there's been a significant change in the types of customers that play in those end markets. That really leads to a more balanced relationship when we think about how memory is delivered to the market and how Micron can differentiate and diversify its products on a go-forward basis. The third leg of the stool is really all around what's happening with supply, and I'll talk a little bit about what's happening from a technology migration perspective and how that influences the supply and demand balance in the market moving forward.
First of all, the nice thing about the memory business is it's a large business, but it's still a rapidly growing business, and it's a business where there's lots of opportunity to innovate. When you look at this slide and the five sort of big trends that are driving the information of the world today, and you look at the growth that those drive, it's really a virtuous system that continues to build on itself. Starting at the left with networking, you've got a big system today that as it feeds machine-to-machine communications and those feed mobile networks, which feed the cloud, which feed big data, none of these segments go away as we ripple to the right and drive more and more memory usage into the system. They all just build on each other and drive memory usage going forward.
Note that memory then goes not just in the computing segment. It goes into a whole plethora of value-added segments where differentiation can perform specific functions to make the customer's end system more valuable. Whether we're talking about computing today, and the memory we supply to our historical customers and OEMs, or the ability to customize and differentiate those solutions for large data centers or for storage companies, or whether we're talking about the automotive segment, where car companies are building out new automotive networks to support advanced infotainment systems, navigation, crash avoidance, voice recognition, all the types of things that people want to do in their automobiles going forward, Micron can take its memory solutions and work in close partnership with those customers, as we've had a history of doing to drive differentiated solutions. This is very different than the memory industry used to be.
It's not a Wintel platform. There are multiple enablers across the whole spectrum of end applications. There are multiple operating systems across the whole spectrum of end applications. In each customer, in each segment, Micron can go in as the largest memory manufacturer to the non-captive memory supplier, bigger even than Samsung when you talk about the bits that Micron delivers to customers and doesn't consume internally. We can go in and craft those solutions in close cooperation with these customers. More and more of these customers on the right are looking to Micron to do exactly that.
The other thing to note on here is that one of those three big DRAM suppliers is also present, obviously, in the customer base, and in many cases, that presents unique opportunities for Micron then to go in and be the partner of choice, and capture the ability to innovate for the future that may not be open to all players. I said I'd talk a little bit about supply. This is a long-term trend on what's happening to supply of bits in both the DRAM space and the NAND space. A few things to note. First of all, the light blue lines are what we think the aggregate supply to the marketplace is likely to be today and over the next few years.
I probably should have gone back in time a little bit further so that you could recognize that bit supply in the DRAM business used to be 80% per annum, and bit supply in the NAND business just five years ago used to be 100% per annum. We're clearly seeing a long-term trend in terms of the reduction of bit supply to the market. In particular, what you're seeing is a stretching out in the blue piece of these bars, which is the bits supplied by virtue of technology moving forward. Most memory suppliers in the memory space whether it's NAND, NOR, DRAM, want to operate at the efficient frontier. They want to have their manufacturing capacity aligned so that they can drive the competitive manufacturing operation, and technology has been a big part of that.
It means that in order to have a low cost per bit, you have to leverage advanced technology to get more bits per square centimeter because you're building square centimeters. Historically, there's been a push in the industry to drive to that efficient operating frontier, and that has not changed. Micron today, just like I believe our competitors out in the marketplace, will probably continue to drive to that efficient operating frontier by deploying advanced technology. In fact, for Micron, that's one of the things we want to focus on over the next couple of years. The net result of doing that is now a smaller increase per annum in the bit supply to the marketplace. That's a different situation than we've seen historically where those bits supplied to the marketplace have an easy or a high probability of potentially outrunning the end market demand.
In an environment like this where we have large end market growth, large end market demand growth, and some of the other guys will come back later and talk to the drivers of that and what we think those numbers look like, it's very difficult for technology to keep pace with that. Suppliers in the marketplace, in order to outrun demand, have to go out and add new wafer capacity. There's a calculus or a decision to make there that is now separated from what's required to be at the efficient operating frontier. There's a separate calculus around what is best for my company relative to gross margin dollars and to shareholder return on investment.
That balance, I believe, and I think Ron will talk a little bit more to this here in a little while, will lead to an environment where gross margins can be very good on a go-forward basis and on a sustainable basis. The other thing this slowing trend enables is products like those on the right, like the Hybrid Memory Cube, which is a complicated memory system, high-density DRAM stacked with an ASIC that performs all sorts of functionality, as well as a high-speed interface to the CPU. Those types of products take longer to develop and to interface to customer end systems, but they can deliver value in an environment where they're not competing with a commoditized product next year that's 40% cheaper than the one the previous year.
That's why you're starting to see now an explosion in the applications that are adopting memory systems as opposed to pure commoditized memory components. We think we'll see a lot more of that moving into the future. Again, you'll hear more of that from Brian Shirley and Mike Rayfield. Enterprise SSDs is another example, whether it's enterprise or client, enterprise being more complicated systems. There is a growing set of end applications for NAND flash where true value is delivered, not just through the advanced technology in the NAND component itself, but in the NAND system that Micron can deliver to the customers. In summary, relative to the memory industry, three main changes in the marketplace today. Consolidated suppliers, each with sufficient scale on a go-forward basis.
Limited new wafer capacity entering the market by virtue of economic analysis that optimizes return for shareholders and slowing technology migrations. Diversifying end markets and customers that provide the opportunity to drive true differentiated value on a sustainable basis, all leading to the situation we find ourselves in today, and hopefully why all of you are here in such large numbers to hear about Micron's story going forward. What's Micron focused on? We're really focused on three things. One is driving internal operational performance. We've done a lot of work on that, and Mark Adams is going to talk in detail about some of the things we're doing in our manufacturing network, as well as some of the things we're doing to close the gap in terms of advanced technology deployment relative to some of our competitors. That's important, I think.
For a stable industry, Micron needs to be at the forefront from a technology deployment perspective, and that's where we intend to be. Operational excellence and deployment of advanced technology in our business continues to be a focus of the company, and we'll continue to make the investments we need, not only from a manufacturing technology deployment perspective, but also from a product diversification perspective. We need to be very focused on the future now, because the future for Micron is about delivering differentiated solutions to all those end markets and more and more building in system-level functionality to those products and delivering them in a truly value-added way that becomes more intimately married with our end customers. That's what Micron's doing.
You've read over the last six months about various incremental improvements we've made to our team, about new capabilities we're bringing in-house, we'll continue to do that both organically and inorganically in order to make sure Micron's positioned for the future and to deliver these more complicated, differentiated products that the customers are going to want and need. Finally, we're going to continue to do what you've seen us do here over the last six to nine months, which is manage our capital structure in a way that manages dilution while driving down debt over time. Ron will give you a little bit more detail about exactly how that looks and what our plans are for the future.
A long-term focus for Micron, and just think about this as you hear more detail about where the company's going, is really to maximize shareholder value and shareholder return, we put that at the forefront of every decision we make. Growth is a piece, differentiation is a piece, but at the end of the day, we're thinking about our shareholders and how we deliver for our shareholders by allocating our resources, both human capital and physical capital, to drive that outcome. I'll be back at the end of the day to answer any questions that we don't cover in sufficient detail, but hopefully these presentations, as we move through here, will give you a sense of the detail behind the big story. Let me turn it over to Ron. Thank you.
Good morning. Great to have you all here. I'm going to cover some questions that we got from you and provide you a little perspective here to set the stage for the speakers who are coming. What metrics should we use to evaluate Micron's performance as we go forward? What are you focused on in terms of your business and operations and optimizing them, and how do you think about capital allocation? I'll address that in sort of three buckets here. Micron performance, talk a little bit about our investment thinking and priorities. Mark mentioned the priorities, but how do we think about supporting that and our capital allocation strategy and what goes into our thinking there. First, to Micron performance.
Clearly, if you look at the semiconductor industry over the last 13 years or so, since 2000, this shows a graph of the semiconductor industry ROIC, excluding memory, which is the dotted line above, and the memory industry over that period average, the bottom black dotted line. Clearly, the memory industry's underperformed its semiconductor peers. You can see in the last year, that's substantially changed. As Mark mentioned, we believe it's changed fundamentally as we go forward, and it's certainly the best performance the industry's had since 2000. In thinking about Micron's performance specifically and how we look at it, this is the first quarter data for Micron. In terms of revenue, $4 billion. We're second only to Samsung in terms of scale of revenue.
Key metrics that we look at, gross margin came in at about 32%, that is below the average of our peer play memory players, part due to the nature of our business model, which I'll elaborate on a little bit, and part due to some performance improvements that Mark mentioned and the people following me will elaborate on. In terms of our business model, not all of our return to our structure shows up in the gross margin line, notably because of our joint ventures. Inotera, a notable example, as well as the Intel joint venture. It does, however, those relationships generate significant return on assets. If you look here, net income on a non-GAAP basis came in at 22% for the quarter. That non-GAAP reconciliation is available in our earnings release, and I've got it following here just for your reference.
You can see we only dropped 10 points from gross margin to net income. That's best in the industry, and although we're a bit below average in gross margin, we are about equal to our peers on the net income line. On ROA, at 25% non-GAAP ROA, that is better than average among our peers in terms of overall return and obviously well above our weighted average cost of capital. When I summarize how we're doing and strengths and areas we're working on improving, property, plant, and equipment turns are a big strength for us. Part of that comes from our business model, our joint ventures, notably Inotera, where we do not consolidate them, so we pay for none of their capital. It's paid for by Inotera, which is a separate public Taiwanese company.
We have an investment in them where we get 35% of their income reported below the operating line, but very good ROA on the total business because of the low asset intensity of that arrangement. The second piece being the acquisition of Elpida, where we got very low-cost assets through that acquisition and have the best-in-class PP&E turns in our marketplace as a result. This gives us low fixed costs and a very high operating cash flow. Also, the tax structure is outstanding. We have a Singapore principal structure. Not to give you all the details, you're probably familiar with other companies that have a similar kind of structure, so we get low tax rates. We also have over $4 billion in NOLs in the U.S., and we also acquired and put on our books about $900 million of NOLs associated with the Elpida acquisition.
What that means in combination with our Singapore principal structure is we can operate in low single-digit cash tax rates for a number of years to come. I'd also mention that in terms of our operating expenses, you see there at 12%, our SG&A costs are best in class. We have very good effective SG&A structure. In R&D, we have the scale and with also funding in our joint venture partnership, to get very significant amount of output and capability from our R&D team and to keep driving our costs down over time. You'll hear more from Scott about that in a little bit. In terms of improvement opportunities, we've got work we can do and are doing on gross margin and inventory turns, specifically, not necessarily an absolute reduction, but inventory turns to improve our asset turns there. You'll hear more from Mark Adams about both those items.
This is the non-GAAP reconciliation for your reference. Another thing I wanted to mention is if you look at how we're performing vis-à-vis the overall semiconductor industry, using consensus estimates in terms of scale, market cap, revenue, or net income, Micron tends to be relatively large in terms of our semiconductor peers in scale and things such as net income using the consensus estimates for 2014 there. However, on valuation, we're below the median of our peers, and we're intent on improving the performance of the company as we go forward and improving our overall results. Turning now just to, briefly, investment priorities and talk a little bit about how we intend to support them. Mark mentioned our investment priorities as he went through that summary there. I just want to talk about how we're going to view that from a financial support perspective.
This graph shows operating cash flow in the green bars at the top. It shows CapEx at the bottom in the dark blue and light blue by year going back about 10 years. First thing to observe is that our operating cash flow has been positive every year over the last 10 years, even in the very difficult, probably the worst market cycle we ever had in 2008, 2009. We still generated, at worst times, about $1 billion of operating cash flow. Further, if you look at our quarterly results, we've had positive operating cash flow every quarter, in contrast to most of our pure-play memory peers who have had at least a quarter or two of negative results. On the CapEx side, the dark blue represents the CapEx we spend on tech node migrations that Mark talked about, and also maintenance CapEx.
The light blue are capacity expansions that we've invested in over the last 10 years. As you can see, it was in the years when we invested in capacity where free cash flow, which is the yellow line here, went negative. Notably, our expansion into the NAND business, where we built the capacity up in 2007 and 2008 with our partner, Intel. That took our free cash flow negative as a result of that capacity expansion, notably. Then we built out the additional NAND capacity in Singapore in 2011, 2012, and you can see that on the graph. Those were the periods when we had the most significant impact on free cash flow. One thing to note is that in our partnership, we consolidate Intel results from the JV partnership. They actually fund half of the JV partnership capital.
This is the total CapEx here, or gross CapEx, if you will. Take, for example, 2011. Intel funded about $1.1 billion of that CapEx, which came in as a capital inflow and is not netted against that number. The actual cash effect to us was less as a result of our partnerships, another example of the value of that relationship financially. Another thing I'll point you to is that I show at the right there, first quarter 2014 annualized. This is the first quarter we have Elpida for the full quarter in our results, so I'm using that for indicative purposes. This is not a forecast.
I'm not giving you new guidance, but rather just taking the first quarter and annualizing it to show you the scale of operation now that we have Elpida fully in our results, and our ability to generate a significantly higher level of operating cash flow. If you look at the CapEx at the bottom of that same bar, first of all, there's no capacity expansion planned in our numbers going forward, and it is somewhat larger. We've got plenty of operating cash flow to cover what we're doing, but we shifted some capital out of 2013 into 2014. I commented on that on earnings calls previously. We also have some incremental CapEx associated with the integration of our two companies. You can think about that in terms of being somewhat additive to the 2014 guidance, and this number's the midpoint of our guidance, $2.6 billion-$3.2 billion, that I show here.
That will probably continue in 2014 and somewhat in fiscal year 2015 as we integrate the two companies, then we should move to a lower level of capital spending. Another thing about capital spending is our capital intensity is also declining. This shows a history of what's been going on, and you can again see the big bubbles in terms of capital, similar to the prior graph, relate to our expansion of our joint venture into the NAND business. It does not have the netting of the Intel results in here either, where they help fund part of it. The thing to note is the trend line is down, and we expect capital intensity to continue to decline, as Mark mentioned.
Although our cost per tech node, our investment cost and capital cost per tech node is moving up as things get more complex, that Mark referred to and Scott will talk more about. Those tech node migrations, from an industry perspective, are actually moving out. The net effect of those two things is we believe that capital intensity will continue to decline over time as we go forward and continue down that trajectory. Finally, now I'll turn to capital allocation strategy and give you a little perspective on what we've been doing. Breaking it down in terms of our strategy, the important thing to understand is our highest priority as we think about our capital management is operational flexibility. It has been strategically critical for us to have the operational flexibility to make the moves we need to make at the time we need to make them.
This, for example, would include things like our Inotera partnership and our purchase of our 35% equity interest in Inotera in 2008, a very challenging year for Micron and the industry, but executing that. Then more recently, the Elpida acquisition that we did in 2013. We had the structure and capability in place and did planning to manage that. That's our highest priority. What does that mean? It means that we have no covenants or very few, if any, covenants. We don't want those to be a problem from our operational execution and operational strategy. We want manageable debt maturities. I don't have that graph in here, but I think you've all seen it. We carefully work to have debt maturities that are relatively low by year, and we don't have any big discontinuities in that, and we're in good shape on that score.
Focusing on stable liquidity and balance sheet flexibility as we go forward. Once we got the operational objectives in place, we focused next on lowest cost of capital. That means to date that we have gone for asset-backed leases, and we do CapEx on a continuous basis. We lease it up on a continuous basis. We amortize it every quarter. They're about four-year cycles, and as we get new CapEx, we just roll it into the mix. You can think of it as a baseline of standard financing that we have in place that amortizes fairly smoothly over time. Secondarily, we've used convertible notes because they come without covenants and gave us the flexibility we needed. Assuming you can manage your equity, they have relatively low cost of capital.
I think you all know we have a high-class problem, and that is that the equity's become a little bit more challenging. I'll talk about that in a minute. Going forward, we now as a company are in a place where we believe that we can add some straight debt to the mix, and this will give us the operational flexibility we need. What we just announced on Friday and closed this week is an example of that. It was our first foray into the debt market as a company. We did it because the rates were in the right range in terms of cost of capital, but more importantly, back to our strategy, it enabled us to have the operational flexibility we required and don't have any constraints on our ability to do what we need to do.
Going forward, you can imagine that we will hopefully, if things continue in the current structure, have some straight debt in our mix, along with the other parts of our strategy that we've had historically. What do we do with surplus capital? As Mark mentioned, we believe we're in a better place as an industry and as a company going forward. Clearly, we want to maintain minimum cash levels. This is part of our liquidity management to support operations as we go forward in the business. Those minimum cash levels we target to be comfortably north of $2 billion at any point in time. We focus on dilution management and dilution reduction. We've made a number of moves recently here, and I'll show you the net effect of that.
We want to reduce debt in an absolute sense, as well as reduce our leverage ratios over time. What you can expect from us is that we will drive absolute debt down over time over the next couple of years, and we want to bring the debt to capital ratio, which we target at 20%-25% long run, back into that range in a couple of years. We obviously, as part of that, intend to get to net cash positive. Once we do, our other options to optimize shareholder value would include probably top of our list share buybacks as a consideration. Finally, I wanted to show you what we've done with dilution reduction efforts from our recent moves. This graph shows sort of the baseline in the dashed line at the top there, the pre-transaction diluted share count.
As you know, with converts, it varies with stock price, it moves up over time somewhat. The transactions we did in November, the moves we made in January, and the moves we made in this last week all have brought us to a place where we're about 107 million fewer shares diluted exposure, including the benefit of our $950 million of capped calls that we took against our converts at a $24 stock price. A significant move, but there's still a base of 50-plus million shares there if you look at that bottom line that we have in our diluted exposure. We'll continue to look at that and manage that over time as part of our capital strategy. With that, I'll wrap it up and take any questions you have. Ivan's got a mic if you have a question. Mehdi.
Ron, you had the chart talking about estimates and then valuation. I would like for comparison, you had [senders]. They do provide revenue and margin guidance. Is there any way you can do to do so? Maybe that would close the valuation gap. I have a follow-up.
That's a good question. I think the challenge in our industry, and our company certainly has been that it's difficult to project ASPs. We are certainly going to be monitoring and watching as things go forward here. We believe that the industry's moving to a better place. With that kind of stability and improvement, we can look at different ways of approaching the way we communicate our forward-looking views. Certainly under discussion, no big conclusions at this point.
The follow-up has to do with, you had the free cash flow. How should we think about the remaining payments to Elpida shareholders? To what extent is that going to consume some of the extra free cash flows that you're generating?
Sure. Another good question. We got JPY 140 billion installment payment obligation to our Elpida creditors as a result of the acquisition. That's scheduled out, and we set it up in our offer such that we would stagger those payment maturities over time to meet our objective of not having unusual bumps in our debt amortization. Another thing to note is that those Elpida installment obligations are an obligation of the Elpida subsidiary, which is in bankruptcy proceedings in Japan. There's no parent guarantee on those obligations. That's the obligation of the subsidiary to pay them. You can think of it, Elpida has to pay their installment payments, as our subsidiary, and they have to pay their CapEx requirements. Obviously, we monitor this carefully over time.
When we brought them in, actually right now, they've got about $1 billion in cash, in Elpida, $1.4 billion in total, if you include Rexchip. They're in a relatively good cash position right now. As we go forward, in our transfer pricing mechanisms, we can modulate and monitor the capital obligations and cash requirements of our various operations through that arrangement. I would mention one other thing, and that is in December, we converted them to the Micron legal transfer pricing structure. Starting December 1st, they're now on a cost-plus basis, as are all our fabs, cost-plus payments back to our Singapore principal operating organization. That's the structure I mentioned about that we have in place with a Singapore principal structure.
We've got a good balance of cash in Elpida right now, and over time, we'll monitor the cash that goes in there with our cost-plus transfer pricing, from Elpida, where they pay Singapore that cost-plus payment.
My question is for the Inotera. You have renegotiated your agreement with Inotera. Could you help us understand the new terms, how does it impact your financial statements, how long is this new agreement for? Is there any chance that you could renegotiate it again, like 12 months down or 18 months down the line?
Okay. The question was about the Inotera agreements. We renegotiated those. We actually set in place a three-year agreement. It has provisions to revisit some of those terms annually. This was the first year. We started January 2013 with that arrangement. We had a review and reassessment of that being the first year of our arrangement. We did, as they also committed, have some adjustments that were built into that structure. Those went into effect January 1st. It's basically an ASP. For those of you who don't know, it's a market minus or ASP-minus pricing model that has some adjusters based upon their profitability. That has an effect as we go into this calendar year.
As I mentioned on the earnings call, it will not have an effect in our fiscal second quarter because there's a three-month averaging period and lag effect that go on that will shift that out in time. Of course, it always depends upon the prices in effect, in DRAM and how they're moving over time, how that calculus works. That's what's going on. Most of that was put in place at the beginning. We tweaked it, if you will, in this latest cycle. We have the opportunity, to your question, to address it every year as part of that agreement and then renew the agreement, fairly straightforward and automatically over time. That's something that we'll be looking at each year as we come to the renewal of that activity.
Ron, Mark has talked about an increased focus on system-level products. What type of additional investments, new type of investments, do you anticipate making to support that system-level focus?
Well, I'm going to let Mark a little bit later address that if he wants to elaborate on it. From a pure financial perspective, we're certainly looking at it as a company in terms of our internal capabilities and having the internal resources and fitting those in our budget as we go forward. We're doing a lot of that internally. We will regularly look at inorganic opportunities. I'll leave it to Mark if he wants to comment more on that later. We do that all the time. We would be remiss if we weren't evaluating all the opportunities in our space as they come around. We do that on a regular basis. We are pretty regimented in how we evaluate and consider opportunities.
You see what we did with acquisitions like Inotera, or the joint venture with Inotera and the acquisition of a minority interest in Numonyx, and more recently with Elpida. We put pretty significant constraints on outside moves we make to justify them. I don't think you're going to see a change in that behavior from a financial perspective. Clearly, we want to make our strategy paramount and execute what we need to do to pull that off.
Ron, can you talk about the yen? What's your hedging strategy there, and what's the sensitivity to the gross margin and the net income line? Thanks.
Yes, I do remember, 79.16 was the yen exchange rate. We spent a lot of time thinking about what we were going to do to manage our exposures and do since. In terms of the sensitivity of our cost structure, rough order of magnitude for our Japanese operations, about half of the costs are in yen. You can think of it that way, about half of the cost. Just to make sure I understand what we do with currency exposure, we fully hedge our net balance sheet today. Most of it's done with natural hedges, both in Taiwan for the Rexchip piece and in Japan. There can be relatively small income loss effects that might flow through on our hedging exposure on the balance sheet, but we use natural hedges for most of it.
We just got a natural offset, and we focus on that natural offset to make that work. For example, I mentioned about the cash we have in Elpida. We're carrying a lot of it in yen, and we can balance that against the debt exposure. That's how we do that. It's a pretty good structure in terms of that. In terms of operating expenses and some of the participants in our industry who have big yen exposures do longer-term operational hedging. We have not engaged in that at this point. We might consider that over time once we get a better view of the accuracy of our predictions about those activities. Also when you do cash flow hedges like that, it just delays the inevitable, if you will.
It doesn't give you fundamental economic value, it just gives you stability over some forward period of time. We'll be evaluating that and deciding if we want to hedge our operational costs. We aren't at this time.
Ron, I just want to make sure I heard you correctly. You said that there would be incremental CapEx for integration with Elpida. Can you quantify what that might look like over the next couple of years? Secondly, as you think about a few years down the line, buybacks, and I'd like to get your thoughts on dividends. Can you help us understand, given the corporate structure, cash onshore versus cash offshore generation?
Sure. In terms of our cash availability, let me start with where our cash is. I mentioned we have cash in Elpida. We think that's pretty well-balanced, and we can use that to manage those obligations over time. If you look at our strategy and executing our strategy and our capital requirements, we have some flexibility in our structure to move capital around. In general, we can be somewhat flexible. We have the same sorts of constraints that a U.S. company has that has a structure the way we have with deferral arrangements. One difference, we have significant NOLs, and we have set up a structure where we have some greater flexibility there, and we can preserve that very valuable asset in both Japan and the United States over time, and it gives us some degree of fungibility.
We're monitoring that closely in terms of how we need our capital and where we need it. In terms of your question about buybacks, obviously, that needs to be U.S.-based cash. One of the ways our structure works is that our Singapore operation pays royalties to the United States because the United States owns all of our intellectual property. All the IP value is in the United States, and Singapore pays royalties to the United States. As they do that, NOLs offset those royalty payments if there's net income in the United States, and we balance it that way, and we can do that for a number of years. We can move cash a number of different ways to do what we need to do. Share buybacks, for example. In terms of dividends, that's down the pathway.
Once we get the cash surplus, we'll focus on buybacks as our first priority.
Can you quantify the integration CapEx?
The comment I made about integration CapEx, I don't have a specific number for you. It is really hard to quantify because it's completely intertwined with the integration node of our business. For example, in DRAM, it's the 20-nanometer node. It's real hard to sit down and try to ask our operations, "Well, what piece of that is integration-related?" All I can tell you, it's a few hundred million effect probably in fiscal 2014, and some of that will carry over into 2015. I don't have a good way to quantify it for you. Okay. Thank you all. I'll turn it over to Mark.
Thanks, Ron. Good morning. I'd like to also thank you for making the trek to the non-snowy East Coast and a very dangerous golf course for me yesterday. I hope those of you who played with us were far away from many of my drives. I wanted to take the chance today to talk a little bit about what we're doing from a business and operations standpoint. Mark talked about the context and the environment in which we're operating as a company, both in terms of the changing industry structure, changing customer relationships, as well as where the end products are going. Ron has certainly talked about the financial environment in which we're operating and the leverage that we have there to drive our business.
If it's okay, I want to take a step back as I think about what motivates all the employees at Micron on a day-in and day basis. It's pretty staggering to think about where memory is going today and how complex the process is. I imagine many of you have had the opportunity to tour a wafer fab. Memory technology is a complex manufacturing process and a complex technology in general. In fact, if you look at DRAM, for example, there's roughly 40-50 layers inside the manufacturing process of DRAM. In terms of total process steps, it's close to 500 process steps in memory.
If you remove irrational government investment, if you remove bailouts on debt, that's all that stuff that's happened over the last 10 years, then you think about where memory is going, what motivates us on a day-in-day basis is we fundamentally believe the memory market is set up to be a profitable and good return business for the right reasons. Look at the applications today. In every one of these applications, consumer, automotive, networking, data service and storage, in all these examples in end applications for memory, if you don't have memory, you don't have the applications.
Memory is an integral part of what we do on a day-in-day basis, as all the employees in Micron, as we think about what we're enabling with our technology, we think that it's time for the memory business to get the spoils and get the right return on our business, we think we're positioned to do so. We are driven to be the world's best memory company, what drives us are basically five strategic objectives. First and foremost, we're trying to take our precious capacity and drive it to the highest value opportunities to get the most for our investment in capacity. Part of enabling us to do that is being a leader in quality. That definition and investment in quality gets to be more critical when you think about the end technology segments we're going into today. Technology leadership is expanding for us.
In the past, technology leadership, you might have thought, what's the next process node that Micron's going to? Well, as you'll hear from Scott DeBoer later in our presentation, technology leadership isn't just around process. It's around things like packaging and advanced controller development. Our business is shifting to be more than just a silicon-based business, our leadership and technology needs to drive us there. Customer relationships. It's an interesting dynamic which is going in as we're evolving from more of a differentiated solutions opportunity. Our relationships with our customers are shifting, I'll talk to that during my content today. Finally, we have to do all of this in the context of being world-class in terms of operating efficiency.
I've got a couple slides in terms of the types of things we're focused on in our business to drive operational efficiency against some pretty fierce competitors. A lot of you asked last night at dinner and yesterday, asked me, how would you articulate what's different in the memory business? Why is it different? This is a pretty staggering slide. When you look at the nature of where memory was just 10 years ago, it was pretty concentrated, and it was pretty focused around DRAM, volatile memory applications, mostly compute, some networking. Even just 10 years later, you can see the diversification, not just on the technology side, where we've grown into a flash memory presence and leadership on the technology side, as well as NOR and emerging memory types.
More importantly, we've started to develop a competency to explore vertical markets and how we add value to that business. Our focus operationally and from a go-to-market perspective is driving more of that capacity to these higher-margin opportunities for the company, better returns, and actually places where people get an appreciation for what memory can do to their end systems. That's been a driving focus to get us to where we are today, and we think with the combination of Elpida and the new Micron, we have a lot more leverage to drive better margin into new application segments. A lot of questions are about capacity in the industry. What's going to happen with capacity? As Mark stated earlier, we're fine with capacity. In the last 18 months, we've increased our capacity 90%, slightly over 90%, of capacity that was in the industry.
It has given us a model where today we've got plenty of capacity to drive to these new segments for us. When we think about our business today, whereas you might expect a few years back when we were 10% of NAND trade and 13% of DRAM, our customers were pushing us for more capacity, that's not the conversation we're having with customers today. The conversation we're having today is how do we use that capacity to enable our customer relationships? This is putting us in a position where right now we're saying, how do we get the most out of this capacity? That's not just about market segments, but it's about the things we do. One of the areas that you've seen us focus on over the last couple of years is the manufacturing network in general.
When compared to our competitors, Micron has more of a distributed footprint, and part of that openly has come from some of the inorganic activities we've taken over the last 10 years around consolidation. Over the last two years, the team's done a phenomenal job of helping us define a blueprint for what that network can look like and should look like going forward. Today, this is what it looks like. We've centralized basically around three high-volume manufacturing areas. In Hiroshima, as part of the Elpida acquisition, we now have a dedicated facility with a primary focus on mobile. Of course, we're building our manufacturing network to have flexibility to shift with the way the market moves.
With this mobile base, both from the R&D perspective and capacity perspective, we're leveraging the best of Elpida's technology on the R&D side and being able to be closely aligned with what we get out of Hiroshima. In Taiwan, with both Inotera and Rexchip, we've got a wonderful high-volume base for computing and server DRAM technology, leading technology qualified in all of our customers moving forward to drive a very good focus on high-volume DRAM manufacturing outside of mobile. As I said, we have the ability to flex in and out of some of that capacity, but again, a centralized R&D focus to drive the most efficient operations in those fabs. In Singapore, that is our primary location for non-volatile high-volume manufacturing.
As you know, we are converting our Fab 7, as we've talked about in the past, from DRAM to NAND, that gives us a very good focused high-volume geographic location tied into our R&D and efforts to drive non-volatile technology. The U.S. presence, we have basically two sites. Our Lehi site is part of course, the IMFT relationship, that continues to be very important for us for part of the development of not only today's non-volatile, but emerging technologies for us in non-volatile applications. Our Manassas, Virginia location is a great site for us for some of our legacy businesses. When you consider something like our Embedded Solutions Group led by Tom Eby, that capacity there serves that business very well, and we're able to consolidate a lot of his needs in a business like Embedded out of a site like Manassas.
What you don't see now are our three facilities that were not as strategic for us. In the last three years, we got out of facilities, and we made sure that we exited these sites in a very strategic way for the employees, then we moved them to industries that were better for those sites and better for Micron. Israel and Italy and those types of sites for us no longer are in our network because they didn't serve the right financial model for us going forward from a technology base. We continue to look at ways, not just on the front-end manufacturing, but also on the back end and how we assemble and test, as that becomes more critical to our solutions to drive our business. Inventory management's been a big focus for us.
Notice we don't say inventory reduction, because we're driving in a market like today to have the best terms, as Ron commented on in his section. We've had some pretty good results over the last two years. Our inventory's down dramatically relative to where we were. Our terms are in better shape. The difference right now is we're serving a much broader end market segment, and each of those segments have different behaviors. I'll go back to the Embedded example. Relative to commodity memory, we don't have huge turns in Embedded because part of that is they need reliability of supply for their mission-critical applications, and they pay for it. As we think going forward in our business about how we react to industry supply and demand, we don't think that the immediate answer is to lower price and clear inventory.
That's not how we think about a solutions-driven memory business. As we think about our business going forward, we have many options to drive our capacity to different segments. The nature of how we look at inventory will drive us to think about, well, maybe we could move some of that capacity to better opportunities in the coming month and a quarter, as opposed to what historically has been a business that has been driven by end-of-quarter deals from our competitors and from customers knowing our fiscal year better than we do. The change is we don't feel compelled that we have to drive inventory down just to succeed in the memory business. As a matter of fact, strategically, we may hold inventory for better margin opportunities.
We will continue to drive optimum performance, but also not feel the pressure necessarily to react by cutting prices and selling inventory because we believe we have good homes for this capacity, and that is a big shift in how the memory business can operate going forward. A lot of what we're doing in the manufacturing and back end and throughout the company is trying to drive better performance in how we operate, from the time we buy materials all the way to finished goods. We've gotten some outside help in the last two years in terms of how we operate our fabs. Quite frankly, the last quarter that we reported, Q1, we got a pretty noticeable benefit for us in cycle time improvements.
I use the example here in NAND, both in the 20 and the 16 nanometer process, that we're driving improved performance in our fabs. That comes from taking a look at everything we do in the fabs and understanding our practices versus other peers, not just in memory, but in the semiconductor industry. We've been able to distill a much harder and rigorous process into driving performance. Again, not just in the front end of our business, but also in the back end of our business. Having third-party suppliers come closer to us and operate their businesses on campuses of ours to drive better performance and better inventory optimization.
You're going to see us continue to talk about these things in coming calls and meetings such as this to let you know how we're doing to become the world's best memory manufacturer on the manufacturing and operational footprint. Separate from things like how we look at the asset and capacity of inventory, and how we look at driving better utilization of our fabs in terms of throughput and cycle times, is how we are engaging with our customers. As you saw in both Mark and myself's content earlier, is when you go engage with an automobile manufacturer or a large financial institution looking for enterprise storage, those conversations are wildly different. On top of that, when you have the solution as opposed to an ingredient, those conversations are entirely different.
One of the things we're focused on at Micron is to continue to bring in resources who understand the markets that we're selling our technology into and their requirements. You can see that's a pretty big shift from a decade ago, where memory was primarily a demand fulfillment model. As you think about our ability to go identify working relationships with large-scale customers in the cloud, enterprise storage, if you think about those conversations, they're more about, hey, what can we do together? Not what is the price of your product. What can we do together to enable a differentiation for my solution? You'll hear a lot from Brian Shirley today in terms of how we look at the overall memory solutions opportunity at our customer base and what those conversations are like.
We have to be able to be set up to engage with customers that way, both in pre-sale and understanding their requirements, implementation, and really post-sale. That dynamic is what we're investing in at Micron, is putting together not just a sales organization, but business unit mentality and marketing capabilities around understanding requirements and how to help our customers innovate. Finally, in the past, you hear a lot about modules and industry standard and JEDEC and what have you. In each one of our segments, you see unique product configurations. If you take a look at mobile, for example, you've got DRAMs and NOR, DRAMs and NAND, and packaged together in a solution. That solution is unique to a customer. I can't take that solution and give it to the next customer.
That's coming from an environment where you could literally take a module from customer A, and if they didn't want it, customer B could buy it. The shift in how we engage with our customers shows up in terms of our product development, and it's a core focus we have to continue to invest in to drive our ability to bring products to market with our customers. The opportunity is differentiation, developing a product for a customer. It's more of a one-to-one relationship. Beyond that, what we're developing, it's no longer about just a standard module configuration. It's about things like controller technology, firmware, software, and even as far down the path of developing storage appliances, for example, to enable our customer base. The shift in our business is driven by the demand for memory in these unique application segments.
The opportunity for us is to take the capacity, which used to be primarily what we brought to the market, and add value to that capacity through higher-value capabilities on top of the silicon. As we do that, we think we can drive better returns for the company, and we feel very optimistic that we're in a good position to take advantage of that. Again, Brian will talk more about that as it relates to how we see all of memory solutions going into our customer relationships. With that, I'd like to stop and take any questions on what I presented here or just general operational questions for the company.
Hi.
Hi, Joyce.
In order to kind of incent the sales force to think in the same way that you're thinking in terms of inventory management, have you changed the way they're incented, i.e., are they given gross margin goals? Are they given volume goals? Are they given this percentage of your business needs to be system sales?
Sure. There's definitely a change in the motivation incentives of really the whole company. The issues that you're talking about as it relates to, shows up a lot in product mix, for example, too, how we're trying to incentivize market share opportunities in the higher-margin segments, as an example. That is very well aligned with what we're trying to drive the behavior of the team in, not just in sales, by the way, and the BU marketing teams in trying to drive relationships that enable us to grow our share in high-value segments as an area that also beyond sales is driving behavior.
Just as a follow-up to that, the very last slide you had, where the solutions would enable you to offer higher-margin products. Of the five points you have there, five solutions. What are the key solutions that you think you have a core competency inside the company, and what are some of the areas that you think you can benefit by going into the market and making acquisitions?
I think suffice it to say, the way we've listed these, from the base upwards, become areas that we're going to continue to invest in more and more. If you think about components, that's kind of been our core capability of the company. We've got good controller technology in the company as well. Although, in certain applications, primarily commodity-based applications, we may look to leverage third parties where we don't add a lot of value to controller development. We have a pretty strong controller organization within Micron. As you get up the stack here, we're looking to both organically build out these teams, and then see if there are other opportunities that we may look to acquire. Each of these elements are definitely critical to succeed to advance us down the food chain.
I have two questions, if I could. One would be on your 3D NAND strategy. From an operational standpoint, how big do you think 3D NAND is going to be for Micron over the course of the next one to two years in terms of your own capacity? How much you allocate on planar versus 3D? I have a follow-up.
Okay, I'll come back to the follow-up. We're not set to talk about that for a variety of reasons, one of which is, we don't want to send any more competitive information than we need to on our 3D NAND strategy, per se. I will let Scott talk a little bit. We feel pretty good about what we're doing around 3D. We made some choices at the technology level, on the architecture of our 3D NAND, and we think both the technology we've developed and where we're going from an end market segment initially, will be pretty successful relative to what we know our competitors are doing.
That's helpful. Thanks. As a follow-up, I think Mark laid out some overall supply growth forecasts for the industry that you see for 2014 and 2015 and beyond. In the past, you've said where Micron is going to end up relative to those industry targets. Could you maybe share with us those-
I don't think we're prepared to talk about that other than what we've shared with you so far on fiscal year this year. I think the message that I'd like to reiterate that Mark stated is, we're going to be awfully careful about how we look at that. We don't feel, in the current industry structure, that this is a market share game. This is a return game. I think that's both the message that both Ron and Mark have tried to hit you with is, and I was stating earlier, if we just took our current capacity, and as we execute moving those to better value segments, we'll have a much healthier business, without trying to nail the number per se on supply and demand in terms of that. That's been a tough challenge.
Hi, thank you. Can you talk about how you think about the speed that 16-nanometer NAND, the nodes you're ramping right now, will depreciate over time, and how the speed of that depreciation impacts your desire to hold inventory? Just in addition to that, how long you think that planar NAND will be sold into the marketplace. Is that something that will be sold in 2016, 2017, 2018, or will it be supplanted by a different technology? Nothing specific.
Yeah.
How is that playing into?
Sure
your desire to hold inventory? Thank you.
Well, maybe the best examples are to think about markets like NOR that have been a pretty successful market for a lot longer than people might have forecasted. When we think about planar NAND and then the 16-nanometer platform, which does drive a lot what we think. In the inventory perspective of the old days where people were running fast and process migrations were every 12 to 18 months, and costing, that's one set of parameters. Today where there's longer lithography curves, and in fact, 16-nanometer looks like a very mature technology relative to anything NAND planar after that, we have options, and businesses will require that. We have options that need product support for 3 to 5 years at a minimum, even longer. As we think about the behavior in NAND, we think the planar NAND does have some long-lasting opportunities in the marketplace.
As I showed that chart earlier about certain parts of our manufacturing network, we're set up to drive legacy performance that way because, quite frankly, the business wants it. It's justified. Not everything is a $ per gigabyte model, and as the business evolves, we think there'll be more of that.
Yeah, I have a question about the way you're talking about managing inventory a little bit differently. Can you put that in the context of NAND this quarter, where you had a bit of incremental NAND supply, and it looks like you're hitting the price a little bit as you move that out. How'd you balance what you're doing there versus holding inventory and kind of looking for a better price once you get new customers qualified?
Well, it's interesting. In the past, in the memory business, you'd say, well, over time, memory pricing's going down, so you better get rid of it early and get out, right? That's kind of the old memory model. As we look at our business right now, we don't see that. We don't think that way in terms of our business. There are end markets that can consume that memory, maybe not in the current quarter. Maybe it's not in component form, or maybe it's not in USB form, or maybe it's not in mobile in this quarter, but maybe that there's a high demand for another product that we will move into the following quarter. We look at our overall inventory perspective a lot more on the demand side of where can we move this capacity rather than just dump it, per se.
I think that's a behavior that's consistent with the rest of the memory industry. At these end market segments, it might not be ready in the current period, whatever that period might be, but we feel pretty confident that we can move these to other application segments without putting a lot of this inventory out just to move it from our inventory.
quarter where you have some incremental NAND-
Yeah
supply coming on, you are moving it out. Why not? You talked about on the call the sort of getting new customers
Well, I think the issue you're talking about just in the current-
Yeah
is more driven by the material that's coming out of a new fab coming online for us.
Yeah.
Those products aren't yet qualified. I made a comment on the earnings call that whenever you bring up a new fab in that capacity, the customer qualifications have to kind of come in at a certain time that's lined up. There's always a little bit of an early pre-ramp material that we have to get rid of in some shape, way, or form. I think that's more an exception than the norm going forward.
Thanks. Ron had mentioned one of the areas of improvement were gross margins. I'm wondering if you could talk a little bit about the path to gross margin improvement, broken by on the DRAM side as well as on the NAND side. On the NAND side, just a little bit, you're below some of the competitors.
Yep.
What tailwinds do you see on that side? On the DRAM side, if you include the equity income from Inotera, your non-GAAP DRAM gross margins are higher. Maybe you could talk a little bit about how you look at the DRAM gross margins as well.
Sure. I think both Brian and Scott will get into more detail, but let me just make a couple comments on that question, which is around gross margin performance. On the DRAM side, we've had a pretty good history of driving our capacity in the segments. When you look at some of the performance in server and networking, in automotive, in those segments, we've already got products and customer relationships that drive us to a leadership position there. Where we are on the DRAM side is Elpida and Micron in parallel are trying to drive process technology to be the best in the industry. I think that's where, for us, in addition to continuing to grow our share in some of these segments, innovating and leading us to a leadership position in DRAM process, it will be critical.
That's one more of technology investment, and I think one of the hidden benefits of Elpida is, we have now two R&D teams coming together as one and accelerating our path to faster technology migration in DRAM. On the NAND side, there are a couple of different issues that I would say that we're working on right now. First of all, a number of our competitors have done a fairly good job migrating their capacity to TLC, triple-level cell. In the past, our philosophy, and we think it worked well, was the triple-level cell, while it did allow for a cost reduction in comparable per gigabyte opportunities when you're selling these components, it also came at a price degradation because the products didn't warrant it. Today, we're finding out that our competitors have done a better job at enabling triple-level cell. That is not something we can't do.
As a matter of fact, we're in process. It's not new news. We're trying to catch up. We think that has some capabilities. We think that in addition to that, we are investing heavier to grow our mobile share in NAND. Around eMMC, for example, a market that we have not participated on as well because of our kind of legacy Numonyx relationships brought us into more of the feature phone market. Then we were later to the smartphone business, and we're playing catch up there in the business, I think you'll hear Mike talk about a little bit later, Mike Rayfield. We feel pretty good about our business there. We actually think that the issues that drive the NAND margin gap are more around how we go to market and product and technology roadmaps, and we think it's correctable, and we're on a path. Yep.
Time for one more question.
Just as a follow-up to that, 50% of your NAND bits goes to SSD market, right?
Right.
Could you maybe talk about how much is enterprise SSD part of that, and what do you think your market share in enterprise SSD market could be over the next three, four, five years?
Well, there's two issues going on. In the SSD sector, obviously, there's client and enterprise. We've seen some pretty aggressive, competitive pricing on the client side. This is a good example of could we have taken some material and maybe thrown some more in the client market? Yeah, we think it would've been bad from a margin and overall return perspective. On the enterprise side, we continue to develop and grow, and we had another good quarter in Q1. The client business is an interesting one because we want to see how that behaves. Because, yes, it's a higher ASP business, but there's obviously a lot of cost going into it, and in a scenario that you have a large competitor trying to gain share, we want to be careful about moving our capacity to an environment like that when there's other alternatives.
We think in the short term, we're going to continue to drive enterprise, but we think the client one has to be watched because we don't want that to become just another commodity memory business. That's just not what we believe is in the right interest of the company. Now, long term in memory around solid state storage, we fundamentally believe that you have to have access to NAND. That's skin in the game. If you look at today's business, people in the business trying to go out, they just haven't been able to succeed. There's a number of customers, both early companies and even a couple public companies, who haven't made it because they didn't have access to the core technology, and they're buying technology and basically integrating it in an SSD environment.
We don't think that works both from a scale perspective, and we also don't think it works from a technology perspective because when you think about enterprise storage and things like reliability and endurance, performance, and all the metrics that a financial institution or a large insurance company, these companies are driving enterprise storage applications NAND has to be enabled to perform at that level, and that enablement comes with deep cell technology level understanding and mapping with controller and firmware technology. There's got to be access to the technology. We think in the enterprise space specifically, if you don't have access to a NAND, forget the fact that you just can't build as much as you'd like because you don't have the capacity necessarily, you don't own it.
You need to have the technical know-how to actually enable that NAND to be world-class in storage, and we think that's pretty important over the long term, and that's really why we like our position over the next three to five-year horizon in the space. Okay, thanks very much.
Okay, we're going to take about a 15-minute break now. If you want to go use the restroom, grab yourself a refreshment. That puts us back here at about 8:40 to restart the next session. Thanks everybody.
The trophy that I made for earth and fur and gold. Got into the wrong pair of hands and truth was sold. We bought it for oh so much less and it was worth. Every man that touched it found a heaven on earth. Heaven is a feeling I get in your arms. Heaven is a feeling I get in your arms. Heaven is a feeling I get in your arms. Heaven is a feeling I get in your arms. The queen and the court, yes, their clocks adored. Their hearts fell to overdrive and nerves distraught. Mercy this and mercy that, let justice prevail. I just want my trophy back, it's not for sale. Heaven is a feeling I get in your arms. Heaven is a feeling I get in your arms. Heaven is a feeling I get in your arms.
Heaven is a feeling I get in your arms. Creatures of mercy, shoot. Shoot them down and set me free. Creatures of mercy, shoot. Shoot them down and set me free. When I got my trophy back, it looked unkind. All its hips and ghosts in black and suit with lines. When I put it back inside and locked the door. A trophy of mercy is a trophy no more. Heaven is a feeling I get in your arms. Heaven is a feeling I get in your arms. Heaven is a feeling I get in your arms. Heaven is a feeling I get in your arms. Creatures of mercy, shoot. Shoot them down and set me free. Creatures of mercy, shoot. Shoot them down and set me free. Oh, me.
I'm tapping sleeping cities, fading by degrees. Not believing all I see to be so. I'm flying over backyards, country homes and ranches. Watching life between the branches below. It's hard to say who you are these days. You run on anyway, don't you, baby? You keep running for another place to find that second grace. I'm moving on along the holy ground Statues that atone for my sins. There's a guard on every door and a twink on every floor. Overflowing with a thousand amens. It's hard to say who you are these days. You run on anyway, don't you, baby? You keep running for another place. To find that steady grace. Don't you, baby? You're rolling up the carpet of your father's two-room mansion. No headroom for expansion no more. There's a corner of the floor they're telling you is yours.
You're confident, but not really sure. It's hard to say who you are these days. You run on anyway, don't you, baby? You keep running for another place. To find that steady grace. Don't you, baby? You keep running for another place. To find that steady grace. Don't you, baby? Like a babe moving in stereo. Like a babe step on the scene. Like a babe shaking and trembling. Like a babe making solid sounds. It's so easy to blow up your problems. It's so easy to play up your breakdown. It's so easy to fly through a window. It's so easy to make good sounds. It's so tough to get up. So tough. It's so tough to live up. So tough on you. Like a babe moving in stereo. Like a babe step on the scene. Like a babe shaking and trembling.
Like a babe making solid sounds
I can't win. I can't wait. I will never win this game without you, without you. I am lost. I am vain. I will never be the same without you, without you. I won't run. I won't hide. I will never make it through without you, without you. I can't rest. I can't fight. All I need is you and I without you, without you. Oh, you. Without you. I can't erase, so I'll take blame, but I can't accept that we're estranged without you, without you. I can't quit now, this can't be right. I can't take one more sleepless night without you, without you. I won't stop. I won't claim, but you're not here like paralyzed without you, without you. I can't look. I'm so blind. I lost my heart. I lost my mind without you, without you. Oh, you. Without you. I am lost.
I am vain. I will never be the same without you, without you. Without you.
A glitter of my life. It's where I was born and raised. A piece of people they come and go. It's here that I remain. You know no matter the day, no matter the hour, you know where I'll be found. Oh, in the neighborhood. Down my road I hear the sound of glass. From my sister, she's still upstairs with you. My cousin Bobby smokes too much. Guess he's there in two ways. Ooh, Loretta from the coffee shop. She's a little piece of dynamite. I've been talking to her for years I think it's going back. At the end of the day, down at happy hour, you know where I'll be found. Same old scene, same old crowd, there's always someone around. Oh, in the neighborhood. Oh, in the neighborhood. Oh, I feel like I'm a million. Oh, I'll never be a stranger.
Home, I've been home. Feels like home. Charlie, he plays the guitar. Down in the courtyard every night. Maria, she's his girlfriend. All I can notice in a fight. My mama she needs a lot of. She's a little afraid of the dark. We go together and it's all right. 'Cause I'm named for me and I fall. At the end of the day, down at happy hour, you know where I'll be found. Same old scene, same old crowd, there's always someone around. Oh, in the neighborhood. Oh, in the neighborhood. Oh, in the neighborhood. In the neighborhood. I would turn on the TV. It's so embarrassing. To see all the other people. I don't know what they mean. It was magic at first. We spoke without sound. Now this world is gonna hurt. You better turn that thing down.
Turn it around. "Well, it wasn't me," says the boy with the gun. "Sure, I pulled the trigger, but it needed to be done. Life's been killing me ever since it begun. You can't blame me 'cause I'm too young." "You can't blame me," sure the killer was my son. I didn't teach him to pull the trigger of the gun. It's the killing on the TV screen. You can't blame me, it's those images seen. Well, "You can't blame me," says the media man. Well, I wasn't the one who came up with the plan. I just point my camera where the people want to see. Man, it's a two-way mirror, and you can't blame me. "You can't blame me," says the singer of the song. The maker of the movie which your faces I hung. It's only entertainment, as anyone can see.
The smoke machines will make 'em, but you can't fool me. It was you, it was me, it was every man. We've all got the blood on our hands. We only receive what we demand. If we want hell, then hell is what we'll have. I would turn on the TV. It's so embarrassing. To see all the other people. Don't even know what they mean. It was magic at first. It let everyone down. Now this world is gonna hurt. You better turn that thing down. Turn it around.
Okay. If everybody can please take your seat, we need to get started and get things rolling here. Please take your seat, and the next section is going to be
Scott DeBoer, Vice President of R&D. Okay. First, I'd start off with a little review of some of the progress over the past year, and it kind of leads into some of the questions that we're going to be going through and trying to address, and then also the big challenges of our focused effort of the next year. On the DRAM side in 2013, really like the rest of the company from a technology development point of view, it was all about how we put together one R&D team out of the talent in Hiroshima and other parts of Japan, along with our R&D team from around the rest of the world. Generally, I'll talk in quite a bit of detail about what the plan is, but we're really happy with how this has come together.
I think the focus has been great. I think we're in a really strong position now. Coming out of the year, we're in a solid position on 25-nanometer DRAM with the node reaching maturity and ramping fast. Also on our 20-nanometer effort on yield ramp is well underway, and we're in a solid position there. On the NAND front, we've had a really successful story on planar NAND over the past year from a development point of view on 16 nanometer. That's now doing great and ramping in manufacturing. I'll give you some more detail on that. That's really the picture or the story of NAND technology for 2014 in terms of volume manufacturing.
The other big thing from a kind of a milestone point of view in 2013 was really around the enablement of system technology in the form of HMC, Hybrid Memory Cube. In addition to Micron delivering products that Brian's going to talk about in a little while on the memory cubes to key customers, a lot of the technology story is around the infrastructure maturity that's occurred from our suppliers and from the overall technology base that supports 3D integration and TSV technology. We're really pleased with those. The upcoming year, and all these we'll talk in more detail about, the upcoming year is all about ramping 20-nanometer DRAM, the rollout in the year on 3D NAND, where we're still intending to sample some product in the first half of the year. The real story then happens beyond that in terms of manufacturing enablement.
The continued push into the space with real products and commitments to customers on system-level technology around HMC. Jumping into some of the key questions that have come up, first is all about what do we look like in the DRAM world following the Elpida acquisition? A lot of questions about what is our view of what happens to DRAM after 20-nanometer technology. I'll try to touch on those. First, a discussion around what our view of DRAM technology development looks like following the Elpida combination. I think this is a really good story. I think it's going to continue to play out over the next couple of years in terms of our technology position overall in the DRAM space.
One of the key pieces to how we put this together was we did have a joint development program with Elpida that we put together prior to the closing of the agreement. We did it about five or six months ahead of time. What that really enabled us to do is combine the teams and come up with a strategy so that we hit the ground running at close with a real execution plan. Certainly, I'm sure all of you can appreciate when you put two big entities like this together, there's a risk for missing a step on technology execution while you try to figure out how you're going to work together. I think the pre-close JDP really allowed us to hit the ground running and not miss a beat.
When we look at the components of our organization now to drive this DRAM roadmap, there's some real interesting dynamics around not just the bandwidth, which is, of course, very important, what the Elpida acquisition brings to us in terms of technology bandwidth, in terms of fab infrastructure and characterization, and all the kind of hard asset type things related to technology development. But also a team of very talented development people in the Elpida organization that come over and really bring a different perspective and a unique experience base. The combination, I think, is exceptionally strong and really positions us well going forward.
When we look at how we're going to do technology development, I think it really brings the best of both worlds from Elpida and Micron together, where we make use of the final development in manufacturing model that Elpida has talked about prior to acquisition for a long time. We have the nodes rolling out and ready to ramp up in a big volume manufacturing fab as soon as they're ready, and that's a key piece to the model going forward. And they're currently focused on 20-nanometer yield improvement. In Boise, we still have the capability to do some of the more disruptive technology developments that's going to be required as we face these difficult nodes coming up. And we can do some new materials and maybe more little riskier development in Boise without risking our manufacturing base in Japan.
It's a great combination, and I think it's going to position us really strong. When we look at our technology evolution path right now, our 20-nanometer focus right now is it's a hard technology development. It's going well. It's completely in Hiroshima. The two nodes behind that, in my view, are still difficult, but relatively evolutionary. I think we see a clear path to how we're going to do what we call a 1X and a 1Y node, but really they're the next two nodes on DRAM behind our 20-nanometer. I think we have a clear path on those, and I think the node behind that is the one that is potentially very disruptive and going to require some more innovation.
A year from now, we may talk about the fact that we think it's another evolutionary node, and certainly we're going to work very hard to make that happen. But I think we've got some other ideas for how we have to come in on the 1Z node in a little different way. Okay, on the NAND strategy, and I know there's a lot of questions around 3D, some of them already came out this morning. I'll talk a little bit about how we transition from planar to 3D and kind of what the status of that looks like right now. On planar technology overall, our 16-nanometer node is looking very strong and it's ramping in manufacturing at this point.
When we look at a couple of the key aspects of the 16-nanometer technology, the first one is that I've talked about at previous meetings, is the way that we've designed it's a very effective capital efficiency node for us. When we look at the cost of transition for Micron from our 20-nanometer node to our 16-nanometer node, it's one of the lowest in our history, at least. It was designed in a way to come in with minimal fab impact. That's looking really good still. The other thing that's really good is through a bunch of kind of strategy adjustments, we've also managed to really improve the quality early on in the ramp of this technology.
If we compare on this chart the performance of the yield improvement from the time we introduced it into manufacturing, this is by far the best NAND technology introduction that we've had. The green one is the previous one from time. Planar's looking pretty solid. We have our first SSD product qualified on the 16-nanometer node as well. We're feeling very good about this. What this does is set a very high bar for 3D to come in as a more cost-effective solution. We have an aggressive time on technology enablement for 3D, and then a little different timeline when we talk about where it becomes cost-effective and really a real cost advantage over this, what we think is the next 116-nanometer node. The 3D is progressing nicely.
One of the things that I think is important when we look at 3D right now, at the highest level, it really is starting to feel like you can look at it almost as just another shrink when you model how it's going to work out in terms of almost comparable to historical basis. If you look at the bits per wafer, obviously they're very significant. When you add in the complexity, the process complexity with it winds up being a cost reduction that's relatively similar to previous main shrinks. I'll talk a little bit more about that in a minute on the roadmap also.
The next big topic is around what we see as in the core technology area, disruptive technology that we have to be cautious with or understand and have a roadmap for, and some other trends where we think we need to be investing, and some of them have been mentioned quite a bit. On the core technology piece, one of the things to emphasize, again, when we look at the amount of bandwidth we bring in from the Elpida acquisition, certainly it's a big benefit on the DRAM side. It also carries over to being a significant benefit on the new memory technology, on packaging development, and through some reallocation of our internal resources, it winds up boosting up the bandwidth overall for pretty much every program we have in R&D right now.
When we look at the roadmap for the memory space overall and for the memory solutions that are required to cover it, we still see NAND and DRAM as being exclusive coverage for certain parts of the memory business for at least the next 5 years and really indefinitely on certain parts. If you look at the real high-performance end of the spectrum, which is shown in this graph, the DRAM is just very difficult to displace in this spot because of endurance and speed. Now, the cost structure of DRAM obviously doesn't compete with NAND as you go towards NAND, but DRAM itself has a very long life in terms of covering a portion of the memory solution space. It's going to be difficult to displace.
On the other side of the spectrum, there's lots of discussion about how Resistive RAM and other things fit into the picture, but on the pure cost point of view side of this equation, NAND is not going to be touched over the next 5 years at a minimum, no matter whether the technology comes in. When you look at the discussion around these new memory technologies, we believe the opportunities are really in a focused area on very high performance with non-volatility and kind of a medium space that we call Storage Class Memory, which is non-volatile, a little better performance than NAND, but a worse cost structure than DRAM. We see the world kind of evolving towards those two things as not replacement opportunities necessarily for DRAM or NAND, but for a memory performance solution-enabling type of technology.
We're putting a lot of focus into that segment of new memory technology development. In addition to that, when we look at how we keep differentiating in NAND and DRAM, we think there's a lot of room in that space also. Part of that brings us back to how we manage that memory, some of the controller development, different things that Mark alluded to earlier, putting more focus on that, as well as package technology, which we think is a true differentiator over the next several years when you combine it with the different kinds of memory technology. We really think HMC is just the tip of the iceberg as we look at what unique packaging technology can enable in terms of new system solutions. Okay.
On the roadmap side for both NAND, DRAM and new technology, I'll kind of talk through each of those, where Micron sees the opportunity. There's a few kind of subtle points from this too that become important. When we look at DRAM technology specifically, we talk about node cadence reducing, and eventually that does become the case. From a pure technology point of view, at this moment in time, our focus is on actually solidifying our DRAM technology base from where it's been. We have some catch up to do, and we believe the combination of Elpida and Micron puts us in a strong position to get this 20 nanometer node out and really start putting us back in a solid position on DRAM.
When you look at the 20 nanometer timing, then the next node behind it, which is 1X nanometer timing, those are kind of on a historic type pace, it's really because we're coming from a bit of a catch-up position, we need to nail those two technologies to put ourselves in a strong position relative to our competition. The 1Y node beyond it then starts looking a little bit farther out, as I mentioned, the node beyond that is one with a lot of disruption to it potentially timing challenges. On the NAND side, this shows our 16 nanometer technology, which is out in the market now, it's going to really be the dominant piece of planar technology ramping up in Micron's fabs over the next year.
We think that's a really solid node, as I mentioned before, it is going to impact the timing of where 3D NAND really makes sense. On this particular graph, I'm showing where volume capability is, which we could start with. The technology will be available from Micron's point of view to start looking at ramping up volume in 2015. When we look at the projection, we think it's more like a second half of 2015 story, when it actually starts becoming something significant in the marketplace. A lot of that is, again, driven by the fact that the planar technology is so cost competitive.
Last thing on NAND is beyond this initial generation of 3D NAND, there's still a couple more generations beyond that of 3D NAND cost improvement, density improvement, that we see on the roadmap right now and have significant programs behind enabling already. Down to the emerging memory technology line. We rolled out phase-change material technology in 2013. We produced volume of phase change. Going forward, we're focused specifically on the two kind of new memory technology pieces that I mentioned on the previous slide. One is in the ultra-high performance kind of place with non-volatility. The other is in this medium density, medium performance kind of space. The new memory technology we're looking to have positioned to roll out in 2015, 2016, 2017, all fit into those type of categories.
We're not providing much detail on what those technologies are right now for competitive reasons, we believe we're really well-positioned with the technology to come in and solve the solution space that we've identified as critical. Overall, when we look at our technology position, the roadmap we're going to enable, the performance we're getting the bandwidth that we've added with Elpida over the past year, we're feeling really good about our current position of the technologies that's rolling out also about our competitive position going forward. Okay. I'm happy to take any questions on those topics or other technology topics.
Scott, earlier there was some reference in an earlier presentation that the ability for customers to use memory to differentiate their products was more prevalent today than it has been in the past. Would you discuss kind of the perspective of how expensive it is for Micron to develop those solutions relative to the price you're able to garner and just how that equation might work and how you're looking at it?
Yep. There's a couple pieces to that. Obviously, from a core component point of view, we try to build in the capability that can cover a wide space of custom applications. We try to leverage as much of our core capability across, whether it's unique mobile applications or server applications or in package memory applications. We try to leverage the real expensive piece, which is the silicon development of the 20 nanometer node, for example, to cover that type of broad space. The second thing is when we go and look to the next stage of the cost around, say, packaging development for 3D-IC, TSV type things, again, we look at a space of applications, and we try to leverage our technology development across as big of a space as we can.
If you look at TSV, the stuff that enables HMC, largely also is very relevant for enabling new mobile packaging technology and Wide I/O type applications. We leverage it at that point. Where the cost starts coming in is just business rationale relative to unique designs, and the product engineering infrastructure to go make those happen. There's additional costs, but the big expensive pieces of it, we try to build upfront to cover the wide space of technology in general. To the bottom line R&D number, there is some impact, but it's not as big as you might think.
Scott, you mentioned on the cost of transition for 3D, that's about twice on the graph that you showed the cost of transition for 20 nanometer. Can you go over where is the cost involved? What is the effective way for how do you get when you go 3D versus that same 20 nanometer shrink that you showed? Thanks.
Okay. First off, on where is the cost involved, you mean from a fab tool perspective or? Yeah. When we look at a transition to 3D, a lot of the planar infrastructure actually is still totally usable. Most of the additional complexity from a process point of view is in dry etch film technology and CMP. Those drive, as you might imagine, when you're building a big stack of material, those drive a lot of additional fab capacity requirements in those areas. From others, like lithography, it relaxes the requirements. The bulk of it is in those areas. When we look at a conversion to 3D NAND from planar, there is a significant impact on the number of wafers you get out of a given square footage.
I think you can think of it probably roughly somewhere between a 1.4 to maybe a 1.8 kind of trade ratio in terms of wafers. Somewhere in that range of reduction. Does that get your questions?
That's a 50%.
For every wafer you would've started before on a planar node, well, for one wafer you're starting on a 3D NAND node, you have to subtract one and a half or somewhere in that ballpark for.
Got it. Thanks.
Just a question on your planar NAND roadmap. Some of the other NAND OEMs have talked about shrinks beyond 16 nanometer going to 1Y, 1Z, et cetera. From a Micron perspective, should we think about 16 nanometer as the last planar shrink prior to 3D?
Yeah. From a Micron point of view, roughly, the shrink looks about like a 1Y that other people talk about. The 1Z node that's out there from some people is a pretty marginal shrink on the 1Y node. It kind of depends on maybe your confidence in 3D and some other things, whether that node actually makes sense.
Just a couple of questions. Going back to 3D, when the second generation is commercialized, do you think it's predominantly used for MLC or TLC or both?
Both.
Both.
From our point of view, at least, when we go to vertical NAND, MLC and TLC will be enabled.
Okay. How should we think about the cost or CapEx difference from gen one to gen two?
I think back on a more traditional kind of NAND node conversion range. If you look at it, I think it'll be similar from when we went from 25 nanometer down to 20 nanometer. From first generation 3D to second generation 3D, it'll be kind of in that ballpark.
Is that because you're reducing repetitive steps, or you're reusing planar capacity?
There's not as much new steps in a second generation relative to a first generation. It's not as big a change.
And then when-
The reuse will actually happen on gen 1. When we go from planar to 3D, that's when you'll get some benefit from the reuse.
One last one, back to DRAM. Is there really a capacitor issue at 19-nanometer?
Well, from a DRAM point of view, I've been here 20 years, there's always a capacitor issue.
Is this one different than the past 20 years?
Yeah. The challenge that actually happens when you start going, at some node, whether you want to call it 1Y or 1Z, the challenge that starts happening is you're getting to the point where you can't fit the number of films that it takes to build a capacitor inside the space that you have available for the capacitor. You either have to come up with new films that can be substantially thinner, which is a possibility, or you have a capacitor problem.
Does your roadmap account for this?
Yeah. Capacitor technology is one of the fundamental limiters on where DRAM goes. Yeah.
Hi, Scott. Back on DRAM?
Yeah.
20 nanometer node seems to be the first node where you're merging the roadmaps for both Elpida and Micron, yet you have, I would assume, fairly different tool sets for both companies. I'm just curious how you're going to manage that. I think historically, both Micron and Nanya and Elpida have been pretty savvy in terms of shrinking at what's fairly cheap compared to your competitors.
Yeah
in terms of your CapEx. Is that going to be changing at 20 nanometer when you're merging these roadmaps, just how can we think about how these roadmaps can be merged considering the different tool sets?
If you look at our base for where 20 nanometer is going, we have Hiroshima and Rexchip, which are historically relatively well-aligned. They're part of the Elpida family, right? The technology from that point of view is going into two fabs that are pretty consistent with the new technology on the 20 nanometer node that we've developed. On the Inotera side, the story's a little bit different, we go through and make sure we do the best we can in terms of capital efficiency for Inotera to help them out with that. Then it's down to technical judgment on which tools can be used for steps. It won't be completely matched through significant parts of the process flow. Critical parts will be matched, and Inotera will have to go fund those.
I think the question's a good one, the challenge is there, we did choose the technology node to make sure that it was the most capital efficient, as you say, for that rollout, 20 nanometer is where we're aligning things.
Thanks. Hey, Scott, can you talk a little bit about the roadmap for Wide I/O as a application of TSV? Maybe talk about what timeframe you see Wide I/O in the industry from other players.
Wide I/O for us, or Wide I/O 2, is an important direction relative to mobile, Brian can talk about it a little bit more, probably Mike also, when we talk about the mobile space. I think it's, again, it's more of a late 2014, 2015 kind of story with real enablement in 2015 for some kind of volume that would be meaningful, if that answers your question.
Okay.
Sorry? Okay.
Scott, sorry if you said this and I missed it, how important is EUV to 1Y and 1Z? Can you do it without EUV?
Yeah. We can do it without EUV. EUV is important from a cost point of view, and it's important for several different memory technologies that we look at, but from a cost point of view, not from an enablement to the node itself. We have technology paths for 1Y, even for 1Z, without EUV, but at a cost point that EUV provides an advantage, then it fits in nicely to both 1Y and 1Z DRAM, as well as some of the other new memories. Well, we are assuming that we don't have EUV. With the rollout of 20 nanometer technology across our broad DRAM base, we're constantly making decisions that make EUV less appealing because we enable more and more double patterning capability, triple, quadruple, whatever patterning capability. Every time we go farther down this path, EUV has a higher hurdle to get to to work for us. Okay.
We're out of minutes? Okay.
Okay. Good morning to everyone. For those of you that I haven't met, my name's Brian Shirley. I oversee what we call the DRAM Solutions Group, which is one of our four business units. This morning, I'm going to be talking to you both about DRAM as well as NAND. I think the key question we want to try to help answer this morning is really as we look forward, what are some of the key market trends, the product trends out there, the needs for both DRAM and NAND solutions, and how does that overlay our ability to put those trends into good products using the processes and technologies that Scott's team develops? How do we get these into good, high value-added solutions for our customers?
To start off that discussion, I think the landscape I'd like to show you is shown here, essentially drawn into four segments, a pretty important segment, by the way, in the middle, holding all of these together. These four segments, you've seen this earlier today. This is a pretty good proxy for how we've organized the company in terms of our four business units. Thinking in terms of computing, a lot of what happens in servers, as well as, frankly, the networking pieces that hold it all together, obviously, the historical base of DRAM with PCs and now moving to ultrabooks, also game consoles that frankly can almost be thought of as compute platforms themselves.
Moving over to storage, obviously, the world moving quickly through new generations of enterprise SSDs and the world of cloud storage and making sure that the right parameters are in place for the necessary compute data coming over for persistency. Moving down to embedded, obviously, we have the world of automotive, industrial, medical, military, some pretty exciting applications down here. Historically, a little bit smaller DRAM content. Mike's going to speak a little bit more about that today, Mike Rayfield. Finally, over to mobile, with just, obviously, explosive growth. Frankly, as we look across this landscape, I have to tell you, from someone that's been in the memory industry for more than two decades, this is about as exciting of a landscape as we've ever seen.
Opportunities across this entire spectrum for higher value-added products that are not dictated by a particular JEDEC, wide-enabling spec, products that start to look a little bit stickier, higher value-added, in some senses, almost like ASIC-type opportunities for the customers that need the performance out of the memory. Some specifics on that, as we look at the DRAM side, our view, looking at a five-year demand bit CAGR here of about 27%. PCs, obviously, not the driver there, but the good news here going forward is that the growth, specifically in mobile as well as infrastructure, is just staggering. Mike will talk further about the handset opportunities as well as tablet.
You look across the infrastructure space, in particular, driven by the demand of servers, not just at the unit level, but frankly, high content per box at the server level, really giving us just a wide range of opportunities to play in here. It's just much more than a PC game, and I think that's evidenced by the graph. Some specifics on that, you think about the graphics and consumer world. We'll show you some products coming up here momentarily. New generation of game consoles that, as I mentioned, really are more computing platforms these days. These are not just graphics boxes. Using, to a T, 16 times the memory bit content of the previous game console. That's a pretty nice memory requirement out there from a memory supplier perspective. Networking, another huge opportunity for Micron. This has been a good historical segment for us.
When you look at the next generation of gear coming forward, the networking guys are talking about platforms that just flat can't be done with traditional DDR-type memory architectures. That's where Reduced Latency DRAM, where HMC, frankly, are not just interesting memory architectures. They are necessary to make the next-generation platform actually function. Again, getting down to enterprise, spectacular growth here that's driven partially by the number of servers, data center applications, corporate usage, the cloud, et cetera. Probably even more importantly, the amount of content per box. As, for instance, the industry-standard server platforms have moved from Sandy Bridge to Ivy Bridge, the amount of memory that can be placed inside of a server has taken some nice steps forward, and the server users are responding to that.
They're filling up their boxes with a lot more memory, we see that daily, having a hard time keeping the server modules in stock. Some great opportunities there. What you see here is a rundown. This is actually for fiscal first quarter 2014, Micron's revenue perspective across DRAM in total. This is the first quarter that we were able to put together the Elpida operation as well as the Micron operation. What you see here is, obviously, a nice set of products going into mobile, and what you also see is a good specialty segment penetration driven both by the Micron networking portfolio, automotive, industrial, medical, et cetera, as well as the graphics portfolio from the Elpida operation that was integrated. Server in there as well.
All of this server demand, as we've talked to you in the past, Micron in the past, Micron by itself, doing a good job, we think, getting our product server-capable, and for the size of the company prior to integration, a good percentage of our material out to server market. You'll see this server percentage of our bits grow going forward as we now take the 25-nanometer process at Hiroshima and Rexchip and move that into server-capable applications. That generally will come out of PC and mobile. You end up with three good, nice large segments with just a ton of value-added opportunities, and obviously the specialty segments of graphics, networking, and automotive and AIM applications as well. Turning over, talking some specifics on product lines here, I want to talk briefly about GDDR5. You talk about a success story out of integration.
If you had asked Micron 6 months ago, as we closed the deal, how would we gauge success of the integration with Elpida from a real perspective on product introductions, on technology? As Scott indicated, some phenomenal progress, not just on 25-nanometer ramps, but also 20-nanometer development. From a product standpoint, really the key things that Mike Rayfield and myself were keeping an eye on were making sure that through this integration, there were no hiccups, specifically in getting some next-generation graphics parts launched, and on Mike's side, some new LP architectures. Graphics, just a huge success story here. Since closed, our new team members in Japan and Germany have sampled, launched, and ramped a leading-edge 25-nanometer, 4 gigabit GDDR5 component, which frankly is the backbone of some of the obvious game console boxes out there. This is just a spectacular product for us.
We like it as well because when you talk about the crossover between, with our new consolidated customer base, graphics opportunities, as well as server opportunities, GDDR5 ends up being a component that has applications in both. For HPC opportunities, graphics DRAM is actually a pretty nice performance point being used in a lot of the upper-end HPC platforms. It starts to show you some of the synergies we think are already in execution as we go forward through the integration. Just a product line that we're very proud of to have marched forward through this integration. Another product that is here and now, something called Non-Volatile DIMM. This launched last year as a DDR3 module. What this is inside of a server today, all industry-standard servers using 8 gigabyte and 16 gigabyte DRAM modules.
What Non-Volatile DIMM is all about, to the user, the DIMM looks just like a memory module. You plug it in the same way, same interface, but on that module, there's actually a controller and Micron NAND behind that. For error data recovery, for system checkpointing, the module is actually capable of persistent storage. This ends up being something very interesting to just about every server OEM we're selling to. This is here and now in DDR3, getting some great traction, and in 2014, you'll see this extended up to DDR4. An HMC update at Supercomputer '13, last November in Denver. We had demos with several leading industry OEMs out there. We are now sampling the production device. This continues to go fantastically. HMC Gen2 here and now.
What you saw at SC13, several platforms from Fujitsu, from several of the notable FPGA vendors, and 18-plus design wins and counting for Gen2. Very pleased with HMC progress. You'll see this be something that's getting more traction and continuing to diversify in terms of variance of HMC for other applications. A lot happening with HMC over the next 2 years. Finally, something that we call the Automata Processor. Just about every officer at Micron has a different way of pronouncing this. Tomato, tomato, I don't know. This is Automata Processor, and what this is all about, this is something that we unveiled at Supercomputer '13 as well. This is something that's been in development for about a good 6, 7 years at Micron behind the scenes.
What this is an array of DRAM memory arrays coupled with processing elements in situ to make a very powerful pattern recognition engine. What that is capable of is you would take this in certain applications, load it up with a large pattern that you would like to match against, then run real-time data through this engine looking for matches. That ends up being a pretty powerful capability that takes certain computing problems, things in vision recognition issues, network virus detection, bioinformatics, for instance. It takes the problem and rather having it be an exponential problem as those problems get larger, it makes it a linear problem. This is getting a ton of interest. I should caution, this is not a 2014 revenue kind of product. This is a long-term play.
We think one of the most exciting developments out there, showing how memory is moving up in the system, what it can really do. This will be present in high-performance computing systems. You talk about a lot of the problems that are being sent to the cloud today. It is all around something called unstructured data. Frankly, to sort through that kind of data, you need a pattern recognition engine like this. Something we are pretty excited about. Now moving over to the NAND side. The five-year CAGR here continues to just grow by leaps and bounds, 38% Micron view, and you see how that breaks down. Really, the key areas of interest here, obviously on the enterprise SSD side, we see a very strong move to PCIe interfaces. Frankly, that is something Micron has been working on as well. You will see more on that coming up.
Just an interface that gets the most out of the NAND and makes sure that in enterprise SSD applications, the goodness of the NAND and the component technologies we have been working on comes through. Client SSD attach rates driven by cost reductions on 20-nanometer NAND as well as 16-nanometer NAND continue to grow. We heard from a number of OEMs at CES this year that attach rates are moving up quickly as we go through 2014. Some OEMs talking about attach rates as high as 25%, even 30%. Just, we think, a relatively conservative view here on client SSDs, something that will be a big volume driver going forward. Then obviously, on the mobile side, specifically in handsets and tablets, everything happening with eMMC and soon a UFS architecture to really drive the next-generation memory interfaces.
This is a breakdown, same as we showed on the DRAM side, fiscal Q1 2014, Micron's position in NAND. Just a bit over 40% of our capacity working its way into system solutions, and by that, we mean solid-state drives. This is SSDs moving through Micron channels, through our own consumer channels, for instance. Obviously, a mobile piece that you will see continue to grow. A nice piece of embedded that Tom Eby is growing in ESG, driven by the needs of automotive and Tele nav systems, for instance. Then, obviously with our Lexar brand, and something we call Consumer Products Group, removable media and other doing well. In terms of the SSD portfolio, very broad portfolio out there. A lot of different needs.
Working from the bottom up, what you really see are the personal storage drives, driven by the traditional SATA interface, generally focused on making sure that client applications have enough persistent storage for the applications at hand. Moving upwards in terms of cloud storage, obviously mission-critical. Then you get to these I/O accelerators, which frankly is where the value of NAND and PCIe in general really comes through. To that point, Micron, in production today with a PCIe drive. This was actually an internal controller using obviously our own NAND, an internally developed controller, something called a P420, and it's out there doing very well today. The reviews are coming in. The graph of that, just bear with me for a second here. What this is in an enterprise application using an SQL database.
The kind of database format that's really used in several corporate and data center applications today is a good measurement. There's an open standard measurement tool set called Sysbench. What this is an hour's worth of test points across a variety of applications, measuring the different competitors across the X-axis and transactions per second on the Y-axis. The key thing is making sure not only that the performance level is high, getting the most out of the NAND, but probably most critically, that it's consistent. The way that you get consistency, frankly, is by being the builder of the NAND component, as well as the controller.
Trying to get the goodness out of both the NAND and the controller, working with each other to make sure that not only is the drive a high-performance drive, but making sure that that high performance comes through, no matter how the SQL database hits it. You can see here just really a leadership position, and this has been validated by a number of the test benches out there. Overall NAND focus going forward. We are believers, as both Mark and Mark said, and certainly Scott, the foundation of that strategy is really around good component technology, certainly a leadership position in 16 nanometer, getting the most out of planar and a transition to vertical here. Controller side, as Mark indicated, we do have internal bandwidth. We are always making sure that we use that internal bandwidth where it can do the most good.
We don't feel like we have to do everything internally, there are several areas where for fast time to market, for critical IT, you will see us using our base of internal controller designers and making sure we're in a good leadership position there. Then using external controllers for the follow-on product. It's a pretty diverse space here. Then moving through the drives, then obviously the software pieces as we start seeing really a bigger push by ourselves, working directly with the true end customers on the needs of their application, be it application-specific software, be it around virtualization, several areas here, that we believe software really will be the next piece of the value-add equation for enterprise SSDs in particular. Then, as you saw with Micron and our purchase of Virtensys, working our optimized drives into the appliance space as well.
Pretty exciting place here. This is a long-term battle, more to play through. You will see Micron keeping a close eye out there, partnerships, investments, whatever we need to do, both organically and inorganically, to make sure that from that foundation, from the base of the component itself, we are capturing the true value add of NAND inside of the system. With that, I'd like to go ahead and open it up to Q&A. Yeah, Doug.
Doug Freedman, RBC. If you could answer, what do you think Micron needs to do on the product portfolio side in terms of, what are some of the critical parameters or specifications that the products need to deliver in either NAND or DRAM, or both, to drive up the value chain? As investors, we sort of have a hard time telling what is the key spec or performance metric that we need to be paying attention to in each of those categories.
Great question, Doug. I'll answer that first looking at the DRAM side. I think the key things we're looking at from a value add perspective, it ultimately comes down to performance of the end system. Admittedly, that is getting a little bit tougher to measure because it's tough to separate, as it once was, the performance of just the memory from the actual end system. A lot of the goodness of HMC, a lot of the goodness of what we call in-package memory, which frankly, on the DRAM side is where we're putting a lot of resources. It's all about, certainly the end bandwidth, as well as the latency figures. How we measure that internally is our share of the wafers or the product space going into that. On the DRAM side, frankly, we think we've done a pretty good job.
We are excited taking that forward to 20 nanometer, and using the core capabilities that Scott talked about to continue to get the goodness out of the DRAM wafers. NAND side, frankly, I think we have more to do, and really, the key parameters on the NAND side, ultimately, what we're doing day in and day out, taking a look at the end applications out there, the real customers driving those end applications, and looking for a share of market in those opportunities, where ultimately from a true system performance view, the value of NAND comes through. The share of market we have in there will really be our indicator of how we're doing in that space.
Brian, you gave us some color on the 18 design wins around your second-generation HMC.
Sure.
Can you maybe give us some scope around the revenue opportunity and how you're feeling about the ROI on that product line?
Sure
Similar questions around DDR5 and also Automata.
Good question. HMC, as we've been saying for some time, first of all, for the big picture on the design wins, when we started the program several years ago, we really started it with a view around high-end computing, really driven more by the HPC, by the cloud opportunities. I'd say the pleasant surprise that happened over time was that it really ended up becoming the networking guys that came in with a view that we need HMC, and unless we have it, there's just no way to solve the problems that need to happen. For us, a lot of enabling horsepower has moved, or been added specifically, not just to target the HPC space, but really after networking in particular. I think as we've been saying for some time, we're not going to give any figures. 2014, there will be revenue on HMC.
In 2015, we think you'll start to see it as a needle mover. We're pleased with that. Automata Processor, frankly, it's further out. I'll tell you bluntly, we are as excited about that product line from a long-term view. When you're in the memory industry, you think in terms of the long term. These developments are not easy. They take some time. Something like the Automata Processor, it does take software behind it and really an additional layer of programming. By the way, we've invested a lot of resources to make that programming easy. There's something called an SDK, a Software Design Kit, that is available at the point of launch for this processor. A wide array of application-specific tools that go around that, helping us to try to get the ROI out of this earlier, frankly, by helping customers make it value-added in their system.
Again, I would not call that anything like a 2014 opportunity. Long-term opportunity, we think as we go through the applications that it can uniquely help to solve, we think is staggering. You have to be thinking kind of 10-year blocks for something like Automata. Okay? Yeah.
Brian, on the enterprise SSD side, I think two years back, you guys bought Virtensys, the flash appliance guys.
Yeah.
When do you start to see products roll out from that family, from the Virtensys, from the flash appliance side? Thanks.
Good question. Really looking at the Virtensys end applications, I think, really as we start to get into 2015, you'll start to see significant product opportunity with the kinds of converged PCIe switch that box enables, helping to really bring the network closer to the actual server CPU. That's what the Virtensys architecture was all about. As we look forward to 2015, we think that's where it really starts to fly. Yeah.
Hi, Brian. I have a question on the UltraDIMM product, where people are trying to put flash directly on the DIMM bus. First of all, how do you think that product stacks against your PCIe solution or any PCIe solution? Secondly, would you look to bring some product like that in the market?
I'm sorry, Monica, I missed that. PCIe versus-
UltraDIMM.
Versus UltraDIMM.
Yes.
Good question. In the NAND space right now, there's just this explosion of different architectures, different ways to try to figure out how to use NAND. We're believers that the best places to put it from a true storage standpoint, really as you're looking at real-time caching of the data, is on the PCIe bus itself. Then for that system checkpointing that we described earlier, that's where NVDIMM comes into play. NVDIMM and UltraDIMM are really two different ways of solving things. UltraDIMM is really looking at the problem slightly differently. We think there could be some applications there, but generally, the way these systems work, what's been enabled off of the DIMM bus versus what's uniquely enabled off of the PCIe bus, that's where we're putting our focus. Questions? Okay.
With that, I'm going to go ahead and turn it over to Mike Rayfield, We'll go from there.
Mobile's going to bat cleanup today. I'm going to talk a little bit about a handful of things. We're going to talk about how we view the mobile market, because it's pretty rapidly evolving, and I think how you view it'll give you an idea of how we behave. I'm going to talk about what's really driving performance requirements in mobile, because they're changing pretty dramatically. I'll talk about the opportunity itself and why, at least I think, a lot of us are under-calling it. It'll give you an idea of how we look at the market. Talk about where we're focusing at Micron in mobile, Then give ourselves an update on the progress we're making.
I guess we met six months ago or so in New York, I was talking about the smartphone market being a billion unit a year opportunity that was just the start of mobile computing. A bunch of you afterwards came and said, "A billion units a year is never the start of anything, right? It's either the middle or the end, but those kinds of numbers can't be the start." I think we've started to see in mobile how it affects markets that are adjacent to smartphones, if you will. Ultimately, it ends up being a much larger opportunity. There was an article that came out earlier this week. It was Walt Mossberg's new site, the headline was, "Apple, the largest personal computer supplier in the world now." That gets you to open it, right?
The reality, in order to get to that, all you do is you say tablets are personal computers. If you take it the next step, you say, ultimately, that smartphones are personal computers. You start to all of a sudden have a whole new set of opportunities opens up, they're all based on the same architecture. If you go and say, I've got automotive, which uses exactly a mobile architecture, ultimately, I'm going to have televisions that use a mobile architecture, it becomes many, many billions of units, it also leverages all the innovation, all the work we're doing on mobile right now. Where there's that amount of computing, there's always going to be more memory. What happens is, we've got all of these different form factors.
We've got our smartphone, we've got our tablet, we've got what's in our car, we expect the same experience from every one of those. What that says, your low-cost devices have to be able to do the same thing that your high-end devices do, or else you're just not going to buy them. The ability to do that says that low cost all of a sudden has significantly higher functionality than any of us had factored in before. The last adjacent market, you think about taking your smartphone, taking the glass off, taking the plastic off, putting an edge connector, you've got a server. Now I've got the technology we developed now in mobile that spans all those marketplaces ultimately crosses over into the cloud. It's a pretty exciting space. What's driving the additional computational requirements in mobile?
We talk about gaming and video, things like that. You've heard about 4K video. It was only four or five years ago, I was walking around telling people that 720p was going to be important on phones, and they said, for the 100th time, I was crazy. The reality is 4K video is going to come to your phone. It's not because you're going to have full-length movies. A full-length movie uncompressed would be 500 GB. You are going to want to take small video clips at 4K. Blow them up, cut it out. I still have a 1080p video that I can go off and show people because I've been able to zoom in and have very high quality. Things like that are going to start pushing more and more memory content, whether it be NAND or DRAM.
Larger displays, the innovation is going on in terms of how pleasing the devices are to work with how do you get that? You get a better user interface, you get more layers in the user interface, it operates more rapidly. That tends to drive more memory in all these devices. If you've seen the specials on people carrying their smartphones over to the Olympics in Russia, they turn them on, an hour later, they've been hacked and everything on the smartphone is gone. The reality is now these things have so much processing capability and there's so much content in them, we have to put the same kinds of security things that we put on our PCs and our desktop computers. That's going to drive more performance as well as more memory.
Finally, all the things I do on my mobile device are replicated many times in the cloud. I take pictures all over the world. I store them on Dropbox. My wife would rather see them on SmugMug. They get stored on SmugMug. The whole phone's backed up somewhere else. That opens up another interesting adjacent opportunity. As we put storage in the cloud, the more successful we are in the mobile business, the more business we create for ourselves in the cloud. In terms of the market, there's no argument about it's big, right. I think the debate comes down to where are the different segments of it. A couple of years ago, people said, are people going to want to carry a mobile phone, use it as a computer? I've already got two screens. I got a television, I've got a computer.
That debate's pretty much over. The devices are accessible. People own two, three, four devices. They're comfortable with that. That drives pretty significant growth. People talk about the high end of the market is slowing down, and therefore, that's where all the memory usage is. The mid-range and low end is what's speeding up. If you remember, when we talked six months ago, we talked about Xiaomi had a device that was a two-gigabyte phone for $130 unsubsidized in China. That's a new point that didn't exist. You look more recently, even at companies like Motorola, right. The Moto X is a two-gigabyte phone. It's about $320 unsubsidized. You think about Lenovo gets behind that brand, pushes that level of functionality at that price point. That's half of what that functionality was available for only a couple of months ago.
What happens is when we talk about, well, only the mid-range is growing, that's great. The mid-range or mid-priced devices and low-priced devices now have the same amount and more functionality than the $700 devices had only a couple of months ago. It's going to keep moving in that direction. I think if you just keep watching those devices, look at the graph on the bottom, it shows entry-level devices not going above a gigabyte until, what, the year 2017. It's racing upon us right now, right. If you start to bury that yellow bar just a little bit and it starts to approach what the high-end phones and tablets are, automatically the market becomes significantly larger than any of us have modeled. I think that's a pretty exciting thing to watch happen. Let's talk about the opportunity.
People will come and say, mobile's great, it's big. It's about to turn into the PC market. It's just going to be commodity. You're going to go down the same path that ultimately PCs went. I don't believe that's the case. There are a couple of reasons, right. We focus on two things. We focus on LPDRAM, and we focus on managed memory or managed NAND. We're not trying to own the whole market. We're going to pick those pieces of the market that make sense from a margin standpoint. The reason those things are stickier or make more sense is they're harder to do. When you go and get qualified on LP3 at a company, you go to the SoC or the chipset supplier, you go work on getting that qualified, making sure the system works optimally so their end customers get in production.
You can't just yank that part and put another one in. The reason is, first thing, it was hard to get to that point. Second thing, when you do that, you got to go back to the carrier and get qualified again, and a carrier qualification takes some number of months, and you've all of a sudden missed the window. The onus is on us to be a better supplier, more reliable supplier, if we take that responsibility to get that design in. It is stickier. It is worth more to people, and that's why I believe that it's going to have a different attribute than historically the PC market has had. When we talked last time, I talked about the focus being integration of Elpida and managed NAND. We now call it LPDRAM and managed NAND because Elpida's been integrated.
Our low-power DRAM strategy is the Elpida product, and what we've been able to do with that is they had a relatively narrow customer base. We now take that and spread it across a much wider customer base that Micron had, and it allows us to spread it across geographies and equipments, different customers, and I think create a much more stable and growing business than the two companies could have had separately. The other thing we're doing is I've been told a number of times in the last two days that we're behind on mobile memory, mobile NAND, eMMC. That's right. We're making good progress catching up, and the way we catch up is, again, we bring some amount of value to the customers that somebody else doesn't. What that means is we use our unique NAND technology. We add our own firmware.
We work with third-party controller companies to modify their hardware, and ultimately, at the high end, we'll have our own hardware. We get in with the customers, we get in with the OEMs and the chipset suppliers early and go off and make sure that we bring them something that somebody else can't, and a lot of that is in firmware. Finally, I grew up in the SoC world. I spend a lot of time with those folks, and the reason is they're the only people that really know 100% what memory interface is going to ship in three years because they're designing it right now. If I work in collaboration with them, I get my memory to work very closely with their new interface. In three years' time, all that work I've done just turns into design wins.
That coupling allows us to be a much better partner to theirs and allows the OEMs to get to production much quicker. An example of that, back just before Christmas, we sampled LPDDR4, the first in the industry to sample to our chipset customers and to the OEMs. Why that's important, LPDDR4 is higher bandwidth than existing PC DRAM now. Extremely high bandwidth, extremely high performance, extremely difficult to build, and extremely difficult to get integrated into the SoC. Because it's high performance, you want to get the most out of it at the lowest power.
We're doing that integration right now, our partners are bringing up their SoCs with our parts, our partners are bringing up their systems with that, and in a year and a half or 2 years when that ramps to production, our goal is we are the best coupled of those, and we have the best solution to get to market quickly, and it's all based on the work we're doing now. That's, again, another reason this is much stickier than it's historically been is because it's just plain hard and we've got to work together earlier. Then we talked about firmware. This is the place we differentiate and we catch up in managed NAND, and you'll start to see the progress on that going forward.
When I look at how do I measure progress, I talk to customers every day, and the conversations start to shift from, how many DRAMs can I have or how much NAND can I have to, I've got a problem, can you help me solve it? We solve that through early working with the chipset partners, through early working with the OEMs, and for solving some of these issues that they've got in firmware. As long as the conversation keeps moving in that direction, I think mobile continues to be very differentiated. It continues to be stickier in a design win business. Ultimately, I think that's the best way we bring value to the company. With that, happy to answer any questions. There's a microphone somewhere. There we go.
Mike, it's pretty well known that a very large mobile customer had a contract with Elpida that expired at the end of last year. Can you give some color around how you expect that relationship to progress, maybe if a new contract is in place, and does that relationship move forward into mobile DDR4?
Without talking about specific customers, Elpida had great relationships with a number of customers. We've got great relationships with a number of customers. Our goal is fundamentally to add those 2 together and continue to have great relationships with all those customers. It's a pretty concentrated business, and I think that the big customers and the big suppliers learn a lot from each other, and there's no reason why we shouldn't just continue to have great relationships with all the folks that Elpida had them with.
Can you talk a little bit about 64-bit going into handsets and what you expect that might do for mobile RAM requirements? Is that a driver? Is screen form factor a bigger driver? Then with 4K, did you mention when you thought we'd start to see that in high-end handsets?
On 64-bit, the simple tactical challenge in the short term is it's dual-channel memory, I've got to make sure for 3 gig, I've got 2 one and a halfs, and for 2 gig, I've got 2 ones. I think the real interesting thing with 64-bit is it ultimately gets a lot more capability. The GPUs on 64-bit systems are going to be significantly higher performance, and living in the GPU world for a while, it drags a huge amount of data. You need a lot of data, you need it very quickly. I think what it will do is drive significantly higher performance and tighter coupling with the memory, which is all good. It requires a lot more work to get it done. It requires LPDDR4, requires next generations.
With that, we'll get additional capabilities, whether it be graphics, whether it be better UIs, and it'll just continue to build and add, I believe, to the growing both NAND and DRAM need in phones. Wait, you had a second one.
4K.
4K. I think 6 months ago, had you said anybody would have 4K, they would've laughed you off the stage. It's raging up on us. I can't imagine that at Mobile World Congress, which is in what, 2 or 3 weeks, somebody isn't going to show something with 4K. Whether it's a prototype or production or whatever it might be, the people's insatiable appetite for great content, it won't stop. We'll just keep clicking along, I'm convinced. I'll buy one, of course.
Hey, Mike. We're seeing some pretty significant changes in the landscape in the handset market. You've got the big guys, Apple, Samsung, China Inc. seems to be on the rise. Can you talk about maybe some of the geographic challenges and how getting to be better with a broader base of customers is important to your business model?
Yeah. A couple of years ago, the mobile business was pretty concentrated. Two big guys and everybody else. It's still relatively concentrated, there's some shifts over the last little while that are pretty significant. China Inc., if you will, has got some people that are showing up as real technology leaders, not followers. We've got local resources, both technical and support that live in those regions that support it. Clearly, the Lenovo-Motorola deal, as that goes through, Motorola has a great relationship with carriers around the world. You imagine the might of Lenovo behind that. It could be pretty interesting as well. We could have additional big players. What we found, the cost of a design win business is you've got to win the design.
We're pretty aggressive at putting the design resources, support resources, right next to the customers to make sure that the designs come up quickly and they get to production quickly.
Talk about what you're seeing in terms of demand out of that marketplace in terms of are you shipping wafers to these customers, MCPs, and how important is the broadness of your portfolio?
Elpida had a great known good die business. Both for support of putting memory on top of processors and modems, as well as working with MCP and EMCP partners. We've very rapidly seen that range of the market move from NOR to MCP to EMCP, and it's been over the last three quarters very rapidly, and it will continue to move on to eMMC. The focus on firmware for EMCPs and eMMCs is, I think, going to pay off pretty quickly as those guys transition to that, because it wasn't long ago that that region was all about low end and cost, and now low cost happens to have a pretty significant amount of functionality, and they're going to have to go to a more aggressive memory strategy.
You touched on the opportunity with 64-bit. I'm wondering if you're expecting, based on the design work going on, that this upcoming 2014 holiday season could be a bigger opportunity for you in that area, or is it more of a 2015?
64-bit is here. It's not everywhere, but it's here. Once the first thing falls, it goes pretty quick. In 2014, there's going to be a lot of 64-bit solutions. Ultimately, that's going to be great for the consumers. It's snappier. You can do content that you couldn't have done before. It'll be richer as you put it on a large display like a tablet. It's going to be a lot better than the notebook computers most of you are using in terms of processing power right now. Great. Thank you very much. All right. Hopefully, by this point, we've covered most of the questions that you guys wanted to hear us talk about. Certainly, I'm here to try and wrap up any loose ends or cover any additional ground, and I'll rely on some of the folks that already spoke if I need to.
Let me just wrap up quickly by saying we think this business is in great shape on a go-forward basis. We think all the cogs of the wheel or all the legs on the stool are there, and we anticipate a favorable industry dynamic moving forward. We're building and continuing to build and improve on our balance sheet and planning our use of assets in a way that will not only build the company on a go-forward basis and achieve our operational goals and place our product portfolio where we need it to be long-term successful, but also deliver a great return for the shareholders. Part of that process is going to be continuing to invest in our business, whether that's in optimizing our supply chain or making sure we're deploying advanced technology to stay at the efficient operating frontier.
We're going to continue to focus internally on running our business cost-effectively. We're also going to continue to invest in the system-level solutions and differentiated products and make sure that as the future comes, we've got all the right memory to go with all the developing end applications. Finally, we're targeting the value-added segments. We're making sure that not only our manufacturing footprint is flexible, but we have all the right assets out in the field close to the customers to give us the opportunity to move our capacity and our output to the most value-added segments in the future. All that's being done with an eye to taking care of our shareholders. Let me stop and try and wrap up here with a few questions. We'll go from there.
Mark, there's talk of Hynix having some percentage of their capacity they can't migrate to the next-gen node. They've got to put in some new leading-edge, I guess, capacity. It's supposed to be capacity net neutral. Do you have a percentage of your capacity where you can say, "Hey-
This isn't going to be able to move forward. At some point, I don't know if it's two years down the road or three years, you need to put in a new facility like Hynix is doing. If so, what might that look like? What kind of capital raise, or not raise, but cost would that entail?
Okay. I think there's a couple of questions buried in there. The first one is, relative to Hynix and their need to add some cleaning space for the future, to me, that rings quite true and something that makes complete sense for them to do on a number of different planes. For Micron, we've been optimizing our capacity footprint over the last year. As you know, we've shed a lot of eight-inch assets, either sold them or diminished our ownership share or our obligations relative to offtake of capacity relative to some of those. We've consolidated our flash operations in Singapore. We're on the roadmap to finish that out moving forward. We have identified certain fabs where we're going to leave capacity behind.
For instance, in Virginia, where we anticipate less capital investment there to migrate that technology because the reality is we've got a lot of customers that need long-term commitments for the products. The more value add and the stickier these products become, the more the customers are interested in long-term support, and that makes financial sense to us, and we're going to do that. I think as we sit here today, most of the rest of our existing capacity is in pretty good shape, at least the 300-millimeter capacity, to continue to migrate. We do have some white space in Singapore to facilitate a 3D NAND transition, not the totality of it, but a significant piece of it. We're in pretty good shape there.
Having said all that, when we think about the future and the need for incremental capacity and how we might want that to come on board, at some point, it probably makes sense for Micron to have some additional clean room space. Not necessarily even increase the number of wafers that we put into the marketplace, but to make sure that we can continue to migrate because these technology nodes do become more complicated as we move through time. To the extent we decide we need that, we would look to do it on a cost-effective basis, but not necessarily with a mind of building out a big new fab. We would look to have clean room space available, as others in the industry are doing, to facilitate our transitions and to enable maybe small incremental additions of capacity on a go-forward basis.
Hopefully, I covered all the different angles in there.
Mark, could you maybe talk about the DRAM and NAND pricing trends, what you're seeing in the market? If you see weaknesses, any segment of the memory markets, steps you could take to stabilize that market?
Yeah. I don't want to get into sort of updating guidance or anything like that. I think you guys can go out and look at spot market trends. You're probably aware that spot market numbers for DRAM have been relatively flat. That's kind of as we thought it would be when we gave guidance. Without updating that, I'll just say that at least is still in alignment. Relative to NAND, there has been weakness in the spot market, as you guys can go out and look and see for yourselves. That's also kind of as we expected when we gave guidance back in January. The market is about how we expected it to be, at least as you look at what's going on in the spot market.
Relative to what Micron is going to do with its capacity, we're going to continue to work on those value-added segments. One of the things that hopefully came across today is that Micron structurally is different than it's been historically. We have a relatively low fixed cost in our overall operations. Part of that is by virtue of the way we've acquired low-cost assets. Part of it is that the capital intensity in our business has been decreasing. We look at our business today, we've got a pretty variable cost structure, and we've got lots of flexibility in terms of how we deal with capacity to have the right market environment out there.
Mark?
Yeah.
Mark, a couple questions. Where do you stand on kind of the vision next 3 to 5 years on 450-millimeter wafers? In addition, is there anything that you can do on the testing and packaging side to also offer cost improvements outside the standard stuff that we've been talking about in terms of capacity utilization? Thanks.
Certainly. On both of those, relative to 450, if that ever happens, which I think is in doubt at this point. We could talk about that if you want to ask a follow-up. I am not at all convinced the 450-millimeter will ever happen at this point. To the extent it does, it's a long way out in the future. I feel really good that our investment in 300-millimeter capacity is one that's got an enduring lifetime. There's not a lot of necessity for Micron, at least over the next 5 years, to be spending a lot of money on 450. Of course, it's something we're going to have to watch. There's a lot of investment that needs to go on in the equipment community to make that happen.
The value at the end of the day to the customers that would buy that equipment, I think, is dubious. We're not really focused on that. Sorry, the second piece of the question was? Oh, testing and packaging, yeah. Micron, I think we probably should put it in our technology section because there's a whole technology platform at Micron around test capability that ripples through all our various products. We design all our own testers, and we manufacture them in-house, and we enable our products to use those testers in clever and interesting ways that drive a lot of capital out of our backend structure as well. As we are migrating recently acquired capacity to advanced technology nodes, we're also embedding in those products some of the things that are needed to facilitate the Micron backend test strategy.
Yes, we anticipate significant cost reductions as that rolls out through the capacity that was recently acquired that didn't necessarily have that capability before. We actually started that as part of a JDP prior to closing with Elpida, many of the 25 nanometer products that we'll roll out here over the next year will be able to take advantage of that.
Mark?
Yes.
Hi. On the shareholder returns, Ron mentioned about wanting some operational flexibility and mentioned about potentially wanting to get to a net cash position. I wonder if you could just give us a little bit more guidance on how we should think about shareholder returns. Is it going to be something that might happen this year, or is it more like a 2015 type of story? Do you have in mind, like, a net debt zero or actually net cash position that you actually want to have?
Yeah. Obviously I don't want to box myself in relative to when that might happen. The first priority for us is to cover the operational needs and importantly, to make sure that we're making the right investments to position the company for a successful future. That's what comes first. We will, of course, continue to nibble away at the debt and try and do that in a way that is anti-dilutive. Think of us in terms of definitely continuing to take a look at our converts on a go-forward basis. Beyond that, I don't want to get into projecting exactly when we might take some other steps, because frankly, we don't know exactly what the market's going to look like. We don't know how much success we're going to have with our products.
We think it's going to be pretty positive, and we're going to have a good run here, but I want to maintain flexibility and make the investments we need to make to continue to grow the company, as well as, potentially return some stuff to shareholders beyond that, some dollars to shareholders.
Mark, a couple of questions. Historically, there's been some fungibility of capacity between DRAM and NAND.
Yep.
As you start moving below 20 nanometer for DRAM and start moving to 3D for NAND, can you just help us understand what happens to the fungibility?
Yep.
Does that help the supply situation? My second question, what sort of sustainable margins would you need to see before you went out and did a greenfield?
Relative to the capacity fungibility question, between technology nodes, that's going to get tougher and tougher because the technologies are going to continue to diverge. As we move to 3D NAND, as Scott talked about, there's a relatively large capital investment that's required to facilitate that transition. Those tools are not necessarily fungible over to the DRAM side. I think while you may see competitors in the space continue to optimize here while we're running significant amounts of planar NAND, that'll probably go away over time, and you won't see as much of that. Relative to the implications of that, I don't know that they're all that huge. I think the DRAM capacity that's in place looks like it's pretty appropriate for the market on a go-forward basis, and likewise on the NAND side.
I think there is certainly a faster growth trajectory in terms of how many bits are needed for storage in the world, that one may break down more quickly over time. As I look at the business, both DRAM and NAND today, I don't foresee that there's a need for a lot of additional capacity, and I think that it's going to stabilize in about the right spot. The second part of your question? What would the sustainable margin number be for Micron to say, "Hey, I need new capacity"? I think the best way to answer that is to say it's a big, big number. The reason is quite simple. When you think about adding new capacity to the marketplace, first of all, the new capacity, the greenfield capacity, is much more expensive.
You're buying all new tools as opposed to making small incremental investments to facilitate technology transitions, you've got a much larger fixed cost associated with that then has to depreciate over time. Your margin on that incremental capacity is lower. You've got to factor in, that's a small increment relative to your installed base, and if it's a significant increment, it's going to impact the ASP you get across the rest of your business. When I think about building a new fab, I just think it's a long time in the future because I've got to be sure I'm going to get a return on that incremental capacity over an extended period of time, and I got to be sure that I'm not building so much that I impact the rest of my business.
I think it's possible that people will look at what's the demand curve out there in the future, our margins getting so high that we're doing damage to our customers and pinching off end market demand. If margins got so high that were a problem, my response would be, I want to have small amounts of clean room available where I can add small amounts of incremental capacity, not build big greenfield fabs. That seems like the way we would approach the future, and that's certainly how I think about it today.
Hang on a sec.
Given the present state of NAND pricing in the first quarter being weaker than I think some might have expected, what would it take for you to slow down your NAND capacity expansion going on in Singapore?
Yeah. Again, we're not really interested in commenting on exactly what's going on there with that transition, and pegging any detail to it. I'll answer a bigger question, which is, how does Micron look at running its capacity full out or flat out in either segment, DRAM or NAND, versus market conditions where there's significant ASP erosion? I just said we've got a pretty variable cost structure right now. We'll look at our end markets, and we'll look at what we think the relative balance over time is, and we'll make adjustments as we see fit.
Mark, that sounds like a very different answer than we've heard for many years. Number one, it sounds like your variable component is much higher, and therefore, your willingness to pull back or not expand at the same rate that you had originally projected, that's changed. Is that a fair interpretation?
The industry is very different, Micron is very different than it's been historically. Yes.
Another question on your capital structure. If the world plays out the way I think you and everyone in this room hopes it does over the next two, three, four years, would you like to get to the point where Micron's capital structure, whatever debt you have, is all straight debt? Would you like to eliminate the converts totally over time, or do you envision converts always playing a role in the company's capital structure?
I think we're on the road in that direction. I think three or four years in this business is a long time, I don't know what other growth initiatives or opportunities or situations might be out there. I think the trend definitely for Micron is going to be more straight debt and less converts.
Mark Adams did a good job of laying out how you're striking strategic partnerships with a number of customers, it makes a lot of sense. There's two areas, though, where it seems slightly problematic to me, that's in the server market, where a lot of the big OEMs are under pressure from the cloud vendors who seem to care about low cost more than anything else. Also in the mobile category, where the best growth is coming from those emerging markets, which are also very cost sensitive. Can you talk about how you strike partnerships in those areas?
I think any large markets, they're always going to be cost sensitive. For us to have partnerships in those areas, we've got to be delivering real value that helps the customer at the end of the day. That's how we approach it. As you mentioned, the server business is kind of diverging a little bit. There's sort of mission-critical, high-value servers, and there's more and more some of these data center servers that are just moving data around the internet, but they are more cost sensitive, and there is less value that can be supported through what has historically been the value proposition in servers, which is quality and bit error rates and fit rates over time, et cetera. That doesn't mean there aren't opportunities to do things that are unique for those data center customers that can add value for them.
That's how we approach those kinds of customers is, what do you need in your business where we can add differentiated value? So far, with a number of them, we've been able to find opportunities. I think over time, in the mobile space, we'll see similar things like that where we can potentially help them drop memory tiers out of the overall system functionality or where we can bring differentiated power performance by getting more closely integrated with their end systems that maybe help them strip costs out of other parts of the BOM. We just look for how do we get the memory more closely integrated into the end system in a way that delivers value and then try and share that value with the partner. Maybe we take one or two more. One more? Okay.
Hey, Mark. You talked a lot about focusing on solutions for customers. At some point, can you move the model more to a backlog model, like a lot of other companies in the sector?
Yep. It's a great question, and I can't tell you. I think the answer is yes, but I can't tell you when, because there's a long history in this business that I think Ron showed you about. People or customers remember that and are nervous about it. I think on the one hand, they would like to have long-term contracts with more built-in stability. On the other hand, they're nervous that they miss on an opportunity to take advantage of a misbalance in supply and demand at some future point in the business. Structuring those deals is something that we're certainly interested in doing, and that we are open to doing and are talking with customers about. There's a history on our side as well, which says you better hold the money because those things are tough to enforce.
I think it's a matter of time and relationship building before we can get to that end state. I think it is a trend that you will see more of that in the memory industry over time. All right. I want to thank you all again for coming. Hopefully, we answered a lot of questions, and we certainly tried to cover a lot of ground. It's great to see a lot of old friends here again, and glad we're able to have all the momentum we have in the marketplace today. Thanks for coming.