All right, thanks everybody for coming today. Really appreciate you coming out. Packed house, which is great. I'm just kicking it off real quick and give you the format for the day. We obviously are pretty short on time. We've got 2 hours, so we're going to try to really focus on the keys that we think investors are most interested right now. Actually, we're doing it in basically a fireside chat Q&A sort of format based on questions that we received from everyone here in the audience. We solicited feedback for questions on what is really important to you as investors, as analysts, and that's the basis for the presentation today. We hope you enjoy it.
We're going to have Mark Durcan first, then Mark Adams and Ron Foster, then we'll also add a couple of our business units with Mike Rayfield and Brian Shirley, and then Scott DeBoer, VP of R&D as well. That's the flow for the day, Mark, come up.
All right. Thanks, Ivan, thank you everyone for coming today. It's kind of an exciting morning for us, ringing the bell and looking at our pictures. Now we're getting down to work, I think we've got a great presentation for you. I want to kick it off here. I'll skip through that. Just talk a little bit about the value proposition to Micron of this Elpida transaction we just closed. First of all, I'm not going to steal Ron's thunder, I'll tell you right now that the economics are very compelling. In my mind, that was always going to be true. Obviously, as things have changed over the last year and we've ground through the process of bringing this thing to a close, the situation has changed.
I think you'll really get to see some numbers today that drive home what a significant and beneficial acquisition this was to Micron from a pure asset purchase perspective. Now, that's not the primary reason we did the deal, though. There are a number of different factors that play in here. The product portfolio itself is very complementary and synergistic, gives us a huge leg up in the mobile market with their mobile DRAM portfolio, but also a lot of strength in graphics. Those happen to be two places where Micron historically has not focused. You put that together with the things that we've been working on in enterprise networking, computing DRAM, automotive, industrial, et cetera, and it's really very, very complementary. The R&D resources, the technology resources that come with Elpida are not to be underestimated.
They have really a world-class team in terms of technology development. Our plan is to leverage that for continued DRAM development and take more of our resources and deploy that to some of the other things we want to work on. We've got a lot of great DRAM technology. We've been working on a joint development basis with them for over nine months now on the 25-nanometer node and for the last three or four months on 20-nanometer, and it's amazing what we're learning from each other. We're already seeing the benefits of that, and we've got that technology path moving forward quickly.
There's going to be significant synergies as we bring these companies together, not only from the product portfolio, one plus one leading to more than two, but also from the activities that we're doing when we start thinking about all the infrastructure that goes into running the company, how we can take advantage of a combined supply chain, a combined procurement activity, eliminate some of the overlap in common designs for the more commoditized products where we're doing multiple designs that are actually the same thing. We're going to recognize significant synergies. One we've talked about in the past, backend technology, not to be underestimated, the importance of that and getting that optimized for the products we want to build in the future, also going to be significant for Micron. Finally, we now have the scale that our customers have wanted us to have for a long time.
The customers out there are rooting for Micron. We're the world's only global memory company with smart folks interacting with customers around the world, really understanding their applications. One of the impediments we've had to date has been we haven't had the scale for the customers to really bet on us to be able to deliver not only niche products that are differentiated, but to really work with them on their high-volume applications and add value at the system level with those advanced products. We have that now, and it is no longer an issue for us. You put the whole package together, and it's really going to be very compelling for Micron. If you think about how we're going to execute on this thing, we've had a lot of history in growth through acquisition.
The timeline on the bottom there, you can see how we move through time adding assets around the world, from TI back in 1998 to more recently Toshiba, followed by Numonyx. Most recently, obviously, the Elpida transaction that just closed. As we did that, we've added scale efficiently without adding capacity to the marketplace. This is just the last step in that transition. What you see from Micron now in terms of capacity is really significant growth over the last year without impacting market capacity. Really distributed globally in a fashion that we think makes the most sense and is very advantageous for us on a go-forward basis. Ivan said we were going to try and answer some of the questions that were submitted. I'm not going to read the questions.
You can read them up front here, and those of you on the web can see them. I think the theme of the first question here is this industry changing, and how is behavior going to change as a result of that? What I would say to you is this industry has already changed, and it has changed for good. There are a lot of reasons for that. First of all, I would say there's been a lot of growth in the memory industry. There's been a lot of fantastic innovation from a technology perspective that's driven significant productivity increases in terms of transistors and bits per square inch of silicon over the last couple of decades.
That trend is playing itself out, I'll show you the data here in a little bit in terms of the addition of bits from technology migration that's been going on in both NAND and DRAM over recent years, and you can see the slowing of that. That has a fundamental implication on the nature of the memory industry going forward, which is that we will not have unrestrained growth of supply as competitors in the marketplace struggle to adopt the leading-edge technology in order to remain cost-competitive. Yes, we'll still have technology migration, we'll have introduction of new technologies, we'll have innovation that moves the ball forward. It's not an all-out economic decision anymore that one must move aggressively to the new technology node in order to stay competitive. It'll play differently in the broad base of applications that we service.
That's the second fundamental change. We no longer ship all our bits into the compute segment, and even the mobile segment is, while growing rapidly, is not a majority of where the bits go, and we'll show you that data here in a little bit. Very broad swath of end applications now being served, which tends to dampen the volatility and increase our ability to differentiate our product portfolio, which begins the process of de-commoditizing the industry, really, I think, will lead to a totally different market dynamic. The industry is consolidated, it's been painful. As we look back through history, we've seen a lot of government industrial policy where governments have been active in the market, disrupting what would be a normal market environment through the support of excess capacity in the marketplace. That trend has played out. We're not going to see that anymore.
For the first time, really, in the industry now, we have rational competitors looking at what is the best way to deal with their capacity in order to optimize the value that they're delivering to the customers. Our focus, really for the first time, is how do we tune our technology development to deliver value-added systems to the end customer, memory systems. What that means is more of our R&D investment is tuned not to advancing process technology nodes, but to building the infrastructure that supports those advanced memory systems that are much more tuned to the end application they're going into. That's a much less capital-intensive business. It drives a different capital spend for the industry on a go-forward basis. How do we think about optimizing our returns for the future? It's no longer about market share.
It's no longer about driving the most advanced technology to full volume production as quickly as possible. It's about delivering value-added systems to the end customer. That's a very different state. Historically, of course, return on invested capital in the memory industry has been bad. Even as compared to the rest of the semiconductor industry here, you can see a pretty good dislocation in terms of what's the average return through the cycle. You can see it's been pretty volatile. Big swings from the top of the cycle to the bottom of the cycle. For all those reasons I just talked about, that's not the case going forward. This industry has fundamentally changed. How are we going to run our business?
We're going to run our business by making sure that when we make investments, they're predicated on a decision process that evaluates what is the long-term return on that investment through the cycle of the capital that's invested. Are we really getting the full return? We're not investing for the peak of the cycle. We're not investing in order to stay in the game, so to speak. We're going to be investing to make sure we have the capacity in place for all those applications we're delivering. That may mean that as we migrate technology, it's a smaller slice of the total capacity that we move in order to support a particular application.
There's another trend that's going on, and there's been a lot of discussion, and one of the questions or themes of questions that came to us is, well, the equipment folks are telling us you guys are going to spend a lot more money on capital. The reason you're going to do that is because the capital's just a lot less productive. I'd ask you, does that really make sense? If something becomes less productive in terms of what it can deliver to you, is your natural reaction, well, I'll just go buy more of that? That's certainly not how we're thinking about our business.
It is true that as we move through time, we are going to have a harder and harder time scaling, and the equipment supplier is going to have a harder and harder time delivering equipment to us that will facilitate those transitions, which means one of two things. That we behave irrationally and pour more and more money into a hole, or that we change our behavior. We're changing our behavior. I think that's what rational folks will do. I think what you'll see as we move through time is that you won't see as many new capacity additions because it's inefficient, and you'll see a slowing in the rate of new technology adoption, not only because the applications will demand a slower fuse in order to adopt new technologies, but also because that's what makes economic sense. How's that going to impact future supply?
I'll get to that in a second, but first I want to show you that the change has already occurred. If you look at what's gone on relative to capital spending in the DRAM industry for four years in a row now, there's a significant lack, or really almost a complete lack of capital spending for new capacity. It's all going into technology migration, and I'll show you in a second that that spending that's going into technology migration is actually producing a slower and slower growth rate in terms of bits supplied into the marketplace. You might say, well, that's going to be a problem.
In my mind, it's a fantastic boon because we're in an environment now where the demand for memory across all the various applications is really exploding, and we've got all the business unit guys here to tell you about the exciting products that we're working on. On the NAND side, you can see a similar trend. On the NAND graph here, we've actually taken a little bit of capital that may get spent in 2014 for new capacity and moved it into 2015. What's going on in the NAND business today is people are building greenfield clean room floor space, not large-scale new capacity, but the spend's going to go on in 2014 in order to facilitate a 3D NAND transition that'll happen slowly in the years beyond that. Scott will talk to you more about how that all plays out.
The capital spend you see going on here in 2014 for NAND capacity, a piece of that obviously is Micron as we transition a fab from DRAM to NAND. A lot of it is clean room space and early pilot line for 3D NAND. Here's the supply situation. Even NAND, which I would argue is the most rapidly growing and the most elastic piece of the memory market, has seen a long-term trend now to reduction in capacity addition, and a long-term trend in the reduction of bit supply to the marketplace as a result of those capital spends. I would argue this doesn't reverse. There's not some miracle that's going to happen that's all of a sudden going to change this dynamic. This dynamic is actually continuing with the same driving forces behind it as you've seen over the recent years.
A supply picture that is relatively muted relative to the growth of demand, and again, we'll have some folks come up here later and show you what the demand picture looks like. On the DRAM side, again, capacity actually coming offline over the last four years. Part of that's Micron's fab again. Generally speaking, you've seen a long-term trend in terms of supply to the market from technology migration continuing to scale down as the process just becomes more difficult. Where we sit now, 2014 is going to be in the low 20s, and the long-term trend's probably under 20. We have to wait and see how that plays out, but generally speaking, these dynamics are in place, and they're here to stay for a while.
Given all of that, many of you are taking a look at Micron and saying, "This is a different company. If we're really going to have a situation where it's not quite as capital-intensive, what are you going to do with all that capital? What's your approach to how you spend it?" I want to let Ron come up and frame up the situation with Elpida and give you guys a better sense of where we're going. It may be a good question to ask a little later in the presentation, too, if you want additional detail. I think you should think about our thought process on a go-forward basis relative to allocation of capital is. We've got some debt we need to pay down. It's very manageable, but we want to get that back into a more normal range.
We've got some investments that we want to make prudently in building our system-level businesses, which are very muted in terms of their capital demand compared to what we experienced historically relative to new equipment. Beyond that, we've got money available to return to the shareholders. What we've done historically has been first to go after the converts. I think that's something that is a strategy that makes good sense for us. Beyond that, there are other ways we can think about using that capital, hopefully it'll be, at least as long as I'm CEO, it will be focused on our shareholders. I want to stop there and obviously, I'll be around on stage here for quite a while, actually, and available for Q&A as we get some more information out in front of you.
Can I invite? Are we going to bring some chairs up, Ivan? Yep. Let's see if we can have Mark Adams and Ron Foster come up and join us. I think we're going to go to Ron Foster next because many of you want to see what he's got to say about Elpida.
Ron, can I give you the clicker? Sure. Where do you want me? Why don't you go here, and I'll sit between you. Okay, all set. Good morning. I'm going to start with a question that, judging from the requests we got in the conversations last night, there may not be a whole lot of interest in, I thought I'd cover it anyway. It's a little hard to see. Let me reorganize things. Just sit over here. You just have the presenter over there. Let me give you a little perspective on Elpida financials. As you know, we closed the transaction July 31st. We got our first look at the balance sheet, there's several steps we have to go through.
One is we have about 7 to 8 days after close to complete the pro forma financials and historical financials for Elpida. That's about a mid-October timeframe. We're beginning our work on that to get those financials ready to go. We also are started on our purchase accounting activities as of close to figure out what the purchase accounting effects are for Elpida. What I do want to do is give you some perspective as best I can on current run rates. I can address most straightforwardly those items which are not materially affected by the GAAP adjustments or the purchase accounting, but I'll also try to give you some perspective on that in a bit. These are per-quarter run rates, first on the P&L.
These are examples of their current performance, if you will, not represented necessarily of any particular timeframe, but roughly what they're running right now. That's the way to think about it. We still have to do all of our work to get their financials into our GAAP statements, but this gives you a view of it. On the revenue side, they're running about $1 billion-$1.1 billion per quarter. The cash gross margin, and I'm giving you that, one, because it's important to our business, and two, because depreciation is harder to calculate, but I'll comment on that in a bit, 47%-53% kind of range in current performance. OpEx is running 11%-13% on the Elpida financials currently.
One important thing to keep in mind is that our total JPY-based cost structure with the Elpida acquisition, with a 1% change, which is approximately a one JPY change today, will move our costs $5 million-$7 million in a quarter. That gives you some frame as you're looking at what's happened to the exchange rates going forward to our total cost structure. Obviously, that affects both COGS and OpEx. The operating cash flow at their current run rates are in the range of $500 million-$550 million. Very good performance on behalf of Elpida, and you'll hear more about that with some of the following speakers. Balance sheet, a couple of pieces I can give you with some degree of accuracy. One, cash, was probably the most accurate estimate I can provide.
It's about $1 billion that we are picking up on the closing balance sheet. The debt is about $1.3 billion. This is the estimated GAAP number after our initial $600 million purchase payment, which pays down some of the debt. About $1.3 billion. In terms of purchase accounting, let me address the elements of that that I can tell you about. As I mentioned, we are just starting our work now, but I'm going to try to give you some ranging views of that and make sure we all understand what it's going to look like when we report our financials at the end of this fiscal quarter, which ends at the end of August, or the end of this fiscal year, for that matter, which ends at the end of August.
Looking at some key items that are affected by purchase accounting on the left, I'll walk through them. I've got the estimated impact with each one ranging and then additional comments. On the inventory side, let me be clear about purchase accounting so we're all on the same page. We are required to value everything on the balance sheet at fair value, assets and liabilities. That's the fundamental requirement, is a fair value measurement of everything that goes on those financial statements. If you look at inventory, it will be booked to fair value for everything that was on the books on the balance sheet as of the date of close. That means that any value that was in the inventory at that time will flow through our financial statements effectively at zero gross margin.
They have about a 3- to 4-month flow of inventory, so it's going to take 3 to 4 months for that to flow through the books. There's one month we'll be reporting the month of August for Elpida, so you can expect to see Elpida in the single-digit gross margin in terms of what's reported on our GAAP effects with purchase accounting. We expect the flow-through of this written-up inventory to normalize after Q1, after our fiscal Q1. On the depreciation side, depreciation per quarter, it's the most difficult thing to estimate right now prior to having all of our purchase accounting done. I've provided a fairly wide range here. It could be between the range of $15 million a quarter of depreciation expense for the acquired Elpida assets or up to $75 million a quarter.
It's affected by the valuation, the fair value exercise on the PP&E, and obviously, the PP&E relates to that number. We also have to estimate the depreciable life of all the assets, which we estimate will be somewhere between an average of 4 and 6 years. That work is also going on. To give you a ranging estimate, it's somewhere in that range. Taxes. We also have a significant effect on our taxes as a result of NOLs, the net operating loss carryforwards that Elpida has. Those have to be valued also at fair value. We will likely have a deferred tax asset that we will put up on our books for Elpida's NOLs. Because of that, our GAAP taxes will be running close to their statutory rate of close to 30%. The way to think about taxes is cash taxes.
Elpida's will be running with significantly low single-digit tax rates for some time because of the NOLs. To give you another benchmark there in the comment section, Micron's total cash taxes, we estimate for fiscal year 2014 to be in the range of $20 million to $40 million for the year. We have a very efficient tax structure, and Elpida comes in with a nice NOL structure, but the GAAP number will be higher because of the DTA we will put up in purchase accounting. In terms of acquisition intangibles and goodwill, any impacts here will be minimal, not material to the discussion as near as we can tell. To summarize the debt situation, the face value of debt that's on the closing balance sheet for Elpida includes an initial payment of $600 million that we will pay down with our initial purchase price in October.
It will be in our year-end close numbers. Future installment payments of $1.4 billion. These are the scheduled payments for the Elpida acquisition. Other acquired debt from Elpida and Rexchip of about $300 million. It's about $2.3 billion. $600 million of it will be paid with our closing acquisition purchase. Of that $2.3 billion, the book value or GAAP value of it's going to be about $1.9 billion. The reason for that is the future installment payments of $1.4 billion, the book value will be about $1 billion. The reason is that they're at zero interest, so we impute interest at our weighted cost of debt capital rate. That flows through in GAAP interest expense, which shows on here, we're estimating it'll be about $15 million-$25 million per quarter, but less than $5 million of that is really cash taxes.
It'll be imputed interest from our purchase accounting. Just so you understand that. For total Micron, we already report about $28 million of non-cash interest related to our converts. When you look at our interest expense line, the vast majority of our interest expense is actually non-cash. In terms of gain on acquisition, we have to go through all our purchase accounting work to figure this out. A number of circumstances have changed since our last published pro forma financials on Elpida back in February, when we did our financing. Obviously, the business is performing better. FX rates have moved significantly favorably for us, and Elpida has significantly more cash on their balance sheet, as you saw, $1 billion. When you add all that up, we have to go back and reassess if there's a gain on the purchase.
If there is a gain, it will be a one-time, non-operating gain in our fiscal fourth quarter that we're currently in. More to come on that on earnings release, which will be in early October. We'll obviously break all this out for you and give you the proper pro forma so you can understand the flow of the business. In terms of Micron and the acquisition, what's our long-term capital structure and financing strategy, and do we need to raise capital? Let me first address capital requirements as an important variable, obviously, when you're thinking about capital structure. This graph shows CapEx dollars per wafer out per year on the dark blue line, and the dotted line is just the trend line of the dark blue line. That's the CapEx dollars per wafer out. The bars behind there is our average wafer capacity per calendar year.
Obviously, we've been increasing our wafer capacity, as you can see, over time, but you can also see on the blue line that our dollars per wafer out of CapEx has been declining over the last decade. The peaks that show here in our history were when we were building out new greenfield capacity in NAND flash at IM Flash Technologies in Utah and IM Flash Singapore. If you take that out, the CapEx run rates are fundamentally lower. As Mark already commented, we don't have any plans for capacity expansion. You can see the trend line going significantly lower in calendar year 2013. As you may recall, in our last earnings conference call, I commented that we were deferring out some capital out of fiscal year 2013 in anticipation of the integration with Elpida.
We're actually projecting to come in just below the bottom end of our guidance at about $1.5 billion of CapEx in fiscal 2013. Some of that's slipping into fiscal 2014, you can see on the line on the curve here that we expect the capital intensity to be in the same general range as we've been running in recent years. We're guiding our 2014 CapEx to be about $2.6 billion-$3.2 billion in fiscal year 2014. Let me turn to capital structure now that you see the view of sort of what our capital requirements are. I'll comment on three areas. First, liquidity, and then debt management and debt to capital ratio. Our fiscal third quarter close for Micron, we report about $2.9 billion in cash.
For your information, this excludes about $400 million in undrawn revolvers that are available to us if we need the liquidity. That's our starting point pre-close. We are targeting to keep a near-term cash balance comfortably exceeding $2 billion as we go through our integration of Elpida over the next year and a half or so. That's the range we plan to stay in. On debt management, we are always focusing on staggering debt maturities and avoiding any large maturities as we go through there. On debt to capital, we're targeting a 20%-25% debt to capital range. We were 31% in the third quarter, and we are going to be in the mid-to-high 30s as we bring in the debt I just showed you from Elpida.
Commenting on each one of these, on the liquidity side, there's been no adverse impact on our liquidity of the Elpida acquisition. We received about $1 billion on the balance sheet of cash, and the purchase price of Elpida plus Rexchip was about $1 billion. Net neutral on that. In terms of debt management, there's no adverse impact on our debt structure over time because we've staggered the zero interest payments over the next six years. We're a $1 billion per year amortization, or less, and less in most years, as you look at our debt maturities. On debt to capital, there will be a short-term increase at the mid-to-high 30s as we bring that debt on. We are expecting, given our current performance and our debt amortization schedule, that we will be back to our target range within two years.
To answer the last question, any plans to raise capital? It is clear from our perspective, given the current status of our integrated balance sheet, the business performance, that we're in good shape, there's no need to raise any significant amounts of new capital. We have no plans to do so. We will continue, as we always do, to engage in asset-backed financing as we amortize off our existing asset-backed financing and roll into new ones. That will continue. As Mark commented in his early comments, just to summarize, as we go forward in terms of our capital structure planning and use of cash, our first priority will be de-leveraging the balance sheet, debt repayment. The second priority will be managing our dilution, most notably with our convertible bonds. With that, I'll turn it over to Mark.
Thanks, Ron. The first question I wanted to address this morning was around the integration, how do we navigate that, and more specifically, around technology transitions. I'm not going to get into a lot of detail on the technology roadmaps. I'm going to leave that to Scott. When you think about technology and the Elpida business case for us way back at the beginning of the process, Micron's portfolio at the time, at a rough run rate about $8 billion-$8.5 billion, was very similar to our largest competitor at the time, was probably in the range of twice as big as us with twice in R&D budget. With the Elpida acquisition on the DRAM side, we've now got an integrated team, that Scott will discuss, able to advance us further and at a faster pace into 25-nanometer volume manufacturing, as well as accelerate our development of 20-nanometer technology.
On the technology side, integration's pretty strong. One thing we don't talk a lot about in terms of beyond the economics and beyond the technology is, over an 18-month process on the integration side, we've got to know these people fairly well. It's a good group. It's a solid organization of people and talent, and we're excited to have this team on board at Micron. When you look at R&D specifically, you've got a core group that is right inside of their Hiroshima facility on the DRAM side, as well as pockets of engineering around the globe, around Rexchip. From an R&D perspective, very solid integrated approach for us in DRAM, and we're pretty pleased there. From a manufacturing perspective, it actually aligns pretty well with our overall manufacturing network.
If you think of some of our announcements in the last year, you've got now Singapore positioned as a non-volatile geographic location for us in manufacturing. You've got Taiwan, in terms of high-volume compute and some mobile in Taiwan. You've got high-volume mobile in Hiroshima. Our network is pretty trued up, and it integrated pretty well into Micron as well. On the BU structure, if you think about Micron's current structure on BU, you've got a DRAM business and a NAND business. You've got a wireless business, and you've got an embedded business. The best way to think about how that will fold into Micron is to think about a mobile business integrating, merged in with Micron's mobile business. Mike Rayfield will be up shortly to talk about the scale and the opportunity there.
You got a pretty sizable business for us of Elpida's folding their business unit into Micron's mobile business for the new mobile WSG group at Micron. The compute business at Elpida will fold nicely under the BU structure that Brian Shirley runs under DSG. From a backend perspective, Elpida Akita was a pretty strong packaging group that's supporting the mobile portfolio, and that will continue to do so under the supply chain and operations and packaging group that we have at Micron that works under Brian Shields. I think the message here is that their organization fits pretty well into Micron. The timing that we've had, yeah, it was a longer process. The timing of integration allowed us to get a much further head start on what it will take day one to bring these two companies together. We're pretty optimistic there.
Lastly, I'll just comment on the customer side. Elpida's business was primarily mobile and computing. On top of that, what we know is it was, relative to Micron's business, a pretty concentrated set of business. We do business with all of their customers, and the overwhelming support there from the customer base has been very positive. From an account standpoint, we're just going through the process of identifying where the relationships are stronger and how we'll serve the customer. Again, we've gotten a lot of support from our customers on this and feel pretty good. The net of the integration discussion is we're pretty excited about it and pretty comfortable that the preparation in the last 18 months has positioned us well for a smooth integration. Next question was around current supply and demand environment, both DRAM and NAND. I'll start off with NAND.
You can see the numbers over the cycle here over the last about five years or so, you can see what the industry's forecasting, somewhere in the mid to low 40s. Micron is going to grow slightly above the industry growth projections for 2014. A lot of that has to do with what we announced in our last earnings call with the conversion in Singapore. Quite frankly, we see a pretty robust demand picture to support this. Primary categories of growth demand obviously are SSDs, both in the client side and the enterprise side, as well as mobile DRAM. We're also seeing strong growth in smaller categories, but higher value categories, things like automotive. Good business for us, a very strong business in terms of market share and starting to use both DRAM and NAND. In this case, the growth for the end market is pretty good.
Tablets is strong. We remain very bullish on NAND, we think the NAND supply and demand are pretty well aligned as best we can see. On the DRAM side, again, lower bit growth, you can see the trend. We're looking at a kind of low 20% range on the bit growth, we're going to be slightly below that, obviously, with conversion in Singapore. Interestingly enough, we think the balance still remains pretty strong here. You can see some diversification of the application market. If you go back to 2006, you see how significant the growth rates were in terms of PC and what have you.
Now you've got areas like server and even mobile DRAM that are driving not only growth, but as Mark talked about in his opening, differentiation for us and how we go to market with packaged solutions that get us away from the commodity business and more of a unique customer engagement model where we're delivering products that are proprietary to their own platforms. That's kind of where we wanted to take this business over the last three to five years. There's been a couple different ways we've talked about capital intensity in the business, I just wanted to talk about it relative to sales and kind of history in the business. If you look back over time, Ron alluded to it back in 2007, as far as the flash manufacturing and our entrance into flash and what that did to this ratio of CapEx versus sales.
For all the reasons Mark talked about in terms of lithography and maturing of the business and evolving to more of a systems play with less capital-intensive investments. As well as, let's not forget, over the last 18 months, we've increased our total capacity, roughly 90% of all existing capacity that was in the industry at phenomenal economic value. When you look at that takes our ratio here and makes it world-class relative to our competitors in any greenfield analysis you want to do. We came out very strong in terms of our capacity profile. We continue to look for opportunities to do that in the past. As Mark mentioned going forward, it's more about enabling this technology into more of a solutions orientation and differentiating our products within the segments we serve.
With that, I was going to hand it back to Mark to broach the one broader question on.
I think the question is, okay, what would change all of that? When would Micron step back and say, "Hey, you know what? This market looks pretty good, and there's robust demand, and let's go ahead and make some significant investments in order to grow our overall profitability"? I think you should think of us as reasonable, rational business people. Hopefully, we've demonstrated that over the last number of years. Our approach is really pretty simple. We need to have good visibility over the life cycle. I'm not going to give you a gross margin number because that's not the answer. The answer is we need to have good visibility that if we invest in capital, that is then water under the bridge, so to speak.
We need to make sure that we're going to get the return over the useful life of that equipment, which is growing. I have to have good visibility that I'm going to get a return on that invested capital that's significantly in excess of my WACC, my weighted average cost of capital. There's uncertainty, first of all. Secondly, just putting that additional capacity is going to impact all my existing assets.
When I look at a new investment, I got to have certainty of the return to some level, and I've got to have a return on invested capital analysis that says not only is this incremental investment positive to me from an economic profit perspective, but also it's significantly well above that threshold that I have confidence I'm not disrupting the rest of the assets I have working for me and for our shareholders in the marketplace. It's a very high hurdle that we have to get over. What other things come into our decision-making process, and in particular, would we ever look at the business more carefully in terms of the supply and demand balance and actually take capacity offline if the market warranted it, or run our manufacturing plants a little bit leaner, optimize in a different way? Well, of course.
Our job here on a go-forward basis, we have the manufacturing scale we need. Our job on a go-forward basis is to deliver value-added solutions to our customers. We don't want to be the biggest memory company. We've got the scale we need. We want to be the best memory company, and that means thinking about our assets and our whatever kind of asset that might be in a fundamentally different way, and that's how we intend to approach the business on a go-forward basis. Good. I think at this point, we want to stop and take some questions because we've covered a lot of ground, and I'm sure in particular, you've got questions for Mark and Ron, but I'm happy to answer any as well. We'll get a couple of shots at this as we go through the day.
Ivan will pass the mic, and you can address them to any of us or just throw it out there, and I'll try and direct the question.
Thanks. David Wong, Wells Fargo. Two questions. I guess they're primarily Ron questions. The first one is, you're coming in with a great asset purchase, but then you have to put in CapEx to upgrade capacity and so on. Does your cost per bit actually rise over the next few years because you're coming from a lower base and then you're having to invest? The second question, I think, Mark, you talked about extending life of equipment. Does your depreciation cycle stretch out because you expect to stay on node for a longer time?
Let me give you a high-level answer, maybe Ron wants to jump in with some more specifics. Absolutely, we are looking at what's the useful life of all the various components of our business, and that could stretch over time. I think that fits with the discussion I gave you earlier. In the short term, will our depreciation load on the Elpida assets go up? Absolutely. There's a high likelihood that as we continue to upgrade that equipment, the depreciation load will increase. We'll have to wait and see what the purchase price accounting sorts out as, but that non-cash expense will likely go up a little bit. In terms of the overall cost per bit, is that going to go up? No.
I don't foresee that happening given the ongoing muted rate of technology expansion or change as well as the ongoing depreciation of the existing assets. Ron, do you want to add anything to that?
I think you covered it pretty well, obviously we have to complete our purchase accounting, we believe we'll have a pretty low PP&E value. In fact, the numbers I showed you, the high end of that depreciation range is roughly the PP&E value that we used back in February. It's going to be at or below that kind of number in terms of PP&E, we think. We have to go through the work. The depreciation on the existing assets will be pretty low. As Mark mentioned, we got to layer in CapEx as we migrate, that is why I showed you that capital intensity graph, and it's not fundamentally changing as we go forward. It's more akin to our normal technology node migrations. We just have more capacity that we acquired at a very low base cost in terms of the depreciation.
The cost per bit should be declining, both for the reason I showed you of the trend over the last decade as well as the well-priced asset we brought in and our plans to convert, which are, as Mark mentioned, are pretty well architected already because we've been working on this for a while. We've been working with them for quite a few months. We have a fairly good view of that roadmap and what it takes to do the conversions, both on the technology side as well as the equipment side and the fabs. The team's done a great job.
We see a pretty consistent trend, and that's why we're communicating that message in terms of cost of CapEx per wafer trend lines continuing on the trend that we've been doing historically and in the recent past, and low-cost bit acquisition as base depreciation, and then adding in capital on top of that. That should do nothing but help our cost per bit.
Hi, Mark and Ron. Just a quick question. This is VJ from Sterne Agee. When you look at the whole CapEx for next year, how do you split it between NAND and DRAM?
Mark, do you want to take that one?
On the CapEx?
NAND, DRAM CapEx.
We're locking in our plans for our fiscal year 2014 as we speak here. My suspicion is it's probably in the two-thirds of that spend will be on DRAM conversion, I think closer to one-third will be in the NAND conversion, as best we see right now.
Got it. Just a housekeeping question, the $1.1 billion that you gave for Elpida, is that the May quarter or the July quarter?
You mean the $1.1 billion I mentioned?
Yeah.
Of PP&E? That was the base PP&E value that we produced in February when we rolled out the first set of financial statements for Elpida.
Got it. Thanks.
It will, yeah. It will actually be lower revenue.
What I'm saying is that it will be at that number or likely lower, given the depreciation scale I showed you of $15 million-$75 million a quarter. 75 would be more towards that end, 15, obviously, a lot lower.
Revenue. He's asking about revenue.
Oh. I'm sorry. What was that?
The revenue you showed. That's our current run rate.
Oh, I'm sorry. I thought you were talking about the comment I made about PP&E. I apologize. That is roughly the current range they're running. We haven't done their historical pro formas, and we also aren't giving you a forecast. I'm just saying, if you look at recent months, that's about what they're running.
Got it. Okay. If you go back to the NAND and DRAM forecast that you gave, any thoughts on how you see mobile DRAM within that? How do you see that growing next year for the industry or for Elpida?
I'm going to let Mike answer that because it's playing out as we speak for kind of our 2014 numbers.
Let me start again. The CapEx projection you provided, does that include some of the catch-up CapEx associated with Elpida Rexchip that were highlighted in the February filing, or is that a separate spending item?
The range that Ron gave you is an all-in number. That's what we expect to spend in fiscal 2013.
2014.
Sorry, in fiscal 2014 for Micron and Elpida.
That includes some deferrals out of 2013, as I mentioned. We actually came in below our guidance for 2013, and everything we're planning to do on 2014 to get the conversions done in that year.
Thanks for taking the question. James Schneider from Goldman Sachs. Just one question, first of all, on the capacity transition you talked about at the tech from DRAM to NAND. Can you tell us over roughly what time that's going to be completed? In other words, if we look out into, is it Q1, Q2 of next year, do you expect that transition to be complete? I guess the second question is, as you look at the rest of the factory network, are there other places like Manassas or elsewhere where you could contemplate some future transitions either way?
First of all, relative to the tech transition, that's obviously something that we're going to continue to monitor and dial with market conditions. We could complete that as quickly as late this calendar year if we so chose to. We're going to reserve the right to change and modify that as we move through time. Still heading in that general direction of a complete conversion or almost complete conversion at some point in the future. That's how we get the best capital asset utilization. Relative to would we go and do that one way or the other at some point in the future, I think absolutely that is a possibility for us.
I'll say also that even relative to compute DRAM and mobile DRAM, there are some slight differences in how we align our capacity, and we are planning to have flexibility in our network to dial that to optimize for market conditions as well. Because as you know, when you've got the growth that we've seen in low-power DRAM relative to what's going on in the PC segment, we want to make sure we get that balance appropriate. For the longer-term stuff, yes, I can see, depending on how the markets play out, that we might want to move capacity one way or the other. My guess would be that's more likely that we would move more to NAND versus DRAM over the long haul, but we don't have any plans to do that today.
We think we're getting it in about the right spot with the move we've got in play, and we'll take a look at it and see how it plays out.
Thanks for taking my question. Could you maybe talk about what are the DRAM pricing trends you're seeing in the market, and if you could address both PC DRAM side of the market and mobile DRAM side of the market?
Sorry, I didn't hear the second half of that.
Both in the PC DRAM side and the mobile DRAM side.
Sure. There's been a little softening in the DRAM market, but we think that's pretty contained to just excess capacity from some of the tier 2 players in the business who are looking to this to generate some cash and get through a technology process, a liquidation process of their inventory. We don't see that. By the way, our exposure in the spot market, which I'm referring to, is pretty light. Our OEM contract pricing has remained pretty strong and stable. We haven't seen any pressure on that. We feel pretty strong about our own performance in that, and we think there's just a little noise in the system. We don't think there's a lot of volume driving this, but of course, if there's small volume out there, that's the current price of the day.
At this point, we're not overly concerned that this is a trend. We think this is more of some inventory movement through the flow in the summertime here. As far as PC and mobile, of course, was pretty strong for most of our fiscal year and had a pretty sizable gap between where the PC business and mobile was. Mobile has been relatively stable during the second half of our fiscal year, yet PC has improved dramatically. I think part of that is because a lot of folks looked at the opportunity and directed some capacity towards mobile, and there's been some pretty good evening out of the capacity. Of course, the PC business, which on the demand side has been flat to less than flat, the alignment of supply and demand has been pretty good, the pricing improves.
It's Eileen Li for Glen Yeung at Citi. Can you sort of describe in more detail what's your end market exposure before and after Elpida? The follow-up is also on pricing. What is your sense of the overall pricing volatility after that industry change?
I have a slide a little bit later in the presentation to address your first question, as far as market exposure. When you think of Elpida, as I mentioned in my earlier comments, primarily think of Elpida as a mobile business and roughly around two-thirds mobile. If you think about the rest of that, it was kind of a commodity PC computing business. That's the opportunity for us, which is pretty exciting. We have the opportunity to take some of that capacity and drive it into Micron's profile, which I will show here shortly on the different segmentation.
When you ask about exposure, we're inheriting a business that was mobile and computing, we're going to have the opportunity to take some of that capacity and put it into some of these segments that we are market leaders in, the server, the networking, automotive, and those types. We're pretty positive there's good value there for us long-term in our transition there. The second question was, I think it was around pricing trends, wasn't it? Overall pricing trends. As I mentioned in the last question here, there's been a little softness on NAND, on DRAM, we think that's just kind of, again, low volume stuff that's out in the spot market, which we're not very exposed to as a company. On the NAND side, it's actually interesting.
On the higher density NAND side, it's pretty tight, that's going into the client SSDs and the enterprise SSDs. We've seen some softening on the lower density products that go into the cars and USB stuff and even some of the feature phone business. Again, overall in NAND, very pleased with where it is today, it'll continue to be pretty tight for us, given that a lot of our capacity is in the better segments, higher density segments.
Unfortunately, we're going to go ahead and stop the Q&A there. There will be more Q&A, though, later on. Hold on to your questions. I know there's still some more. We're going to bring up the other presenters now.
Over here. We'll go with the Mac. I got to come back. Oh, now we don't fall off when I turn.
I better not sit in front of the picture.
Yep. All right. Maybe before we kick off, I can introduce Brian Shirley here from our DRAM Solutions Group. Scott DeBoer runs our technology development, and Michael Rayfield from the Wireless Solutions Group.
I'm going to start off. Mark has covered a lot of this very well already on the synergy. The first question is around the benefits of Elpida from an R&D technology development overall point of view. Certainly, the big message is around scale. The two pieces that I would add to some of the things that have already been said today are really around our combination of Micron's previous expertise in mobile technology, Elpida's previous expertise in mobile technology and low-power technology, and the combination with the inherent high-quality products we have right now on the server side. When we put those two things together, the combination really makes us into a very strong technology company on the DRAM side.
A piece of the DRAM strategy that's fit together very well for us over this period of time is Micron's previous strategy, I've talked to that at previous meetings, really was to go straight to the 20-nanometer node. We've had our R&D resources focused very strongly on the 20-nanometer node over the last two and a half years, preparing for a strong position on that node. During that same period of time, Elpida had all of their technology resources primarily focused on a 25-nanometer node, which is now positioned and starting to ramp. The synergy of how the development was done, even pre-close, has been really good and it's fed into something that's going to put us in a strong position from a DRAM point of view. Mark and Mark talked about technology slowing.
The scale and the resources we're bringing in through this acquisition really position us to deal with a difficult technology scaling path on DRAM and a highly complex roadmap going forward to support the whole product base. Okay. The second question here is around what we think our roadmap position is across all of our products. Some of this has been alluded to. The 25-nanometer technology, which is now yielding well and ramping in Hiroshima and Rexchip, is the current leading-edge production on DRAM for us. In the second half of 2014, we'll be introducing the 20-nanometer node, which will be a joint product where we align our technology roadmaps between previous Elpida and Micron. We come together with one technology. I'm going to talk quite a bit of detail about NAND. I have a couple slides on this.
I know there's a lot of interest in where we're at on NAND technology. I'll just say right now, our leading technology that we announced about a week ago is on 16-nanometer. We view this as the most cost-effective NAND technology in the world right now. We're very pleased with where that positions us, both from a cost per bit point of view, and also from a performance point of view over the next year. Our 3D NAND, this is a big topic right now. Our technology on 3D NAND, this is a 256-gigabit density that we're going to be sampling in the first quarter of 2014. This technology, and the mixture of how it comes in relative to our 16-nanometer technology, is part of the story that I'll be explaining to you today.
Because of the strength in our planar NAND, we really can be very critical of where we introduce 3D NAND and make sure it comes in at a cost and a performance point of view that continues our industry-leading NAND technology position. Across our other technology, right now we have a solid position in NOR and in PCM with a 45 nm product. We're in development on a 20-nanometer node for PCM that'll roll out probably in 2015 timeframe. Okay. Now I'll go into some details really on our view of what 3D NAND means to the industry. Explain the strategy that we have for NAND technology that I think puts us in a position to be the leader in technology and performance through this period of time, really into the foreseeable future with the way we're going to scale.
To start with, at 20-nanometer, we made a technology decision that was unique in the industry to change the NAND cell and to position it for a high-performance process technology that would scale rapidly, and with good cost structure to the 16-nanometer node. Part of our strategy that's playing out right now is that this 16-nanometer node is going to be the most cost-effective technology in the market, in volume ramping up right now, and in volume through 2014. One of the big pieces about our differentiated planar technology is really the CapEx per wafer for this conversion, this structure. This is, I believe, unique in the industry from things that have been put out. We really positioned our 16-nanometer node to be exceptionally low in CapEx for this transition relative to all of our previous nodes.
This technology that enables it and this cost structure are something that we're very proud of, and think puts us in a very strong position over the next year. The other thing to look at on this graph, which tells part of the 3D NAND story, is the CapEx relative to transition a planar node to a 3D NAND node, and it tells you part of the story around how you make the decision on 3D NAND relative to cost structure, number of wafers you get out of the fab, and bits per wafer. I'll talk about that in more detail in the next slide. This is our non-volatile strategy. As we focus on 16-nanometer, as I've discussed, maturing right now. We have 3D NAND technology.
We've made a business decision that the right time to come in is at a cost per bit point of view where it makes financial sense. When you weigh the capital cost to transition the number of wafers you get out of a given fab on 3D versus planar NAND, and the performance, which we believe we have the highest performance 3D NAND solution available right now. We understand that the right strategy is to run planar NAND through a big part of 2014 and introduce this, and then ramp it up over probably a significant period of time, as Mark mentioned earlier, rather than a rapid switch to 3D NAND overall. Beyond that, we have a scaling path already well down the road on scaling to a higher density of vertical NAND. We believe this technology is the position for bulk storage.
It won't be touched by any other technology, in our view, for that SSD application. I'll talk a little bit about that on the next slide. We have a lot of focus on emerging memory for higher performance applications at a cost point, not quite to the NAND space, but significantly better than the DRAM space. We have major partnerships with strong partners, and we're putting a lot of focus on this next memory technology. We believe these two things coexist going forward in the future. Okay. The last topic here is around our emerging memory strategy. We've shown the stoplight chart in the past of all the memory technologies that exist in this space.
We've spent a lot of time understanding with our partners, with our global infrastructure, what the best technologies are that actually fit into good memory applications in the future. We've really narrowed it down to a few key partnerships and a few key technologies. What this graph shows really is the performance and the density between the different types of these memories that we're focused on this page. If you look at the far left and the far right, it shows our belief that the DRAM space really can't be touched from a pure speed point of view into the future relative to these other memory technologies. The far right of this is that NAND goes on, and we believe in our scaling path for NAND technology to maintain a position that really can't be touched by any other technology for that piece of the application.
The middle part, which sits in between the two in performance and density, is where we're putting a lot of focus on new memory technology, on 3D memory technology. We believe that's a space for the future that really enables some new kinds of memory system technology and new kinds of applications. One thing to highlight here, because we've talked about PCM, Micron has a lot in the past. PCM is the only emerging memory that actually emerged. It works. We have products on it, but it isn't our primary focus going forward, as it is something we're working on.
We believe PCM is a platform that we've gone through to enable system products to really understand the technology requirements of a new memory, and we think that just having the platform of PCM has moved us into a pretty strong position as we look at all these other kinds of memory technologies and evaluate which ones are going to work. That's the technology section. I don't know if we're doing questions on this right now or if we're going through all three of us. Okay.
Scott, just to be clear, you said your 16-nanometer process technology, you think it'll be the lowest cost. Is that for the segment you're targeting, you mean high performance?
No.
Even compared to three bit per cell?
Well, the 16-nanometer inherently can be MLC, SLC, or three bit per cell, so it's a technology platform. We believe that technology node of 16-nanometer, given the cost structure and the bits per wafer, will be the leading at each of those segments.
Okay. Thanks, Scott. Ed?
Quick question, how much development do you think you need for going to somewhere in the low teens, beyond 16-nanometer, in terms of development or capital thing, that you might say, "Maybe it's easier for me to switch to 3D?" Do you think that planar has still some more legs to go before you start to see 3D starting to happen?
Right now, the last products that we're developing on a planar roadmap are at 16, and from a pure technology assessment point of view, there's a capability cost and overall performance trade-off. We really think 16-nanometer is the last node that makes fundamental sense before the transition to 3D, but it's physically possible to build something smaller than 16.
This is Scott. Hi, it's John Pitzer with Credit Suisse. A lot of us is focused on 3D from a manufacturing yield ability standpoint. I'm kind of curious if you can give us some insights on the ecosystem around 3D NAND, specifically on the controller side, because clearly, as we think about how quickly 3D NAND can ramp into the supply environment, controllers are going to be pretty critical around that, and I've gotten mixed views when I've talked to people. Is there stable controller technology out there? Is that going to gate or slow down the adoption-
Sure
of NAND and 3D NAND in this marketplace, be helpful.
Brian, if you can.
Yeah, thanks. Brian Shirley here. I'm going to go ahead and field that one. Both in NAND and DRAM, we do see an increasing use of controllers coming, both for the interfaces as well as what we would call the media management. As we bring 3D into production, both in SSDs and mobile, as well as we bring new DRAM technologies into production, I think you're going to see areas where really understanding the physics and getting some layer of logic management of that memory taken care of really is critical. Frankly, to do that well, you have to be the memory company that understands the physics as well as the fabrication behind it. Yes, that will be a piece of the game going forward.
Srinivasan Sundararajan from Summit Research. The question that I have is on 3D NAND. What will you be doing to make it more ROI friendly in the next two years for 3D NAND?
From the point of view of what the cost structure is of 3D NAND, we're targeting a position, and our belief is it has to be above 30 tiers for it to be cost effective relative especially to our planar NAND solution. When you look at the overall cost structure of 3D NAND, you get to a scaling point where you get a lot of bit density increase per wafer, and the financials of it are you offset the extra cost to capital and what we call wafer trade ratio, which especially as you scale, can be anywhere from 1.5-2.0 on number of planar wafers you could get out of the same space as you get out of a 3D NAND process flow. Certainly, the bit density increase trade-off there gets you to where the ROI on it is very positive.
Yes. Do you have any products or packaging synergies between what you do in mobile DRAM speed and what you have in-house with NAND flash, either different types of packages or things relating to eMMC for smartphones and tablets?
One of the big strengths of Elpida's technology, especially in the mobile area, is their packaging technology, and it's definitely a synergy that we're really happy with. I think that like all the parts of R&D right now, there's definitely a blend of technology expertise that's going to make us stronger both on the NAND side and the DRAM side. Just, I think, a really good situation for us on the packaging technology side through the combination.
Let me add a little bit.
Something there. Clearly, for the types of products we want to deliver to our customers going forward, state-of-the-art packaging technology is a key enabler. We're very happy that, as Scott said, to have some of the expertise and capability that came with Elpida imported into our bag of tricks, so to speak. You will see that as we move forward and as we probably spend less money on advanced capital for bit, for square inch of silicon, you'll also see investment that goes to facilitate those system-level solutions and advanced packaging that's demanded in all the end applications we're targeting for the future.
Thank you. One follow-up question. Scott, did you say that you're going to sample 3D NAND Q1 of 2014?
Calendar Q1.
Then on DRAM, if you're going to be migrating to 20-nanometer next year, then what would happen to technology roadmap beyond 20? To follow up on that, would you need EUV to migrate below to 20? How does the cost benefit change?
Sure. There is a node below 20 that we're working on right now. It doesn't absolutely require EUV. It and some of our other technologies benefit from EUV above a certain cost point on the EUV infrastructure. When the wafer per hour of an EUV tool gets to a certain point, it's not necessarily for us a technology enabler, but it's absolutely from a cost point of view, the right thing to do.
If EUV is not available, would double or triple patterning have an adverse impact on the cost structure?
No. By default, if EUV comes out and can run 150 wafers an hour, it will be a benefit to our cost structure. Our plans right now don't assume that EUV is going to be available for 20-nanometer or whatever the 1X nanometer node is on DRAM. The other piece of that is the infrastructure in our fabs is facilitated for double and triple patterning already. Those things are built into our DRAM and our NAND fabs now, and they're already part of our cost structure.
There would be other types of cost savings to offset the increased capital intensity of double, triple patterning?
In my view at least, the cost for us, with the technology we use of double and triple patterning, is sometimes overemphasized in the press. I think we have a very cost-effective methodology for doing double patterning, and it's been in mass production for us, especially on the NAND side, for ever since the 50 nm direction. We're very experienced with it. The tool sets are very mature, and EUV is the thing with the hurdle to get over to be cost effective, not necessarily double patterning.
The different NAND manufacturers are all taking different strategies for the timing of 3D relative to planar, that migration. Which is the more important driver for each of the competitors? Is it their planar technology running out of gas, or is it their 3D technology being superior?
I think a big piece of it is really the quality of the planar technology. In the end, the timing for real volume manufacturing on 3D for the competitors may look very similar. The front end of this may look different because of the timing of or the quality of the planar NAND technology. Certainly if you have a strong position in planar NAND technology and your performance is meeting the spectrum of needs for your highest performance product, then the cost structure of planar NAND is superior through the next years.
Let me add a little on this 3D NAND topic. We've said we're going to sample early in 2014, others are talking about sampling at different points, as Scott has already pointed out, potentially for different applications. We don't believe that there's any significant impact on the NAND marketplace from 3D NAND in 2014 at all. We think it's a much slower boil in terms of when this becomes a volume manufacturing technology. There is socket enablement and work to be done with the customers to smooth the way, and that's why you see the sampling earlier in the year. Other things in there. I'm going to talk a little bit about the mobile business. The way I want to talk about it is, it was five years ago or so, we started talking about our phones becoming computers, and there were doubters.
They just assumed that they were sort of high-end calling devices. They have become computers. There's really no argument, both from a processing capability and what we try to go off and do with them. The diagram here a couple of years ago would have shown servers and workstations and maybe some laptops tied to the cloud to do processing. If you just take a look at some of the things we do with these mobile devices now, a couple of years ago, somebody came up with the idea of, well, upload a bunch of photos to the web, it'll go off and get crunched on a big machine, and you'll come back with a panoramic picture. Right now, we do that on our smartphones. We just go across like this. We get beautiful panoramic pictures, a huge amount of processing necessary, a huge amount of memory necessary.
Got a bunch of frame buffers in there, a whole bunch of crunching going on, ultimately, that's what has opened up this whole category we call mobile computers. As we talked about this last night at dinner, a bunch of you said, "Yeah, but that's only the high-end devices, and the high-end devices are slowing." Earlier this week, a bunch of you probably follow Xiaomi in China, the little smartphone startup that did $2 billion in the first half of this year. They just introduced a $130 phone. It's called the Redmi. It's a quad core, 720p phone, 4.7-inch display, one gigabyte of memory, and four gigabytes of NAND. That's the new definition of low-end phone.
All these models that talked about low-end phones and smartphones having a quarter or a half a gigabyte, at $130, that's squarely in the middle of the low-end smartphone category, and it belongs right here with these devices in terms of its processing capability and where it goes. When we talk about what part of the market's growing, the market, in terms of its capabilities, is crunching together pretty significantly. I think the other thing that we've seen is this architecture that is called smartphones is really the first instantiation of mobile computing. It goes everywhere. If you look into what's in your car entertainment system, what's in the back of your television, what's in your tablets, what's moving into ultra-thin, it's architecturally all the same.
What that means is that this little smartphone opportunity of 1 billion units is really just the tip of the spear on what mobile computing is, and we'll see it grow into an amazing amount of applications, some of which we'll talk about here, and some of which we probably can't even imagine. It really does tie our lifestyle together. I watch all you folks operating these things all the time, and it is where we share data. It's where we have an amazing amount of applications running in the background, and where we go off and actually edit data, which is something people didn't know we go off and do. There's a whole bunch of forces driving this computer and what it's capable of doing, as well as the amount of memory it needs.
The first piece is we talked about all of these mobile devices are architecturally about the same, and if you've got a tablet and a smartphone and you do one application on that tablet, you want to make sure that same application runs on your smartphone. I've got a 15-year-old daughter. When the smartphone runs out of either memory or processing power, it's ugly, right? You need to go off and make sure you've got enough to be able to do the same kinds of things you do on your tablet and your smartphone, and the same things happen in the back of your automobile. The next piece is all these soft services. We share and edit photos. We share and edit all kinds of content.
All of these things run in the background, and if you understand computer architecture, in order to run all these things in the background, they all have a carve-out of memory. It just keeps adding up. If I've got a BSP, I've got all the applications, and nobody wants to shut one down to bring another one up, and it just keeps inching up. We've already talked about the low-end smartphone being a gigabyte now. You saw the Moto X came out at two gigabytes. Samsung's new phones are two gigabytes. There's three gigabytes in larger phones in development, and it's all about the user experience. People want to make sure they can go off and do those same things. The carriers have gotten in the mix. They want you to run a lot of applications. It forces a lot of data.
You see T-Mobile very recently, no contracts. They want you to upgrade phones more rapidly. AT&T has come out with a plan every year you can upgrade. You start to put the forces of very affordable, high-performance devices with very rapid upgrades, you end up with a market that grows pretty dramatically in very high-performance devices. Finally, all these personal electronics, whether it is going to be a watch or glasses or a heart monitor or medical monitor, while those all have a relatively small amount of memory, all of those use the phone as the hub. Now I have just put a whole bunch of relatively critical applications that you are interested in running in the background on your mobile device. This thing is now, as opposed to the cloud being the center of the universe, what it is, it is the center of your universe.
The cool part about it is as you put more memory on that, it also forces more memory in the cloud, which is where Micron's got a huge opportunity. People want to back up to the cloud, but they do not want to be connected all the time. Connectivity is not ubiquitous yet. This allows us to go off and every time there is another device being sold, there is a huge amount of infrastructure that goes in place as well. In terms of just sort of a couple more pieces of data that shows where this is going, the amount of NAND in the smartphone in two years is doubling. It goes to over 20 gigabytes. The amount of tablet goes closer to tripling.
Again, to our earlier conversation, if you can do something in your tablet, you store your information in your tablet, you want to be able to do that same thing in your phone. I have all my books on all the phones and all the tablets. I get a little upset if I cannot fit them on. People go off and want to make sure that is a great experience. Cameras are another great example. When you start doing processing of images, you have multiple copies of the frame buffer. You need to go off and crunch on those and keep them in main memory. As camera resolutions go above 20 megapixels, you saw Nokia's what, the highest one is now 41. You put a couple 41-megapixel images in memory and start processing it is going to take a lot of memory.
If you just look at connectivity and video, LTE drives more memory. Combination initially was because it is a higher performance device. In the end, it is because it allows you, again, to do a lot more processing on images and video. Finally, three or four years ago, I used to walk around and tell people about 720p and 1080p videos on phones. They said, "You are crazy. There is no way everybody is going to want to do that." I do not know if any of you are fans of these GoPro devices. The latest GoPro device, when you take it out of the box, the default video setting is 15 frames a second, 4K. You shoot through 64 gigabytes of data in a very short period of time. That is the same things that is coming in all these mobile devices.
The memory is going to continue to grow pretty dramatically. Finally, what does that mean to us? Obviously, we're a memory supplier, but as we put Micron and Elpida together, and as you've heard earlier, the rate of progression in technology is slowing. It's all about having all the right pieces, NAND, NOR, DRAM controllers, firmware, packaging, and with the combined companies, we've got all those pieces to go off and create interesting solutions, energy-efficient, high-performance solutions for all of these mobile devices. I think that, as Brian had talked about earlier, the real innovation is going to be how do we put all these pieces together? How do we wrap firmware around it? How do we do the right controllers?
How do we control memory like we can better than anybody else, and deliver something to customers that lets them go off and make these mobile computers all that much more mobile and all that much more powerful? That's it.
I'm going to go ahead and finish up on the business unit side here. We're going to talk a bit about what we call the DRAM Solutions Group. Again, this is our computing DRAM portfolio that goes into servers, networking, ultra-thin gaming applications. We want to talk a little bit about how we see Elpida adding to these segments and some of the key trends happening in each of them. You've heard us talk about it before, server, storage, and networking. These have been some hallmark segments for Micron over the years. A lot of key products that we've developed, such as RLDRAM going into core networking. This is a segment that continues to grow leaps and bounds. It is insatiable demand out there, notably on the server side, really for a lot of the trends that Mike talked about.
You've all seen the figures about the number of cell phones driving new data center servers, as well as the networking gear behind it. Folks, we can't keep up with that demand. One of the things we're most excited about with Elpida is just simply additive scale. It taking us away from being in a supply-limited situation in servers to increase our share of market in a segment that, frankly, has been very good to us and that we see quite a bit more room to increase further share with scale. Moving down to ultra-thin PC and tablet. This is a changing market, frankly. As you think about the diversity of memory and the diversity of platforms coming in there, frankly, it's unlike the desktop notebook market of years past.
What you see these days with a variety of different SoC platforms is different memory requirements coming into that space. Really, a decommoditization happening in a segment that, frankly, is still fairly high volume. In this space, we have DDR3 coming into play, now even mobile technologies such as LPDDR3. Really, what we are finding with Elpida is just a perfect match, both with 25-nanometer technology in production at Hiroshima and Rexchip, as well as LPDDR3 capability for always-on, always-connected ultra-thin memory that's getting soldered down, really bringing into play much different customer-supplier relationships than we've had in years past. Again, helping to decommoditize this space. Finally, over on the graphics and consumer side, this is a big win for us with Elpida. Elpida has been a longtime player in the graphics market, something called GDDR memory, Graphics DDR.
Currently, they are in production with something called GDDR5. This is a product line that is perfectly complementary to our existing Bi16 portfolio for mid-range and low-end graphics. GDDR5 really giving us the product portfolio for the game console refresh happening now, as well as advanced graphics in laptops and desktops. Frankly, even moving beyond to some high-performance computing applications. Really, an area that we're pretty enthused about. Perfect complementary product portfolio to what we had already developed and the scale to execute there. Now, in the graphics market, in particular, again, GDDR5. This is a picture of the 4-gigabit GDDR5 Elpida had recently taped out. This is in production as we speak in Hiroshima. This will be used for a number of upper-end game consoles, as well as advanced computing graphics.
Again, one of the things we see here is an opportunity to leverage GDDR5 into upper-end server and high-performance computing applications. You don't hear much about that, but some of the absolute column performance that GDDR5 offers is actually making it the architecture of choice in a number of infrastructure markets. This, by the way, is in production on 25-nanometer technology today. Moving over to the infrastructure side. Again, growing by leaps and bounds. We can't keep server modules on the shelf right now. We have recently moved the product portfolio from what we would call 8-gigabyte registered DIMMs as really the sweet spot of the server space into what are now 16-gigabyte registered DIMMs. That speaks to the kind of content per box increases happening out there. You can see the chart on the upper right here talking about the server unit CAGR. Looking forward.
The really amazing thing, and we see this daily, is the content per box of any given server. Again, that's driven by the mobile trend. That's driven by virtualization. More cores per CPU and more CPUs per server, what it all needs is a lot of memory in that box. That, again, on the Elpida side, is really helping us out as we increase our scale, moving 25-nanometer eventually into server applications and helping fulfill the needs of our key customers. A lot of the market for Micron over the years past, we have generated good share of market in the corporate server, the high-performance computing side of the business. The data center growth has really been a huge volume adder here. The way those data center builds happen, though, is that the customers will come in, and it's a little bit of a clumpy business.
It's something that we call the LVO phenomena, large volume orders. The customers, and you have a good sense of who they are, the large data center providers, will come in and say, "If you can deliver X amount of memory to me next week, you can have 100% of this business." By the way, it may be dark for a couple of weeks after that or a couple of months until the next data center build comes along. The way you win that business is through scale. It sets up as something that turns into a very large percentage of the overall memory market going forward. De-commoditized, specialized server memory driving significant multi-digit percentages of the overall DRAM market.
From a product standpoint, we've situated ourselves beautifully here, both with standard server modules as well as optimized cloud products that really get to high-density DRAM densities. You get into RLDRAM 3 specifically as the architecture of choice today for upper-end routers and switches. Moving into HMC, which is winning the battle for the next generation high-performance computing and networking architectures out there. To that point, let me finalize here with just an update on Hybrid Memory Cube. What we did when we started this program roughly four years ago is we said, we want to gain the upper-end performance point in servers and networking with this architecture, but we also want to deliver it in a way that's easy to use, where the memory can be placed down on a motherboard, doesn't need a silicon interposer, it doesn't need fancy assembly technology.
We take care of all of that for the user and manage the memory. Frankly, this is working out better than we ever could have planned. It's gained the high ground, and it has given a number of server and networking companies an architecture that is ready to use in these upper-end systems. I'm happy to report that the production version is in silicon. We have sampled this outside of the company, and we are on track for a 2014 production launch. We're very excited about this, and it continues to win more design wins in each of these phases today. Now, the interesting thing is that in between standard DDR3, DDR4 memory, and HMC, what we're seeing out there daily is a massive number of custom opportunities coming in across the board. This is not just true in the server and networking space.
This is true in graphics. It's true, frankly, even in upper-end standard client machines. New ways of taking advantage of DRAM memory, that's really changing the business for us. 10 years ago, what we worried about in product development was just standard component designs, bringing in a lot of designers to cost optimize a design, fight through high volume, and really ultimate cost optimization. What I worry about these days, frankly, is bringing in more logic architects, more firmware architects, even software, crafting up, frankly, a much higher value add portfolio, something that competes on value as opposed to just a commoditized cost profile. A number of areas here that, frankly, we're pretty enthused about the memory industry going forward. I think it brings it back around to Mark's point from the beginning of what the R&D starts to transition like with this changing game of memory.
With that, I'm going to turn it back over to Mark Adams.
Thanks, Brian. When you take a look at the technology discussions we've had today, and what I wanted to do, comment a little bit about our business model, because the vision of differentiated product sounds great on PowerPoint, what have you, but from a vision of how we've gone forward, if you went back five years ago, you're going to see on this chart the changing diversification of our business. On the left-hand side, there's a technology discussion around our mix here with DRAM, NAND, and NOR, and other. If you take a look at what we've acquired with Elpida, obviously, nearly 100% DRAM. The combined last 12 months puts you in a ratio of around 61%, over 60% on DRAM, and 27% on NAND. The opportunity for us is to continue to play out this differentiation and multi-segment approach.
If you look at the end of the day, our job is to take this precious resource we call capacity and put it in the right homes. I mentioned earlier, over the last 18 months, we got a lot of capacity. I heard some questions last night about what would it take? Mark addressed the question here is, what would it take? That's not our mindset. Our mindset, really, going forward in our strategy is, how do we take our capacity and put it in the right higher value segments? Micron, if you went back again five years ago and looked at where we've come from, we've done a pretty good job. Personal computing was down under 20% on the native Micron, and now we take over an Elpida capacity that was much higher.
Again, it gives us the opportunity, as Brian was just talking about, to push that capacity to some better segments. Our focus going forward from an operating perspective and from a go-to-market perspective is How do we continue to optimize that? We're not thinking about additional capacity. We're thinking about existing capacity into higher margin segments. Quite frankly, the opportunity is there. Our customers want us to be bigger in server. Our customers are demanding us to be bigger in networking, automotive, and all these premium segments. It's an exciting time for us from a go-to-market standpoint to take that new capacity and continue to drive it into these differentiated markets and really give us some leverage in terms of our P&L and operating performance.
You might want to read a couple of those, Mark. I'm not sure they can all see.
Yeah, I'm sorry. We're blocking there. On the bottom there's personal systems. The personal systems on the Micron, native Micron, was through our last five years of going to market and product development. We got that down to a number of sub 20%. Then you've got mobile here, and the combined mobile and the new business will be about 25%. Then consumer server, networking storage, automotive, broadly referred to as AIM, because you have some industrial and medical. Then SSDs at the top, you can see, keep growing in materiality. The shift here is, as we balance out the equation on the left-hand side from a DRAM NAND technology mix, is how do we leverage that into the right markets? We're pretty comfortable on the scale piece. We've got a really good story there.
You'll see us continue to drive more in these higher value segments. You just use the left-hand side as a benchmark on how we've done from a diversification. The excitement we have around Elpida is we can take that capacity to do the same and get leverage in the business. Okay, thanks. I think we're going to go to a broader group Q&A.
Come on back, guys. Come on. We're down a little bit here. Why don't you stay there? There we go. Yes. Get in here. There we go.
Yes. John Pitzer again with Credit Suisse. Just two questions here, one for Ron and one for Mark Durcan. I guess, Ron, real quickly, I've heard the word scale multiple times in this presentation. I guess, just given your newfound size with Elpida, how do we think about your purchasing power, either with raw materials and/or capital equipment today versus what it has been? For Mark Durcan, Mark, you guys put out a lot of arguments as to why the industry is different today versus the last 15 to 20 years. One that I didn't hear was the fact that your largest competitor, Samsung, today, for the first time in their corporate history, also has a sizable market share as an OEM.
I'm wondering whether or not that's allowing them to view the component business differently, because clearly, stable to up pricing for them now is very profitable. It makes them more competitive in their handset business. How important is that dynamic to your new worldview on memory, and does it open up opportunities with competitors of Samsung to drive economics? Thanks.
Go ahead.
Well, John, you heard comments already in terms of relationships on the customer side that I think are going to be really helpful with scale. I think that base was covered pretty well. To your question on the procurement side, obviously, having more scale in the total system is going to be valuable to us, and we're already beginning to work in the last few days with Elpida. We haven't had a chance to really even see that information until we close the transaction. I think there'll be some opportunities there as we're moving ahead, especially on the capital side.
As to the question on Samsung being a large consumer of memory and having, obviously, a vested interest in maximizing their competitive advantage through how they treat their memory business, I think that's definitely a factor in the market today. What Samsung's corporate strategy is over the long haul, I would love to pontificate on, but I probably oughtn't to, because I don't actually know exactly what they're planning to do. It makes sense, and I think the way I look at this business is, I think rational people are rational, and they do what makes sense for their business. I think the dynamic you're talking about, it applies to Samsung, but I think it also applies, by the way, to SK Hynix now with SK, and maybe to a slightly different degree with one of the other big NAND manufacturers.
That is a dynamic in the marketplace. I think it's real, and it gives me more confidence that everyone's got the right approach to this business.
Right. Maybe I could just add one more comment. From a customer perspective, yes, there's the emotional, "Hey, they compete with us. We want to buy from other people," and that certainly helps. If you look at the raw availability of the output to the non-Samsung camp, the rest of the world, when Samsung takes that consumption internally, we're in pretty good shape to be the leading supplier to everyone else but Samsung, just in terms of the raw capacity. That manifests itself in some pretty good customer relationships and these opportunities of development and higher value opportunities at all these other customers. Just the raw capacity we have versus what Samsung has available outside of their company to other customers, it puts us in pretty good shape.
I'll add, since we're on a roll, I'll add one final point to that, which is Samsung is actually a pretty good customer.
Yeah.
We're happy to have them as a customer, they're very happy to have us as a supplier, because if you think about also from their perspective and rational behavior.
Playing out the way one would predict. From just a technology risk mitigation perspective, even if they think they've got a perfect roadmap internal to their own semiconductor operation, they've got to be concerned that there's a lot of bright guys here at Micron that are innovating and have a lot of customer engagement all over the world. To not be engaged with Micron for Samsung is a risk factor that just makes no sense. We actually have a pretty good business relationship with Samsung now, and we're happy to engage in that.
Hi, it's Sandeep Baliga from Jefferies. Thanks for taking my question. Just to follow up on the capital spending question that came up. I think you've made it pretty clear a couple of times that your capital intensity is trending down. It's clear now with scale that you have more power in terms of bargaining with your suppliers, specifically the equipment vendors. Given this, I guess, big picture, how do you think this impacts the equipment supplier industry longer term in terms of their ability to invest in technology development? Clearly, this comes at a time when scaling has become increasingly challenging. Any thoughts you might share on that?
Yes. I think it's an interesting challenge for them. I think, frankly, for both the suppliers and ourselves, the future lies in much closer collaboration in terms of picking the things we work on very strategically and doing them in a more deliberate way rather than maybe more of a spray shot approach. The investments required for a particular capability are probably larger and more complicated, the probability of success if you're not working very closely with a supplier and have buy-in from the beginning that this is something that's going to add value for them, you have a very diminished probability of success at the end of the day. I think, for both the suppliers and for ourselves, that close interaction and relationship is going to be critical to optimizing our outcomes. There was one up here.
Quick question. As the four NAND players embark on this 3D NAND, how should we be thinking about what kind of differentiation that all the four players will have against each other, if at all, there is a differentiation on the 3D NAND in terms of anything in terms of density or any other metric that you might be able to throw out so that we can better understand in terms of the differentiation? Thanks.
Look, maybe I can start, and Scott or Brian, maybe you guys want to jump in, or even Mike. I think fundamentally, the large NAND suppliers all have a slightly different technology or technical approach to 3D NAND. That underlies probably a slightly different strategy in terms of what the optimal timing and first application is. The primary driver behind that is what's the quality level of the data retention and endurance performance of the NAND technology you're deploying. That drives a thought process around, okay, what application makes sense, what kind of configuration, what kind of density, et cetera. Each competitor in the marketplace, I believe today, probably has a slightly different view of what makes sense for them.
The other thing is, I think Scott already pointed out, it kind of depends where you are in your planar technology as well, but less important than the sort of fundamental approach each competitor is taking to the 3D NAND technology itself. Brian or Scott, do you want to add anything?
No, I think you got it.
Just a quick question on the way you're going to manage fab utilization and just sort of historical aspect of the business has been the fabs were full until you kind of get to cash cost. Last year, kind of a watershed event when Toshiba took utilization down unilaterally in a not great environment, but not a horrible environment. How do you guys think about it, and what are the criteria you would use to determine when you'd run the fab less than full?
Yeah. I think it's completely rational behavior from Toshiba. My understanding, anyway, they had a lot of triple-level cell product going in the marketplace in that timeframe, and I think they just looked at it and behaved rationally to optimize their company independently. I think you could assume that Micron and other competitors in the marketplace will do the same thing. We have no interest in pumping extra memory out into the lowest value-add segment in order to simultaneously pump money off our balance sheet. That's just not a behavior that we're interested in engaging in.
Hi, this is Srini from Summit Research, and this is a question for Scott. Can you show the slide where you have the 20-nanometer and the 16-nanometer side by side?
We're going to have to get out of the way, though, guys. Try to get Luigi to back a little bit. Sorry, we're a little crowded in here today. This one?
Yeah.
Yeah.
You can clearly show here how your density is going to be much better compared to 20-nanometer and 16-nanometer. By using the high-K, you were able to compare the scale between.
Yeah.
You better repeat the question.
I'm sorry, I didn't hear his whole question.
My comment is, it's very easy to explain why your 16-nanometer is going to be superior and why 3D NAND is not necessary, just based on this picture.
Yeah, I think, as I mentioned, we did think out this strategy of how to go from 20 to 16 to 3D in great detail with a lot of thought put into this node. We definitely believe this is the most cost-effective direction for the next year and really puts us in a strong position as we transition into 3D NAND.
Let me add. I think the primary point for the benefit of others in the audience is, and we said previously, our 20-nanometer node here is what we call a planar cell. The control gate does not wrap around the floating gate, and he's noticing from the cross-section that in order to facilitate that, you have to have an advanced high-K gate dielectric. That's the approach we adopted at 20 nanometer, driving a low-cost scaling to 16 nanometer.
Roger.
Hey, guys. By buying Elpida, you obviously become a bigger DRAM company.
Right.
For those of us who have been around for 10 or 15 years, in the past, what we'd do in the morning is wake up, and the first thing we'd check is DRAMeXchange and look at the spot price. By that, you could almost gauge what Micron's doing, how well it's doing, and where the share price is going. Certainly, DRAM has been your biggest earner in operating income over the last couple of years. My question is, when we look at that DRAM spot price every day, does that number still have a big relevancy compared now than compared to, say, five or seven years ago? Is it still an important gauge? I know you don't sell into the spot, but certainly the contract follows the spot. Is that still an important number that we wake up and look at every morning?
My second question is, recently, we see the spot has gone below the contract price in DRAM, that's the first time it's happened in a few months. Do you expect contract to follow spot and go down near-term, or do you expect spot to stabilize? I mean, that's hard to predict, but do you expect contract at least to follow spot in the near term? Thank you.
Let me take the first part of the question, maybe I'll ask Mark to comment on how contract, how we think that's going to play out. I would say, for a lot of reasons, we tried to enumerate some of those for you here today, the DRAM business going forward is going to be very different from the DRAM businesses has in the past. I think it's going to be a pretty good business. When you look at, A, where does Micron's products go today? Very few of them, as you pointed out, go into the spot. The spot itself is also very thinly traded and is representative, not of the vast majority of the memory output today and the quality level that's required for a lot of those applications.
It's really representative of the tail end of the quality distribution that can perform in some of the lower value-add applications. I can't predict how you guys are going to trade our stock. I think if you're rational, you'll be less sensitive in aggregate as to what's going on in the spot market. Now, relative to how's the spot going to play with contract, et cetera, certainly we have to make rational decisions. The spot, to us, given everything I just said, does not drive how we approach contract. With that, I'll turn it over to Mark, and maybe he can speculate on where ASPs are going.
Yeah. Thank you. From a mix standpoint, one of the things we've been working on a lot over the last five-plus years is the diversification of the end segments. To address your question about the exposure to the, quote, "spot price" and the commodity type of the business, as you can see on the right-hand side, that first bar, the Micron last 12 months, you can see we've done a pretty good job on the diversification of the business. We believe that we have an outstanding opportunity to take Elpida's business and push it more to that Micron model around DRAM. From a diversification play, that long term, we don't think that index of DRAMeXchange on a daily basis, by the way, we look at it once in a while, but we don't think it's going to be as meaningful as it has in the past.
As far as OEM contract and their behavior versus the spot market, in the here and now, we don't think the linkage is very tight. I mean, we don't think the volume out in the spot market is significant. We think there is just some flowing through of excess inventory, and we don't see the pressure out in the contract. As I mentioned earlier, we're not selling a lot there, so our OEM contract pricing to where we are today has been pretty stable.
Hi, thank you. Historically, your shrewder customers have waited until the end of the quarter for you to make decisions regarding pricing.
Can you talk about that historical behavior and reconcile it with nodes living longer and slower shrinking, and tie together what's going on the technology side with how you're running the business and how you think your customers will or should behave going forward? Thank you.
Sure. One of the byproducts of where we have been pre-Elpida on capacity is we weren't even exposed to that behavior, although they tried, except for extreme conditions where there was oversupply. We have been able to navigate the end of quarter, end of month buying cycle, the pressure at the end. Given some of what Brian's team's been up on in terms of stimulating demand around server and network and what have you, our decision in that scenario is we normally don't try to push out revenue towards the end of the quarter just to do that. Again, part of the theme of the day is it's profit-driven and it's value-driven. So if our decision is do we push out DDR3 inventory just to hit a revenue number versus keeping that for linearity for the next quarter so we can ship servers and networking, it's a no-brainer.
We don't look at our business that way.
You're saying you'd be willing to hold more inventory as-
Well-
opposed to-
Yeah, assuming it economically makes sense. By the way, as we've talked about in our last couple earnings calls, over the last 18 months, our inventory is down about over 20%, because we're trying to manage it well, and so we're not sitting on a lot of pressure towards the end of a quarter. If we had to, we look at that as just good economics, good value. Rather than dish it to a spot market customer on the 31st of the month, it doesn't do anything for our business. We're so constrained right now in the server and networking business that it's clear what we're going to go do.
Roger Gill from Needham & Company. Question, now that you have both NAND and DRAM capability and capacity, how do you look at the trend towards multi-chip package solutions in mobile handsets? Where do you see that going in terms of adoption? How do you think you can exploit that internal capability over your competitor who has to partner with a third party?
As I talked about on the last slide, one of the things to do is have all the pieces. When you've got the NAND, you've got the DRAM, you've got the packaging technology, controller, and firmware, that's where the differentiation is going to happen. As phones and tablets get more and more memory, higher and higher performance, that's where we think we can add a lot of value. It's just the integration of all those pieces, doing from a firmware standpoint, as Brian highlighted, there's a piece of firmware to manage the memory itself. We're probably better at managing our memory than anybody else. A piece of firmware that looks externally where we can add value to whatever that application is. Huge piece of opportunity for us.
By the way, let me mention on the computing side as well, we are seeing mixed DRAM and NAND architectures come into play. As a matter of fact, we've launched a product line with a couple notable design wins, something called NVDIMM. These are standard server DRAM modules with our own NAND on the same module, a controller and firmware in place to manage, in essence, a check-pointing operation in the server where everything that's written into the DRAM can be placed directly into the NAND flash for backup, for checkpoints, for a variety of different applications. It helps having all of the technologies in the bag right now.
Thanks. Steve Fox with Cross Research. You guys have outlined a lot of synergies that are to come from the Elpida acquisition. Obviously, it's too early to put a dollar number on that, but could you maybe just qualify what's the most important synergies to watch for, the timeline that we could start seeing them maybe drop to the bottom line? Just one clarification, Ron. In looking at the purchase accounting for Elpida, you're saying the rest of this quarter and next quarter, you're going to be running Elpida through the non-GAAP numbers at a 0% margin before we see the real numbers from Elpida. Is that correct?
Let me take the first. I haven't actually tried to dollarize the impact of any of these things. I think the product portfolio/customer application diversification, immediate impact and very significant to us. The R&D scale capability and bandwidth to handle future challenges, a big win. The ability to tune our capacity, we started on that in advance of the close, important. Finally, all the various things that we can do as we run the company together in terms of marrying our product portfolios and in terms of getting more efficient relative to a company infrastructure, supply chain, IT, those types of things, probably put that at the bottom. There's plenty to be done there, too. Ron?
Yeah. In terms of the purchase accounting impact on inventory, we've got one month of Elpida we'll have in our fiscal fourth quarter. Anything that was on the balance sheet in inventory at that time, we marked to fair value, which means it will have a zero margin at the stage of completion it was at.
Any margin we get on Elpida business, on our GAAP financials in the fourth quarter, will be value add beyond the inventory that we're shipping that was already on the Elpida books. That's why I said it'll be low single digits impact in the fourth quarter in terms of the margin. I'll plan on pro forma all of this for you so you can understand the fundamental flow when we do our earnings release. They've got a 3- to 4-month inventory flow-through. That's going to wash through that original inventory at zero margin over 3 to 4 months. By the end of the fiscal first quarter, it'll be virtually completely behind us. We'll be reporting margins that are more consistent with what's really happening with the fundamentals.
Just wanted you to understand that because it will affect the numbers. We will plan on doing a pro forma of that and the other effects so you understand the underlying contribution.
Thanks. Last question on CapEx. Vijay from Sterne Agee. When you look at your CapEx for next year, $3 billion, not too high, and you had to push your own CapEx out to next year. Embedded in that, the Elpida portion of the CapEx, is that mostly flattish year-on-year? Any color on that? Thanks.
In terms of the Elpida CapEx trends?
Yeah. Year-on-year, as you look at fiscal 2014 or fiscal 2013 for Elpida.
2013, okay.
Thanks.
You may know from public communications and our earnings calls that we have been involved with the planning, including CapEx planning, and we had an operating plan that we put out there with Elpida when we signed up the sponsorship agreement. They've been working to that operating plan, including capital investments that go with that operating plan, pretty close. Their CapEx activities have been going on exactly as planned and consistent with the rollout of now their 25-nanometer DRAM, for example, that was commented on. As we go into 2014, it's just a continuation of that in terms of CapEx and the merged 20-nanometer DRAM that will be working together. All that's factored into the guidance I gave you of $2.6-$3.2 for 2014.
All right. I think with that, we'll say thank you all very much for coming. Appreciate your attendance and interest, and we'll get back to work. Thank you very much.
Thank you