All right. Good morning, everyone. Welcome to the Boise State University School of Business and Economics in the Micron Building. It's great to see you all here, isn't it? It's a fantastic facility. Hopefully, you enjoy it. They've got a great program here. We're very happy that Micron Foundation was able to help them put this building in place. Hopefully, over the years, as they continue to grow this program, we'll reap many rewards as a result of it. It's great to see so many of you here today. We know that Boise is not the center of the universe. We appreciate you coming out to join us to go through our fall analyst conference. We're going to try a new format, which hopefully is very interactive, not only between us and you, but also between ourselves. Relatively unscripted.
We have culled some general topics of questions from information you sent in in advance, that we want to try and address. We want to spend most of our time actually then responding to your response to that and responding to questions that you have in an interactive way to hopefully give you the best use of your time. As a result of that, we don't have a lot of slides we're going to go through with you today. We do have some that'll directionally help us with some of the points we want to make. As always, we've got lots of that material available for you, relative to what's going on in the marketplace and growth in our markets, et cetera. That's all available as always. Again, thanks for coming. Let me just give you a sense of how we want to proceed today.
I'll start out here just with a few thoughts and comments on the memory industry and what's changing and why it's such an exciting time for us and why we think there are great opportunities ahead. Also try and talk a little bit about what some of the challenges associated with that are. I know that there's a lot of interest in the room relative to what's going on with Elpida. Obviously, there are things I can talk about there and the things I can't talk about there. I'll try and be as responsive as I possibly can to your questions. As we go through those questions, by the way, I'll probably bring Mark Adams up in case he has some slightly different or additive views to my own.
We want to spend a little time, Mark and I, addressing questions you might have relative to my comments on the industry dynamic and Micron strategy moving forward. Jump into some market and product Q&A with the various business unit leaders. We'll have Scott DeBoer is here today to also contribute to a technology discussion along with the guys that are developing the products. Finally, Ron Foster is here to talk a little about finances, how we're thinking about our financial position, our balance sheet, and positioning the company from a financial perspective. As we go throughout that whole process, hopefully, we'll be getting a lot of input from you to help draw out from us the information that you need about Micron.
Without further ado, I really do believe it's a fantastic time to be in the memory industry, unlike any time in the previous 30 years, because the dynamics that are driving the industry now on a go-forward basis are so different than that which we've experienced. Many of you have been around the industry, as I have, for a long time, and you're well aware of the technology treadmill that we all participate in, driven by our ability to add more bits per square inch of silicon, more efficiency, more value through scaling. We're at an inflection point relative to technology, and Scott DeBoer can talk in much more detail about that and how it impacts the various products we're building and how it might play out moving into the future.
I think for now, I'll ask you to take it at some level face value that over the next number of years, we're really going to see a change in the rate of scaling, not only for DRAM but for NAND flash. We're going to see the introduction of a lot of new memory types. What that does is it sets the stage for other levers on the business to become relatively more important. We'll come back to that here in a minute. The other thing that clearly we've experienced over the last number of years is consolidation, not only in the memory sector itself, but also in our suppliers and in our customers.
The graphic I have up here, and those of you on the web watching can see, really tries to quantify that with some bar graphs relative to where is 60% of the output going in terms of number of customers or where is a significant amount of the equipment supply coming from. You can just think about some anecdotes relative to what's been going on relative to ASML. We can talk in more detail about that later if you'd like to or with the investments of certain memory suppliers into the lithography area, and the indicative nature of that as to how important those supplier-customer relationships are relative to the capital equipment realm. We've also seen a pretty good consolidation in our customer base.
With really some pretty significant consolidation, not only in the computing segment, the PC segment, but also in the phone area and across the board relative to the end applications, as well as the introduction, by the way, of new customers that are much, actually even further down the value chain than our historical ones. We've seen the consolidation that's ongoing here in the memory space that I know that you all want to spend a little bit of time talking about today as well. The net result of all of that is that memory suppliers are more important to customers because there are less of us, and because our capacity will be such that we can individually service end customers in ways we couldn't before.
That wouldn't be possible if we didn't also have this changing dynamic relative to the rate of change of technology, because in the world we've had historically where next year the customers can count on a commoditized DRAM product that's 30% cheaper than the one they got this year, the ability to compete with that ever-moving benchmark has been very difficult. By the time you get that custom product in and designed and working in your system, the ability to extract value versus what your competitor may be using in a more commoditized memory is diminished.
That trajectory of technology changes, we believe that you're going to see, and in fact, we are already seeing, and I'll come back to some examples here. We're already seeing an environment where the customers are much more thirsty to get very close with the memory suppliers and work with them on differentiated memory solutions, and beyond that, even memory system solutions that include controller functionality, firmware, software, et cetera. A few examples of that are things like the Hybrid Memory Cube Brian Shirley will talk about a little later on today.
The 2.5D integration where Micron can take bits of its memory intellectual property, circuit elements that maybe have to do with memory control or wear-leveling non-volatile memory or testability or repairability of memory circuits, and couple those with custom interfaces for a particular enabler to give that enabler a strategic leg up on somebody that's not working with Micron. That whole relationship between the larger memory suppliers and the larger memory consumers is really changing in ways that wasn't possible a number of years ago. One example of this is in the mobile area, where many of you may be saying, okay, well, we understand there's LPDDR2, maybe there's some competition from a more commodity DRAM that's slightly smaller. How's that going to go? The answer is different than it's been in the past. Different platforms are choosing different solutions.
Some will choose LPDDR2, some will choose a slightly lower voltage commodity DRAM, and they're all trying to get a slightly different angle based on their exact application. Part of that is because these products are all going to fewer customers, but to more applications. That doesn't mean that there's not a diversity of good opportunities in the embedded space and the smaller market opportunities where we can park products that drive higher value. It does mean that in those large volume markets, the product is going to become stickier with more value add. It's part of why we're so excited, particularly in the long-term prospects for the DRAM market moving forward.
Coming back to the low-power DRAM arena, what you're seeing on a go-forward basis is that the enablers out there, companies like TI or Arm or Qualcomm are all working outside of the mainstream JEDEC process in terms of how do they think about memory. If you go to one of those guys and say, "What's your memory roadmap?" Chances are they're not going to tell you. We look at that as a very good thing, because what it means is they really have embraced this thesis that working closely with a memory supplier on a go-forward basis is how they can drive a lot of value. The lines on the graph down here are directional.
I don't mean by any means to say that the industry structure has changed yet, but I think that the dynamic is clearly there in place where with the consolidation that's going on and with the behavior that you're likely to see from suppliers wanting to be close to memory manufacturers, and with customers wanting to be close to memory manufacturers, more of the value is going to end up with memory manufacturers. That's why we're still excited to be in the business, and that's why we think the future can be pretty bright. To be successful in that world, we think we have to have enough scale that those consolidated customers and those consolidated suppliers want to be or look at Micron as the partner of choice. It's no longer sufficient to be a great technology partner.
They have to be able to look at you and say, "Hey, I need" or, "I'm going to develop a product, and I'm going to need 10,000 wafers a week for the next year and a half or two years, from my partner to meet my demand, and I don't want that to be 50% of their capacity. Because if I ask him for that, I'm not going to have any upside surge, and I'm not going to have the confidence that if there's some sort of hiccup in their system or if some other opportunity comes along, that they're going to continue to meet my demand." In order to take advantage of these opportunities on both sides of the supply chain, you want to be a leading player, and Micron is positioning itself for that with the Elpida acquisition. Will Micron be successful in this new world?
You guys have seen these slides before. These are the companies that have come and gone in the memory industry. This is where we are today. Nanya has already told the market that they are less interested in being a major manufacturer going forward, and they would like to establish a model maybe that looks more like a Winbond, which, over time, we think is a good model for a smaller company, but one that becomes less significant in terms of the overall marketplace. We think we've done a good job through the years from using the levers that have been important in the past.
We think we have the right core competencies moving forward and happy, by the way, to talk about what some of the new ones we'll need are as we get into the Q&A, because clearly we'll need new competencies, and we're moving in that direction as well, in order to thrive in the new world. Generally speaking, we think we're in pretty good shape, and this Elpida acquisition, obviously is a piece of cementing that picture I just showed you relative to making sure that we have the right industry structure from a memory supplier perspective and that Micron has the right wherewithal internally, in order to thrive in that environment. Let me jump to Elpida here, just for a few minutes.
I'm going to bring Mark Adams up, and we want to chat a little bit about everything I said because I know there's going to be a lot of give and take on that. You guys have seen these slides before too in memory presentations or quarterly analyst calls, previous conversations we've had with you before. The combination of Micron and Elpida really does create the world's best memory portfolio and the world's strongest technology company. We have, together, NAND, DRAM, NOR, phase-change memory, a great capability relative to new emerging memory technology development and the technical resources to support customer-specific interaction and development on a go-forward basis. We have scale, and in particular, combined with the-- You'll recall back in April, we completed the transaction with Intel to grow the scale of our NAND business.
Our NAND business now already-- I anticipated coming down because we had some inventory sell-through in the second quarter, but we were 20% of the NAND market, on the NAND merchant market for the first time in Q2. That number really is probably more stable in the 15%-17% range right now. That change in our structure is part of this directional move we're making to become a significant supplier. We look at our capacity as sufficient, as well as our product portfolio and technology portfolio is second to none. When you take Hynix and Elpida-- Sorry, when you take Micron and Elpida and put them together, you have one of the largest semiconductor companies in the world. As suppliers look at us and say, irrespective by the way of whether or not we're making a capital investment in a supplier.
I'm sure I'll get some questions on that. We don't think that's an issue for us. Irrespective of that, we're a big player from a capacity perspective. We look at our capacity as largely fungible across memory technologies, and we'll be well-positioned in order to meet our customer needs as well as be significant from a technology development perspective. I know there's going to be lots of questions on Elpida, let me just say a couple of things, and then I'll bring Mark up, and we'll jump in on some Q&A here. One question, of course, as always, well, what's the timeline? Are you guys going to close this thing? If so, when? I will tell you that I think the timeline is still the same as it was when we announced the signing of the sponsorship agreement.
We still expect to close this in the first half of 2013. Exactly when that is, I can't say with a lot of precision because I think that really the key domino is when we get regulatory approval from all the concerned countries. That's just a process that is not particularly deterministic in all jurisdictions. We think that'll work its way through, in the first half of the calendar year, and we'll get to close. There have been, as you've read in the press, there have been challenges from various classes of creditors, et cetera. We don't anticipate any of that creating a problem for us today.
Really, when you think about who has the wherewithal to provide a meaningful return to the secured creditors on a go-forward basis, it's the Micron offer and the other offers that are out there, at least in our opinion, and we believe that the court will come to this conclusion as well, really don't give much assurance whatsoever, that they'll be anything for anybody. We still think it's a good deal for us. The market's been tough and, I'm happy to answer any questions about the value and where we see the value here in the future. We still think that we struck a good deal that's got the right net present value, that's got the right financing structure, that will be beneficial to Micron shareholders and give the various stakeholders in Elpida the best possible return.
When we put this thing together, I already talked about the manufacturing scale and the R&D capacity on a go-forward basis. We also expect that we will affect significant synergies in the Elpida business. Those reside in a lot of different areas, and I hope to have some more discussion about that with you. We think it's the right net present value. We think it's the right deal, and we think it really completes this picture of a strong Micron in a dramatically improving memory landscape. Again, I know there's going to be a lot of questions on Elpida, and we'll jump into those here in the future. Mark, before we jump into some Q&A, I think maybe we should sit and we have, I think, a slide here, with some strategic questions. Let me click through that. Oh, sorry.
Let's go back to here, we can jump into these strategic questions here. We're going to open it up to Q&A on any of these topics. Mark, maybe I'll take the first one, and then you can take the next one, and we'll see how we're doing. Again, we want to keep this interactive, so I'll interrupt Mark, and Mark will interrupt me, and that's how we normally.
Okay
that's how we normally do business. Don't take that as a sign of disrespect either way. I think the first general issue is, what's our primary focus and what are the challenges and opportunities associated with that? Or why is that our primary focus? I think, first and foremost, of course, we're always focused on execution. We've got a tough market environment out there. We've got to be very cognizant about running the business that is the business today. That involves making sure we're driving cost reductions everywhere we possibly can. We're driving an efficient manufacturing operation. We're optimizing our existing capital. We're running the business effectively. We're fortunate enough to have a great and deep management team here, and that is working very well for us.
We're very proud of the way things are going in terms of the daily operation of the business, despite the tough market conditions. That doesn't mean that we can just count on it to run on autopilot. We continue to stay focused on that. I think the bigger, the longer term issue for us is this issue that I just talked about, which is, we need to make sure we're positioning ourselves to be successful with our customers in these new differentiated product areas. That means, we need to have the resources working on that, which means we have to allocate our company resources differently than we have in the past, because the levers on the business are still the levers on the business, but they have slightly different magnitudes now.
We're very cognizant of the fact that technology scaling, you always have to keep up with, because if you don't, it's fatal. At the same time, the size of that lever is somewhat diminishing and the size of the lever that's associated with making sure we have the right customer partners and customer interactions and are delivering differentiated solutions to them in whatever form that may come. Again, the BU guys are going to talk a lot about what the detail around some of that is here in a little while. We need to be very clear that we need more of our resources working on that. In some cases, we don't have all those core competencies today.
That's right.
As a management team, we're working to bring them in. You're going to get the opportunity to meet Mike Rayfield here in a little while. He's a fantastic addition to our team. He's the kind of guy that we're bringing in all over the company now, to make sure that Micron has the right interactions and the right capabilities to drive differentiated solutions into our customer base. Did you want to add anything on that, Mark?
I just think that the question was opportunities and challenges, and I think Mark's starting also hitting on the challenges for us, and I think we acknowledge the challenge. If you go back 10 years and think about a company that's primarily DRAM selling to computing applications.
We have the ambition to be a leader in enterprise storage. There's a recognition on our management team that we need additional competencies that we haven't had maybe over the last decade. You think about that, and I think that that's a challenge for us as a company to move from a pure semiconductor play to more of a solutions orientation. Yeah. By the way, that comes not only in terms of how we define new hardware solutions or how we partner with enablers or foundries or customers to create the right capabilities, interfaces, control, functionality, et cetera, but it also comes in the software and firmware area, and whether it's SSDs or platforms beyond SSDs that our customers may want us to help them with. Mark, maybe you can take this next one. That's great, because you get the nice, fun, strategic one.
I get the one on, how do I cure the losses in the DRAM business? It's good to be the boss. I think fundamentally what we believe is that when we look at our competitors and our place in the industry, we think we do a pretty good job of segmenting and diversifying away from the pure commodity play. The PC DRAM question for us, when you look at our memory portfolio, specifically on the DRAM side, our success in the specialty market is really unrivaled. When you look at the share we have in server and networking, areas like automotive. PC, I think, is now under 20% of our DRAM business, if I'm correct. Where just less than five years ago, it was probably closer to 60%, 70% of our business.
I think that we will continue to focus on other market opportunities to drive our DRAM capacity and overall capacity. Mark kind of did a little foreshadowing today about what you're going to hear from some of the technology and BU teams around innovation and targeted performance and design capabilities to meet unique application requirements from our customers. It's funny, the slide that Mark talked to earlier about customer concentration, it talked of the revenue piece, that actually revenue consolidation around memory is focused on smaller set of customers as industries consolidate. I would argue that chart on profit looks different. On profit, I would say that the channels and the customers are in that other 40%, are as valuable or more valuable.
If you take a look at Tom Eby's business around embedded, most of Tom's customers will be in the other 40% of revenue, but highly profitable because the engagement we have is differentiated solutions custom to a specific need.
Yeah. I think that's very true. That's the part of our business that's already pretty good. Yeah. We already have a pretty good business in some of these more differentiated applications, more service-oriented segment of the business, where the customer needs higher service levels, higher quality levels, more help integrating their products, et cetera. That business is great for us today. What the historical problem has been in the memory industry is that piece of the business that can overwhelm you in a tough market.
What I was trying to talk to or what I would like you to understand is that piece of the business is going to be healthier on a go-forward basis than it has been historically. It's not just by the way because it's all going to be more stickier products that are oriented to particular customers. It's also going to be just because there's consolidation, and my guess is that on a go-forward basis, you will see the surviving players be more rational in terms of how they approach the market. There's not going to be, I don't think, there's not going to be a new DRAM company pop up in Taiwan this year. I don't think there are a whole lot of countries that are ready to jump in today either.
Because the barriers to entry are just very high at this point, and the capital is expensive. I think we're in an environment now where that supply and demand will be better self-regulating, and will be healthier for the industry. I wanted to jump back here on this first question around opportunities and challenges and just highlight one example, which is this Hybrid Memory Cube solution that you've heard us talk about in the past. This is an extremely high-end solution for high-performance computing and networking, and the buy-in by all the customers in that space is phenomenal in terms of the interest in these types of products. There will be other folks that participate, but Hynix's solution will be different than Micron's solution, will be different than Samsung's solution.
That's the point here, is that the customers won't just buy a Micron memory that's compatible with one from Samsung. They will look to Micron to deliver something that works uniquely well in their application, and they will look to the supplier that they have a lot of comfort with. Hopefully, it's the one with great technology that doesn't compete with them. Mark, why don't you talk about. I'm going to go back here. There's a capacity slide. Why don't you talk a little bit about capacity at Micron and what you see going on-
I want to go here. There's some next ones. I want to go here. Why don't you talk a little about what you see going on with the capacity at Micron and maybe the industry generally, to the extent we have something new?
Sure. I think, fundamentally, if I look at, especially coming from my prior role in sales, the one big ding we had with our customers was that we weren't big enough. In 2011, as recent as 2011, we had about 10% of the NAND market, trade NAND, because the other half went to Intel. Then we had about 13%, 14%, depending on output in a specific quarter, of DRAM. Fundamentally, that put a limit on how engaged we could be with major customers. They like Micron, they love our technology, they think we have great engineering teams, but we just couldn't supply to the level that they wanted. Our biggest customers wanted us to be twice as big.
I think 2012 was a significant year in terms of that lone feature of, we've now got a roadmap to be able to supply our biggest customers to the level they want to engage with Micron. When you couple that with a portfolio of products, our conversations with our customers are dramatically changing. You've now got a business that, a post Elpida close would be in the mid-20% DRAM market share. Then, when you look at the restructuring of our JV with Intel, and this quarter, of course, was upwards of about 20% of overall NAND share. That will go somewhere, depending on output, somewhere between high teens to 20%. You've got a company that's got the capacity to compete and engage much more strategically with a large customer base. From that perspective, it's pretty good.
Okay, but it's DRAM, and DRAM hasn't been very healthy. I think another message that we want to make sure we articulate today is that we have flexibility over time to adjust our output capacity where we need it and what we need in what form. The nice part about this capacity is it wasn't new. It was in the industry. We think as we consolidate this capacity under Micron, we get tremendous scale, we elevate the relationships with our key customers to be much more strategic because of that and the portfolio of products on our roadmap. We also think that with that scale and alignment with our customer base, we'll be able to drive just better business relationships and generate better operating platform.
Let me just make one more point on capacity. You cannot underestimate the importance of the fungibility of this capacity in the market we have today, given the eventual emergence of new memory technologies, and in particular, given the strong growth in the NAND piece of the market, and the relatively muted demand growth in DRAM. I think DRAM, of course, it's soft right now. We all know that, and we all acknowledge that. There is still growth in DRAM bits, and over time, that'll work its way out. The long term, the world's going to lead a lot more NAND capacity than it is DRAM capacity. This capacity, as Mark said, is definitely fungible. The only other point I'd like to make on capacity is capacity is becoming more valuable. Existing capacity is becoming more valuable than it has been historically.
You might say, "Why do you say that, Mark? What's different?" What's different is that because the rate of technology change is slowing, and because lithography is not changing as rapidly as it-- in fact, it's stagnant right now. At some point, EUV will come along, and that will change. Today, lithography is really essentially stagnant. What that means is, and we'll talk more about CapEx later today, but what that means is that existing in-place capacity has more longevity, and therefore delivers a better return. You have more time to amortize that capacity. When Micron goes out and buys cheap capacity for pennies on the dollar that can be upgraded relatively inexpensively, and the cumulative cost is much less than putting in place greenfield capacity, there's a bigger return than there would've been had we done that five years ago.
Obviously, this is a strategy we've employed through the years. It's been pretty successful for us. I think it's actually more important now than it's been historically, notwithstanding the fact that DRAM demand growth right now is somewhat muted. That ought to be enough to get you going and get some good questions kicked off. Obviously, Elpida, anything we said relative to industry landscape, anything we've touched on, we'd love to get in now. Obviously, you'll get another chance at us both, and me in particular later in the day. If we don't get to it right now, we'll get to it later today after you get a chance to hear from some of the other folks.
Yes, hi. I just want to talk about DRAM. You talked about consolidation in the industry. We have seen that since last year, right? Nobody added capacity last year. Nobody added capacity this year. I give it that, but we are still seeing the condition where DRAM is in today, right?
What do you think needs to happen? How do you see this market coming out of this downturn this time?
You want to take it?
Sure. I think that there is a slide in there. That one. I'll allude to this thing. We talked a little bit last night at dinner with some of you that we think that the industry on the supply side is doing all the right things for a correction. We think we have pretty much a demand problem in the short term, and I think that demand shift is just technology shifting from platform to platform, meaning from a pure PC segmentation on the commodity DRAM bits to other applications, i.e., tablets and smartphones, and I'll talk to that in a little bit. The other side of it right now is, if you look at the macroeconomic environment, I just read this week the IMF's recent adjustment down of the global GDP. Look at Europe, still got big debt problems. U.S. isn't exactly nailing it on all cylinders.
China has growth issues, right? Contextually, we think it's a demand problem, and we're doing things structurally to make sure we're prepared for whatever length of time this takes to get the demand problem in better shape. Structurally, on the DRAM side specifically, because I think that's where your question is, we think that's all the things you just referenced. Lower CapEx, as the chart's talking about, production shifts. We think that's all very healthy for the industry. We think it's demand driven, and we think over time, that will work its way out like other cycles.
If the root of the question is, do we think that DRAM demand, bit demand growth will continue to be less than bit demand growth generated through technology migration, I think the answer is definitely not. Okay? Supply growth on a go-forward basis is going to be in this 25%-30% range, and when you look at all the other non-PC sectors, which are a big piece of the total demand now, they're growing at a CAGR 50% and higher. It's not a long-term issue. It's a short-term issue. At some point, will DRAM suppliers take matters into their own hands if a catalyst is needed? Of course. We told you the capacity's all fungible. Of course, those are catalysts that can come into play if we're wrong.
We actually see any sort of reasonable overall macro environment, this thing sort itself out pretty quickly.
Go ahead,
Along the same lines, as the number of suppliers consolidates, can you see yourself unilaterally running the fabs less than full? When there was 15 guys, you sort of always run the fab full unless you get up to cash cost. Can you see a scenario where you're big enough that your own actions could have enough pricing impact that you can run the fab underutilized and see the market rationalize that way?
Yeah. I think that's another way of accomplishing what I just said. One is to move the capacity where it's productive. The other is to not utilize the capacity if that's what makes sense. I'll tell you that's easier to do if you don't have a large fixed cost because you bought the assets cheaply. It's less painful for you than for the other guys if they're doing something similar.
One in the back there.
Your comments about waiting for demand growth to improve, how long is it before you take matters into your own hands and say, Micron does have a strong balance sheet. You can survive many quarters of 1% book value destruction, how long is it before you, the management team, the board say, "Enough is enough, and we have to take matters into our own hands"?
Well, I'd love to answer that, but there's two reasons I can't. One is if I knew, I wouldn't want to telegraph it, and two is I don't know. These decisions always depend not only on where are you today, but what's your view of the future at that point.
I'll also say, Doug, just to the point about uncertainty, I'll be the first one to tell you, I didn't think that the DRAM market was going to recover like it recovered in post-holiday season in the spring of, was it 2009, I think it was. The uncertainty of trying to pick that is what Mark's trying to project, is that we're watching it really closely, and we're trying to take a look at the broader indicators on demand. It can be very uncertain when you see a chart like this and DRAM production bits down to where they've never been. Who knows when that alignment comes up and the market goes into a tighter condition, and we want to watch it, and we have the flexibility to act. I think the answer is we're just not sure how to call it right now.
Yeah. The other thing is, I think we do have a little bit of unfinished business here that we're working through relative to consolidation as well right now.
Hi there. I got two questions, one on your CapEx chart here for DRAM. You're actually down next year. Are you suggesting that you do not plan to have any conversion and related shrink CapEx on your side? Just to think about that piece, it's actually tough to see for me why overall industry CapEx will be down again next year just because of Micron and PD, basically. Secondly, you talk about fungibility of capacity, but actually, if you look at the process flows between DRAM and NAND flash, they're more and more different. Are you telling us that you're actually about to switch a lot of capacity to NAND flash?
Yeah
deal with LPM?
Let me take those, Mark. You can jump in. Let me do them in a reverse order. First of all, relative to this fungibility issue, I think it's very inefficient to ping pong with high frequency. I'm talking about a low-frequency response and a long-term retuning of capacity to meet market demand. I'm not a big believer, other than small amounts at the margin in fabs where you're running maybe both technologies, and we do have some fabs like that. I'm not a big believer that the sort of high-frequency shifting to take advantage of short-term market perturbations is a particularly productive thing to do if you're running a manufacturing operation. There are things you can do with the margin, but I don't think that's a big effect. Now, sorry, the first question again was?
Just to finish on that one.
Oh.
That means you are post-Rexchip, Elpida, Hiroshima coming in.
Yeah.
Potentially the whole of Inotera as well, since Nanya doesn't need the capacity anymore.
Yeah.
What do you do? Do you convert tech? Do you convert Rexchip? Do you convert something else? What actually were you referring to?
I think we should come back to the CapEx discussion in more detail later on when Ron's up here, because he's got some things he wants to tell you about CapEx generally, and that's maybe a good time to address that topic. Relative to why we think DRAM CapEx will be flat to down, we're just showing you our market research relative to what competitors say they're going to do and what suppliers say our competitors are going to do. It's our best guess, as it always is, relative to what's going to happen. That incorporates what we plan to do as well. I think there's a lot of diligence that goes into those numbers. They're always subject to change at Micron and at our competitors, and you got to take them with a grain of salt. It's our best guess.
Now, relative to what Micron might do relative to changing its capacity, we clearly have a 30-nanometer transition that's starting to near completion. 20-nanometer actually is not going to be a big spend for us in 2013. That transition will happen for us a little later. At Elpida, they have a 25-nanometer node, that they have talked about publicly. Probably at this point, I can't say too much about what their plans are. I don't want to say too much about what Micron's plans are relative to moving capacity around because, as I told Doug here a minute ago, I don't know yet what we're going to do there. There are lots of options. We're fortunate now that we've got a large network of 300-millimeter fabs, and we have lots of options we're looking at.
exactly what we do will depend on a lot of different variables, including our view as to is there a shift in timing relative to when we're going to get the Elpida transaction closed? What are the latest data points on macroeconomic effects, as well as the DRAM market health?
You mentioned about, I guess, opportunities and challenges, and you kind of said that, I guess enterprise storage was one of the areas where, I guess, you don't have sufficient capabilities there right now, or you want to expand, I guess, in terms of depth there. Can you expand a bit more about that and what other areas you think that you still feel that you don't have all the pieces in the puzzle of other product areas that you think you should expand to?
I think we're going to have a lot more discussion about that later with the BU folks about particular opportunities they see. Let me characterize it generally. I think enterprise storage is clearly an area where there's a lot of opportunity and there's a broad swath of different types of opportunities, different areas to go after, whether it's SAS or PCIE or SATA or some other formats we're looking at. They all require significant resources, lots of customer interaction, that's clearly an area where we have some building to do. Beyond that, we need more folks that can go out and interact in the mobile arena with the enablers to really make sure we're driving differentiated and value-add solutions for them.
That involves not only the circuit resources to optimize those interfaces and to get the right IP over onto their side of the interface, also involves really folks with deep knowledge of their system-level challenges so that we can help them craft what that solution looks like. I don't want to get a whole lot more specific than that, the BU folks can maybe point you more specifically.
We'll just do one more question, guys.
Okay.
Just one question with two parts. Do you expect or do you see your competitors reallocating more dollars of capital to M&A to build out infrastructure for some of the things that you're thinking of, and therefore, that would limit the incremental capacity added? I have a follow-up.
I'll let Mark answer this one, too, because he's got pretty good eyes and ears out there in the marketplace. I would say yes, absolutely. I think there's a lot of interest, much more so than there has been historically, by some of our competitors in the memory market, out there looking at what are the opportunities to grow inorganically into some of these new areas. They get it as well.
Which, it'd be great if they were a flip of the switch, but I don't think anyone can hope for that much good fortune. The flip side of it is, I think it validates the thesis that we're putting in front of you guys.
For a big picture, with the memory cost of scaling decelerating, and less capacity added going forward compared to in the past, what happens to the margin profile? We go through these ups and downs in prices, especially in NAND. With cost of scaling decelerating, is that going to remove some of the pricing upside that happens when there is a shortage? How do you see the margin profile in the next couple of years, especially with the bottlenecks in lithography that is limiting the cost of scaling?
I think that answer lies in the level of differentiation you do, not just in the silicon itself, but in the overall solution you're delivering. To take in the last question around enterprise storage, there is no DRAM exchange for enterprise storage class SSDs. It's a solution. It's got a number of different components to the solution, and it's valued much differently than a PC module. We think to the extent that we can keep on differentiating and adding value, not just in the core technology, but around the solution controller, firmware, software, and working with our customer for custom features they need to deliver their performance, we think there's value in that's not measured in cost per gigabit or gigabyte.
Let me throw out an analogy you might think about. It's not perfect by any means. I think I would say that, who knows in the short term, but the long-term trend, I think, is to less volatility and higher margins. The analogy I would give you is the NOR market, only a much larger market. It's not a small market, it's a big DRAM-size market. A model that looks more like the NOR market where the products are stickier, and the rate of change of technology migration is slower, and the customers want to make sure the designing cycles are longer and the customer isn't flipping from vendor to vendor as often because the products are all slightly different and interface with the system in a slightly different way. That might be one analogy to think about for the DRAM business on a go-forward basis.
Only it clearly will be much bigger. All right. Hey, great questions. I'm surprised we didn't get more on Elpida, you'll get another shot at me later, we should probably move on now.
Great.
I'm going to step down here, right?
Great.
Yep. Before we get into some of the technology that we alluded to as far as roadmaps and some of the BU focus areas, what I thought I would do is talk about something we don't spend a lot of time is on the kind of outside of the market update I'll give briefly is on the operations of the company. First, let me just talk a little bit about the industry, I think we've already dealt with a lot of this in some of the Q&A. Clearly, right now, DRAM is in a funk. Pricing this quarter to date, probably a little bit off where we thought it would be, a little bit lower, slightly lower, not a major shift, not in a great place, clearly.
On the NAND side, I would say slightly opposite, probably a little bit better than we anticipated quarter to date. On the margin, nothing's shifted dramatically in memory on the ASP side. We think inventory on DRAM is probably a little bit flush, around five or six weeks in the channel. We think NAND's pretty tight. I think the overall message around the market condition is that, yeah, PCs are in a precarious place. If you look at the traditional players in PCs, they're struggling right now. When we think of bit consumption, we think of what I would call personal client devices, whether they are consumers or whether they're in the commercial environment, we think that there's tremendous upside going forward.
If you layer on top of just generic desktop notebooks, you put tablets on top of that, you think about smartphones, you think about what people are doing with smartphones, searching, communicating, watching videos, playing games. If you think about that, you think about what's driving bit consumption, we think that there's still a pretty good future here for DRAM in these segments. Yeah, right now, the PC market's pretty tough, there's some catalysts like Windows 8, there's ultra-thin, we think those are all positive. Beyond that, the overall idea of look at the number of units that are being forecasted for this whole category of personal devices, that it's going to be consuming memory. We're pretty bullish on that.
Right now, certainly in a downturn, especially around DRAM, but we think overall, the demand profile is pretty good as part of the DRAM industry. Not all. We have a lot of different segments we market to, but we think the growth in terms of bit consumption is still there. It's not as dire as one might think. When you think of technology companies, the boring parts are how do you compete at an operating level? Again, I don't think we talk about this a lot as a company. We probably should do more of it because at the end of the day, in a relatively tight margin environment, how we operate can make a big difference in the bottom line of the company.
One of the things we're focused on at Micron is being much more efficient in what we do, and that means tough choices. As we think about what's changed at Micron and the capacity we've had in the company, and improvements, one of the benchmarks is our capacity versus our overall headcount, which normally drives and leads to expense line. We are very focused on becoming world-class in our overall headcount optimization, headcount efficiency, and expense management. Today, I think we have room for improvement, very openly. You can see we've been focused on it, and as you look at our P&L, you look at SG&A as an example. You'll see SG&A managed much more tightly going forward in these market conditions as we look at how we manage our business. You see capacity going up, and you're seeing headcount relatively flat.
Part of that is just making tough decisions on who you are. One of the things that Mark has challenged us with is, we have to stay core to who we are as a memory company. Sometimes large companies, especially during the good times, a year like 2010, you get ambitious. You want to go do more things. I would say that we're being guided with Mark's vision, that we have to focus on the solutions aspect of memory and get away from the tangential businesses that might not be core to who we are. Another thing, inventory. In a business that I would argue that's much more complex, because we've got multiple different types of supply chains. We've got the traditional commodity supply chain around PC applications. We've got embedded, which is a lot more customers and unique requirements around product longevity and delivery performance.
You've got wireless, which is a little more custom in nature than the PC business. Then you've got our NAND business, which is definitely more of a solutions orientation, and newer opportunities around demand that provide us forecasting challenges. Inventory is going to be a critical piece of how we manage our business. We put together a number of programs in the last 12-18 months to help us think about working capital management and how we drive efficiency in our company. You can see here that we're having some good positive results. These two slides is an example of we're going to be much more focused, going forward, on making sure that while we expand in these new markets, we are properly managing the business financially to make sure we're in the right businesses and we're doing them right. Let me see here.
Go back to one slide here. From our perspective, when you look at what we're up against in terms of not knowing the demand picture in the PC business and what have you, then DRAM, we've got to make some choices, and we've already made some. Let me give you a couple of examples. We thought we had a pretty compelling proposition to compete and enter the solar business. As we look at the Elpida opportunity, you have to make trade-offs. Mark and the team, we got together, and we said, "You know what? We are who we are. We're a memory business." We made decisions on memory and display, and we're looking at other areas that we can reshift allocation of resources to drive the best use of our capital going forward.
That is around what we're talking about today, solutions orientation, memory value add marketing segments. From that perspective, you'll see us, today in our communication, talk solely about that in different app, in DRAM and NAND, and how we go to market in wireless and embedded. It's a much different approach than trying to be multiple technologies outside of the core memory business in which we're in. That's a message that you need to understand that's critical to us going forward. You'll see also about technology leadership across all segments. We have to be a leader, and we have to innovate. That might come in the form of partnerships. That might come in the form of redirected deployment of assets. You've talked about how fungible we need to be and react to the market conditions. That's critical to our business.
Premium segments, a leader in server, a leader in networking, a leader in automotive solutions. We're going to continue to invest in differentiation. That's a theme out of today that we'll repeat, and you'll see examples of as we go through the BU focus. That's critical to our business as well. We need to leverage how strong we are in those markets and the dependency our customers have on us as it shifts away from a generic standard part to something that's innovative for the customer, innovative that we need to focus on from a transitioning to a technology-laden customer engagement model. Finally, I would say that Mark referenced Mike Rayfield. Mike just joined us within the last month and is leading our wireless business. We have another executive team member that joined us within the last couple of months on the supply chain side, John Waite.
We've never talked about supply chain in a forum like this. Supply chain for us as it relates to customer delivery around optimizing inventory, around the delivery focus and performance of meeting our customer requirements in different segments, it's critical to who we are, and it's not the old DRAM company serving 10 PC customers. We are looking to round out our management team, whether it's developing people internally or going out and getting the best possible people to run our businesses to help us lead us to these new market opportunities is critical for Micron.
As we look at the challenge for us and the opportunity for us, part of that is filling in the gaps on the technology side, whether it be more solutions orientation capabilities or competencies. Then getting the right people in place and enabling them to drive us to that opportunity. I do want to talk about each of the BUs and what they are focused on. When you think about the message of differentiation, it sounds great. It sounds like a nice place to be, but what does it really mean? Each of the BUs is being tasked to take on this thought process about how we differentiate at the technology level and the market segment level.
If you think about, for example, in DRAM, there is a number of examples that you are going to see today around DRAM, and Mark referenced Hybrid Memory Cube as one of them. As you look at more networking systems going out to support data center and cloud infrastructure, or you look at the server impact to this growing phenomena, there is more DRAM bits going in those systems than there ever has been. Going forward, it will increase. As we think about that, Brian and the DRAM team have to go deliver value to the customer in terms of how the memory fits into new architectures. It is not just a 1 CPU game anymore. There is a number of different CPUs that are popping up across the business units, and how we deal with adding value and innovating where the memory fits and gets around bandwidth issues.
These are unique challenges that are new to the DRAM business. That is really what you are going to hear from Brian today, is how we are developing a roadmap around technology to drive those solutions to capitalize on the market opportunity. If you look at NAND is a little bit different today because you think about the shift from where NAND came from. NAND came from where? Photography cards, dumb USB sticks. It evolved into MP3 music and some storage around video players, what have you. It was mostly consumer that drove NAND. Even today, consumer applications are still a fairly healthy demand driver for NAND. There has been a pretty dramatic shift, and I would say 2012 was a big year in that way. If you go back two or three years ago, we were thinking about what is the magic price that is going to drive SSDs.
I think the first generation, or the first guess at that, was we got to get down to $1 per gigabyte. That did not quite do it. Then we started to get down lower and lower, and about $0.50 a gigabyte, we saw a pretty big shift. We saw industry volumes go up in SSDs, and that has triggered a movement from some of the capacity that was more focused on consumer in the past towards some new applications. The enterprise space has really taken off. Shipments of enterprise products have continued to grow massively in 2012, and it is early in that game. What Glen and his team need to focus on is how do we capitalize on that opportunity?
How do we take those bits that used to be pretty commoditized, supply-and-demand driven bits, how do we add value to solutions where the NAND's part of it, but not all of it? Where your negotiations with customers are more about how can we add value to your solution, customer? How can we add value to that rather than how do we look at gigabyte cost and pricing? The phenomenon around enterprise storage is really interesting because we just met with a startup company in the last week that's got, I think it was a 20 or 44 terabyte system that can daisy chain into 250 terabyte systems, all NAND. Serving a government application, serving large, massively parallel mainframe applications with large corporations. The idea that you would even talk about NAND in this type of environment five years ago was going to be hard to imagine.
Glen's got an opportunity in NAND, is how do we take what's traditionally been a consumer model and then go out and engage with customers and deliver that? I think we're on our way. We've got great partnerships. He'll talk about the elements of NAND that's going to lead us to winning in this environment. You've got controller development capabilities, both in-house with partnerships. You've got firmware and software resources that are needed. The NAND business is actually similar to where DRAM might have been before we drove into the specialty and differentiated markets. I think that that's the opportunity for us, and that's what we got to go execute to. On the embedded side, it's a little bit different. I think the technology pieces are in play.
I think what you'll hear from Tom is more about a leadership position with a lot more opportunity to go dominate. Certainly, a leadership in NOR, in the embedded market, and what we've capitalized on. I remember what we talked about at the time of the Numonyx acquisition was we can take a portfolio of products to this segment, we've actually gone off and done that fairly well, we think there's much more opportunity, meaning that we've got DRAM and NAND driving the embedded market, and we have access to that market through Tom's team in the Embedded Solutions Group. There's a lot more growth for us there. We talked about the cash flow capability of NOR, DRAM, and NAND solutions going into that market. We've got a pretty good roadmap there.
It's more about channel execution and how do we get a company that's traditionally been an OEM engagement model to go in in distributed customer environments, distribution segments or smaller customers that are more high margin, low volume opportunities, a lot different than maybe the PC plays. Tom's opportunity is more about how we go to market, how do we service this customer, and how we drive more accelerated growth. We think there's a really an opportunity for us to drive that business. As you can see from our earnings announcements, a very high margin, good business for us to be in. Finally, our wireless group. I would say that that one has been the most challenged of the four business units for us. I think at the time of the acquisition, we had a customer concentration that was more heavily focused on feature phones.
As a result, our product alignment wasn't necessarily lockstep with the rapid growth over to smartphones. Surely, we're in process of doing that today, and we think that we'll have great results in 2013 and 2014 in terms of moving our product direction, our customer engagement down that path. We're coming from a place where that shifted pretty quickly, and we had some legacy technology that was not deployable to smartphones, and that's something we had to fix. Hopefully, that's why we went out, and we feel like we got the best guy available to us in the industry. Someone to come out and help us rethink about how we go back and win in this business. By the way, winning doesn't mean we're going to be the leading market share in mobile.
I think the mobile business, if you're not careful, has characteristics of the PC commodity business. Winning means how do you be profitable in this business? There's a fine line between customer engagement just for market share and being smart in how you engage with customers. We think that our product portfolio is catching up. We have some exciting low-power offerings coming down the pike from Micron, certainly around eMMC and MCP offerings. We think we're getting more aligned with the right segments in the wireless business, but we have a long way to go from a customer engagement model and go to market.
From that perspective, this is a little bit of a turnaround in our business, and openly, we feel pretty good about where we're going, but it's been a disappointment for us, and we're trying to fix it on a number of different levels. That's the challenge for Mike in his new role, is that we're in a business that's kind of coming from the wrong customer mix and some product misalignment to engaging with the right set of customers and making sure our product roadmap's in place to drive that customer model. With that, I'm happy to take any questions about what I presented. There was certainly some overlap from what Mark and I talked to, but happy to take any questions about that before we move on. Yes, sir.
Just with regard to the Elpida acquisition, how does the Elpida acquisition help the DRAM business unit in terms of the capabilities and what you're trying to do there?
Well, I think that, first and foremost, the point I talked to earlier in the Q&A with Mark was that the customer engagement to know that Micron has the scale going forward to be a significant, potentially number one supplier is very material for us for new opportunities. Secondly, we're not nearly tapped out on our ability to penetrate the specialty markets, and I really think that's the opportunity. If you go back and look at where Elpida was coming from, they primarily were in the PC and wireless business, and we think one of the opportunities is to take that capacity and put it in better homes, and we don't think that's an impossible chore. We think that's kind of what we do well. We like that capacity in the markets we serve.
Both in scale and additional capabilities to drive specialty business are two ways I would answer that.
Can you elaborate on the NOR business? You talked about products, that you had the wrong products. Exactly what does that mean, and how do you turn that into the right product, or is it basically you discontinue it and you develop new products?
Well, I think, maybe to clarify, the NOR wasn't necessarily the culprit for why we didn't succeed. I think it was the overall product portfolio that was featured on a segment of the market that didn't grow as fast as other aspects. You saw basically some shifts in winners and losers in the mobile business. You think of what Apple and Samsung have done in the smartphone segment, and you think of some of the players who have struggled, and you've seen a big, major shift in terms of who's the winners and losers. From that perspective, it's not about NOR, DRAM, or NAND, it's more the packaged solutions and the architectures of what they want in their phones. NOR wasn't necessarily the lone culprit.
We think actually NOR's got some place going forward because the smartphone business really is now even segmented itself. There's now kind of a lower end of the smartphone business. That you're seeing people kind of generate some technology platforms around the entry or value smartphone that you're going to take to emerging markets. We don't think it's a NOR, DRAM, NAND question. We think of it more as just an overall product roadmap alignment to a specific category, feature phones. We hadn't advanced the ball fast enough on the smartphone business piece.
Are you saying there's a market for NOR in low-end smartphones?
I think there's potential for leverage of that, yeah. Okay. One last question.
Quick question. How do you actually differentiate your SSD offering from your competition? What are the things that you need to do to just stay ahead of your competition? Thanks.
I'm going to let Glen speak to that when he's down here. When you think about SSDs, and you think about things about reliability and performance and those are some of the attributes in the application space they serve, whether it be client or enterprise. I don't want to necessarily steal Glen's thunder. If we don't answer that during that session, we'll come back to it. What I wanted to do next was, I wanted to bring up Mike Rafiel, who, again, just joined us within the last couple of weeks. I wanted to let Mike tell you a little bit about his background and also give you a perspective of the wireless business and his assessment and views in a very short time of where Micron might fit. Mike Rafiel.
Great. Thank you. I think I've worked with a number-- Wow. I wouldn't want to see that early in the morning, so I apologize for that. I'm going to talk a little bit about, there's people who've asked, "What's the opportunity for Micron in the mobile business?" I thought, I've worked with a bunch of you guys for a long time. You know I'm a pretty direct person, and you've asked me a number of questions over the last day. Most of them are, "Why did you come to Micron?" I think if I answer that, it'll shine some light on what the opportunity is. You're going to hear a little later from Scott about sort of the core technology, NAND, NOR, and DRAM that we've got in here in Micron. I think what you'll see is the core technology we've got is leadership.
We can go toe-to-toe with anyone on that technology. What was highlighted a little bit earlier, the real key is wrapping the right stuff around that technology, whether it be firmware, whether it be IOs, whether it be developing the relationship with the SoC suppliers or the end customers to go off and make that technology compelling. If you look at the mobile opportunity next year, it's about $18 billion or $20 billion. Pretty significant. Our business last year in mobile was about $1.2 billion, and most of that, as highlighted, was within the feature phone or the entry-level smartphone. It's mostly NOR technology. That was probably $3 billion of that $20 billion.
We've got a pretty good chunk of that already, and above that, you get into higher-end feature phones and super phones and tablets, and we've got the core technology in DRAM, and NAND, and NOR to go off and service those. If you think about China, you get entry-level smartphones in China, they want a lot of these MCP products. You take a look at what we've got in the 30-nanometer DRAM, you stack it with NAND, you've got a great solution to go forward in China, and that's a huge market. We're already starting to win designs and starting to ship on that. All the way through that stack, as we've introduced the new 30-nanometer DRAM stuff, the four-gigabit stuff, we found great success in design wins that allows us to, again, participate in that upper $16 billion of market.
Finally, you look at NAND. You do eMMC, managed NAND, you put that together. Any of you who've got kids, they've got all of these media devices, tablets, phones. They've got a staggering amount of managed NAND on it. It gives us a way to differentiate. We do a little firmware, we do a little software, we put a bunch of our NAND technology in, we've got the ability then to participate all the way up through the marketplace. As you go forward, I think, this slide here, the reality is the top 3 bars are mobile. What they require is mobile sensibility.
If we've gone off and aligned with the people that drive that mobile business, whether they be ecosystem partners or end customers, all of a sudden, those top 3 bars become a place where we can go off and have great success hunting. I think that the most exciting part about coming here is this is probably the largest single opportunity in the industry to go from a $1 billion business to many, many billion-dollar business with the core technology we've got. You add Elpida to that, it gives us scale, it gives us great customer relationships. It's a pretty exciting opportunity.
Let me ask you guys to clarify that chart. Where is the infrastructure? You mean where are the things like networking server and other applications? Sorry. This chart was more around, probably the infrastructure devices, the client devices. Really, probably could've been better labeled, but it wasn't necessarily around the higher-end data center cloud component of DRAM. It was more around the client or portable devices that interplay there. The consumer DRAM, basically. Primarily, or included in that also is anything in the client, commercial, corporate desktop, and that type of stuff. Server, storage, networking. Not in there. That's right. Okay. That's right. Okay.
Okay. I think this is all set.
I think we're all set, except for anyone have any questions on Mike before we let him go too far?
Just on the, if we look at the wireless business operational last four, five quarters, it has posted losses, right? How do you see we come out of that situation, and when do you see that kind of turn around in that business?
It's been about product focus, right? I think there was a little bit of innovator's dilemma. You get focused on the NOR business in feature phones, and you just sort of follow that thing into the ground, right? It's not going to go away. We've got products to service that going forward. Fundamentally, somebody backing up and looking and saying, "Let's use the rest of the core technology we've got. Let's start shipping into these upper classes of devices," we can do that pretty quickly. I mean, we're already starting with a 30-nanometer LPDDR2 DRAM. We're starting to do it with the MCP products in China. We'll have the eMMC or the managed NAND products in the latter part of this year. We can very quickly, I think, have a pretty significant presence in those markets. I optimistically do it pretty quickly.
It might be a little early for you to answer this question, when you look at the mobile market, it doesn't appear that we've seen sort of demand elasticity related with pricing. How should we think about the mobile customer and the way in which they approach buying memory? Is there a bill of materials they're targeting? Is there any chance you're ever going to get these guys to sign up to long-term contracts? They seem to be buying more unique solutions than I think the market necessarily recognizes, and that's because they're constantly getting market-based pricing.
Let me answer it from a mobile industry standpoint, because on my third-week anniversary, I'm probably not going to be able to go much deeper than that. There are a bunch of phone customers that say, "I need to build a phone for 1,000 RMB," and then it's a question of what you can supply them in that, what sort of uniqueness. I think we have the ability to put unique solutions together with SoC suppliers that can maybe get somebody more capability or more functionality in a given price. I think there's an opportunity there. I think after that, from an elasticity standpoint, you have to move fast and be able to differentiate, whether it be firmware on managed NAND, whether it be unique interfaces on DRAM or NOR, like SPI on NOR.
I think if you just stay close to the people that are driving the market, you can find a way to add more value to them and therefore get more value if you can go off and have the right product mix. In the end, I think that's the right answer. Can we do one more question?
Hey, Mike. In targeting this China opportunity, I'd be curious, how important is it to have relationships with the baseband vendors over there, the MediaTek, Spreadtrums of the world, and how would you rate where Micron is there relative to the relationships with Qualcomm and the rest of the baseband vendors?
It's very important to have relationships with all those guys. You named probably the two key guys in China, as well as all the North American people. I think we do okay. I was lucky enough to be a partner with Micron on the other side until a couple of months ago, I think there's more we can do there. Ultimately, what that does is creates a huge worldwide virtual sales organization, that's what we're going off and doing, have them go sell our products. We'll support the hell out of them early on and make sure that they help design us in, we'll be pretty aggressive on that.
Great. Thanks, Mike.
No problem.
We're good with that. So far, we've actually talked about a lot of the corporate initiatives. Mark's talked a little bit about the industry landscape. What we wanted to do next was to bring up our BU leadership team, to talk a little bit about some market conditions and our response to that and some of the themes we've driven home already that they can actually add some more substance around. Brian Shirley and Tom Eby and Glen Hawk.
Thanks, Mark.
Okay. Good morning to everyone. By way of introduction, up here on the stage, we've got Glen Hawk representing the NAND Solutions Group, Tom Eby representing the Embedded Solutions Group, and I'm Brian Shirley, representing the DRAM Solutions Group. I think what we wanted to do today is take a stab at the submitted questions, see if we can lend some insight to those, and then a little bit of an abbreviated session from what we've done in the past, and then have a chance to get into some more direct Q&A as well. With that, I'm going to go ahead and kick off. The first submitted question was, how does Micron reposition its DRAM business with the lack of growth in traditional PC applications? Now, as you've heard Mark and Mark talk about, for us, the answer is pretty easy.
It's really a threefold strategy, but at heart, the big-picture answer is you go where the bits are. That's number 1, wireless client applications on the DRAM side, as Mike indicated. The second big area, which I'm going to spend just a little bit of time on, is really in infrastructure DRAM. This is not a new area for us. This is the conjunction of server and networking DRAM, and it's an area that frankly, we feel like we're doing pretty well in. I'll tell you on the networking side, the demand remains insatiable out there. It always has a feel of a segment that's a little bit difficult to forecast, quite honestly. There's questions about CapEx expenditures and LTE rollouts, et cetera.
When you look at the continued push on mobility, when you see what's happening with data center switching, networking continues to grow by leaps and bounds, and frankly, Micron's pretty well-represented there. By our calibration, we have over 40% of the DRAM $ revenue in this segment flowing through Micron. It's been a good segment. It's been a very profitable segment. Frankly, the rest of the vendors in there are disproportionately fragmented. This is a segment, by the way, this is not a small segment. It actually is nearly as large in terms of revenue for Micron today as the server segment. We've been pretty pleased with that. Some of the solutions you've heard us talk about in the past, such as RLDRAM, have really provided a boom.
The second area is servers, and there has been a little bit of a shift out there. Some of the traditional server applications are a bit on the slow side. Some of the news reports out recently. I will tell you, in aggregate, servers remain very, very strong. There is a shift in the model out there, however, where a lot of the direct vendors are taking over share, and it's the data center applications. They are building their own servers and stuffing them full of memory. We like what's happened out there. Up on the screen, you will see a picture of a motherboard that supports the Romley Sandy Bridge and Ivy Bridge platforms. What we like about this is that those blue and black slots, those are slots for DRAM modules.
That has recently grown from roughly 12 slots up to 16 with the advent of Romley. I'll tell you, the data center vendors are using every single one of those slots. The driver for that is that as has been represented by others in the industry, speaking to the goodness of mobility driving server CPU growth, frankly, that's just as good for the DRAM guys. As a matter of fact, it's better. 600 smartphones, 120 tablets driving a new server. If you go through the math, if you look at the silicon area on a server motherboard, frankly, there is anywhere from 10 to 20 times the amount of DRAM silicon on that motherboard as there is CPU silicon. That's a pretty good thing for us.
There's a good growth in the number of servers, frankly, the DRAM content per server is what's really staggering here, and it's been a good thing for us. I'll also add that in the big data set, it sounds like a little bit of a buzzword. I'll tell you that it's real. These are applications that try to sort through all of this data that's been created, and by and large, it's a DRAM problem. These millisecond transactions that you hear about, these are systems that are loaded up with literally a terabyte of DRAM, because that's the only way you can sort through it fast enough. You're starting to see some of these words out there.
I won't go through the particular vendors, if you pay attention to the space, in-system memory, several other forms of ways, what that all gets to is loading big data sets into DRAM and making sure you can sort through it quickly. We like that trend. The third area, quite honestly, is that, as Mark and Mark indicated, this shift of PC growth being a little bit slower is more than counterbalanced by the rise in NAND silicon going into client applications. Glen is going to speak a bit more to that today as well.
Exactly, Brian. Exactly. Good opportunity where there's a lot of synergy between what we're doing on the DRAM side and the NAND side at Micron.
The second question, what is your outlook on DRAM for next year and beyond? There was a question earlier about where was infrastructure. Here is a graph that actually includes infrastructure DRAM bits, and what you can see up here, frankly, is that going out to 2016, infrastructure is not a corner piece of the market anymore. It's a pretty large piece, which when you combine that with tablets, mobility, and the ultrathins, you've got a pretty diverse set of applications out there, driving still pretty hefty bit demand on the DRAM side. Overall, 2012 to 2016, it's tough to perfectly prognosticate these things, but we are seeing, overall, demand bit growth pretty consistent between 35%-40% CAGR year-over-year. 13 over 12, obviously, a difficult thing to project. As Mark and Mark indicated, the PC market's slow today.
A big piece of that is the global economic situation out there. We do tend to still be somewhat bullish on a pickup after 2012 with Windows 8 and ultrathins and what that's doing. Some rough numbers to keep in mind. We think 13 over 12, somewhere on the order of about 45% bit growth year-over-year. Lastly, on the DRAM side, as both Mark and Mark indicated, just really a completely different landscape from the 25 years I've spent in this industry. Things are changing in a way that makes it feel like this business is, in certain ways, almost turning into an ASIC-like model. The drivers for that are not just industrial. Frankly, it's technical.
As you load up systems with a lot of DRAM becomes an increasingly large part of the power budget, and frankly, when you put a lot of silicon in a system, things slow down, and that hurts this trend of memory not keeping up with processor speed. There are big architectural trends driving us both on the power side as well as the performance side, both in client and infrastructure. A couple notables up here on the left. Micron has introduced something called DDR3RS. We were the first to validate this with Intel. This is shipping today. It's a feature that has to be enabled by DRAM as a way to get lower power out of ultrathin applications. On the right, you have HMC. Some pretty good stuff happening with HMC.
We've spoken about it in the past, I'll just tell you, going to the next slide, we have formed a consortium out there. It did what we wanted it to do, which was establishing HMC as the upper-end solution in the infrastructure space for upper-end DRAM performance. All of the major DRAM developers have joined up. There will be a final spec release coming up here in about the next four months or so. In 2013, you'll actually see us releasing engineering samples of the full production HMC, with volume to start in 2014. Pretty pleased with how this has gone and see it as a huge area for future growth. With that, I'm going to go ahead and turn it over to Glen.
All right. Thanks, Brian. I'll keep this brief just to get to the Q&A quickly, but at the same time, want to do justice to some of the questions you were thoughtful enough to send in ahead of time. We had a couple of questions up front. The other on the screen right here, one's about what's happening with overall industry growth and what are some of the segments that are driving that. I think Mark Adams started us off with the beginning of the story, because I think this is a situation where the past is a pretty good reflection of what's going to happen in the future. The story of NAND flash to date has been that it's changing the world, and when it changes the world, that creates more demand for flash.
Of course, the story that Mark was sharing with everybody was the disruption to film and CD-ROMs, and now most of us can't remember a day when we didn't have our MP3 player, and probably most of us don't have a lot of CD-ROMs or 8-track tapes or those kind of things. That was in the 1990s. What's shown on this chart here is that more traditional NAND flash market. It's shown by the gray bars. Guess what? It's still growing, okay? I think the next wave of innovation came from the mobile handsets. It's no secret that mobile handsets wouldn't be as thin and light, as fun to use, if you didn't have a lot of flash in there. The light green section on this graph represents that growth.
If you just look at the top of that light green line there and you do some compounded annual growth math on this, it turns out that top green line is growing at about the same rate as Moore's Law. Mark Durcan had talked a little bit about the challenges that the memory industry has keeping up with Moore's Law. Scott DeBoer is going to come up and say a lot more about that. The point here is that even if the NAND market were only capped at that top green line, we would still have to treadmill to lift those just as fast to stay even. The big growth story for NAND and the things that have happened more recently have been shown in dark green here is the tablet phenomena. Obviously, that's consuming a lot of flash.
We think that on the right-hand side of this graph, out around 2016, that's going to be more than 15% of all the NAND bits in the world. The other big growth driver, of course, that we're very excited about are the solid-state drives for client devices and, as Brian mentioned earlier, some of the opportunities we see in the cloud. We think that the SSDs are going to comprise over 20% of the bits out in time. Those are two huge new disruptive innovations that are occurring, and I think that we've just started on a couple of them, by the way. What this means is that over time, the industry is going to have to add capacity, and of course, we'll do that in the right way.
There was also a question on what's our product focus, especially there was another question about how do we see the synergies between the DRAM group, particularly in cloud computing. This slide just makes it pretty simple. We're going after the real growth, the real exciting opportunities of those two new disruptive trends that I mentioned. On the client side, there's a lot of portable devices out there in the world today, whether I see some thin-and-light notebooks out here in the audience, I'd like to see a few more of those. As you get into ultra-thins or tablets, I think this looks a lot like some of the products we see in the embedded market, right?
I'll talk a little bit. Some of the migration we'll see from tablet applications processors into embedded is something we can take advantage of, and that's real good.
Definitely
Down to the core technologies that we're delivering, whether it's the NAND silicon or the controllers.
Yep.
It's nice fit for both of these markets, nice synergy here for us at Micron. I think at the bottom here, I think it's pretty well-known now that flash first made its entrance into the data center, into cloud computing, mainly on the storage side. It was pretty easy to take a flash solid-state drive with a SATA or a SAS interface and slide it into the storage racks and see an immediate benefit from power reduction, from reliability. That's certainly where the revolution has started, so to speak. Now there's a lot of excitement about server-side flash. We have some PCIe SSDs that have made a pretty big splash here, and this comes back to Brian's point about it's so nice to be able to walk into a customer with leading-edge flash for servers as well as DRAM.
In fact, I think Brian and his team got our foot in the door, at least with a lot of the work they've been doing on the DRAM side.
That's right, Glen. A number of these customers, frankly, to put it bluntly, they want to be working with one company on the entire memory subsystem. That is a real issue these days. All of this is getting re-architected, that's just much easier to do when you can go to one company and handle both the DRAM side as well as the NAND side.
Yeah. Everything that we've learned about providing high-quality, reliable, consistent supply to mission-critical applications on the DRAM side in the data center, those exact same principles apply to solid-state drives. It's a lot more than the technology that distinguishes us or differentiate us from some of the competitors, as was asked earlier. There were also some questions here, specifically on SSDs. Big focus for us, so I was very glad to see that. The competitive environment, et cetera. Just to show you a few proof points of where we're at on solid-state drives. The client solid-state drives, we've been shipping for a couple of years now. We're in the mid-teens in terms of market segment share. That really helped us get our R&D and our manufacturing lined up in this new, more highly integrated system-level solution. From that, we've started to work on enterprise. I mentioned that PCIe solution.
A photo of it here is shown at the bottom. This is our P320 product, one of the fastest PCIe SSDs on the planet right now. It's definitely targeted at the top of the top high-performance server-side flash applications. A nice couple of announcements this year from EMC and Dell using this product. In addition to those solid-state drives, we're also finding a lot of success of selling our components into other innovators that are out there. There's so much innovation that's driving product proliferation at the system level now. Sure, we can do a lot of these SSDs and PCIe cards or SSDs ourselves, but there's a lot more going on out there. Mark Adams had mentioned one of those companies earlier, Skyera, is the one that has that 44-terabyte storage box that he mentioned. There are 3,000 individual Micron flash chips that make up that solution.
You know what? It was only possible because we're the only NAND flash manufacturer in the world today that has a single-chip 128-gigabit MLC solution. That was a big reason that they selected us, and of course, all the other experience that we have delivering enterprise-grade NAND to the mission-critical applications. The thing I wanted to point out here was that it is about the SSDs for sure, but we're also having great success selling our NAND components into people that are doing similar things, and it's created sort of this virtuous cycle for us. The more of the SSDs that we make and the system-level solutions that we deliver, the more we learn about what we can do differently in the NAND chips or the NAND silicon.
That makes the quality of those components better, a lot of the feedback we get from our customers is that our NAND flash works better in these kind of enterprise applications or even SSD applications than some of our competitors. I think this climbing of the value chain, and the work that we're doing to deliver more fully integrated system-level solutions is a big part of why all this works together well.
Okay, great. Thanks, Glen. A couple questions here. The first one, probably a little bit more remedial, what is the embedded business, and how does the Micron portfolio apply to that? In very short order, we sell the full portfolio of non-volatile and DRAM products to automotive, industrial, medical, multi-market, another word for everything else. Then we focus the non-volatile portfolio primarily on the balance of the segments in networking and storage, consumer, and in PCs, with, at the highest level, a value proposition that leverages the full line. Customers want to buy from one guy. Stability across financials, across roadmap, across supply, because particularly on the left side of this chart, people are making decisions that may last for five or 10 years, so they need to be able to count on you.
last but not least, a very wide range of unique requirements in the embedded market, I'll talk about those a little bit more shortly. Before I do that, just a quick comment about why I get to be the last guy up here. In, I guess, somewhat flipped terms, it's because I get to take advantage of mooching off all these other guys a lot, right?
We'll let you mooch.
Yeah.
Yeah.
Mike may not have figured that out, but he will quickly. It's at a component technology level in NAND and DRAM, it's increasingly at the memory subsystem level, both in NAND and a little bit further out on the DRAM side. The point of it is that the ideal embedded business cannot exist outside of the cocoon of a large, multi-technology, leading-edge, high-volume manufacturing memory company because all the customers up here, including the ones on the segments on the left, they want leading-edge, cost-effective, high-volume manufacturing. They also have a whole host of unique requirements, so that's where the embedded group comes in, right? Oops. What are some of those? Certainly at a product level. We take an eMMC solution that was developed for wireless.
Well, we're going to ship that to an automotive guy, part of their qualification is to build a couple of hundred cars. They'll send 100 of them up to northern Sweden and drive them around in February for a month. They'll take another 100 of them and drive them down to Morocco in July, it's got to work. There's issues at a component level, at a firmware level, at a software level, where we've got to basically adapt and modify at a product solution so that we can meet that need. There are services. Again, more to the left side of this chart, automotive, industrial, and medical. These are products that need to last for 10 years. We have a product longevity program that guarantees, not at the same litho node, but certainly from a functional point of view, a 10-year lifetime.
We have to work with the fab operations to make sure we understand how to do that in a way that works in the network. Of course, as Mark referred to at the beginning, there's a whole bunch of channel activity, because particularly in that industrial, medical, multi-market arena, there are thousands and thousands and thousands of customers, and you can't get to them with a direct sales force. You need to work very carefully with a rep and a distribution channel organization, because individually, they're very profitable, they're just also individually quite modest in size. A number of the areas where we need to work. Before I go on to the second question, just real quickly, the one area where we do contribute a bit more from a technology development perspective is NOR.
Just got back from our fab in Virginia, we call it MTV, a couple of weeks ago. We've seen our first 300-millimeter, 45-nanometer NOR wafers out of that fab. We think that will help us extend, both from a technology and a cost perspective, our leadership there, of course, continue to gain share. Our CapEx, to be clear, going forward, will be predominantly in the NOR space on 300 millimeter going forward. The second question is, how do you grow? How do you defend, and how do you extend your value in this market? There's certainly a whole host of issues there, and channel being a very important one, as Mark alluded to. The lifeblood of the embedded business is design wins.
Over the last about 18 months, we've had a very focused effort with the field sales applications and, of course, business unit resources to reinvigorate a design win process that, frankly, probably got ignored a little bit right after the close of the Numonyx acquisition. We're now able to track on a by segment basis, on a by technology basis, on a by region basis how we're doing as a future predictor of design win success. Clearly, the traction that we're getting here was part of the very significant growth that we saw in the second half of the recently completed fiscal 2012, including what were record revenues and margins in the fourth quarter. Where we're, I think, doing a reasonable job, but frankly, expanding a bit more is at the reference platform level.
If you can get upstream of the OEM design by getting to the chipset guy first, as Mike said, let their sales force be our sales force, that can be that much more effective. There's an increasing level of focus there. Perhaps what's most exciting, though, is we're starting to bring up and make operational a series of regional embedded-focused systems engineering labs. They're intended to be in-region, close to the chipset partners, customers, in some cases, end customers, like automotive. A good example is our lab in Munich, which, amongst other things, is focused on the automotive business. The objective there is to be working with reference platforms or actual products, sometimes from a chipset partner, sometimes from a customer, like an automotive electronic supplier. Some case, even end users, so we are starting to work increasingly with the automotive guys.
Part of it is pre-sale. Let's get the customer to market faster. I think more exciting is understanding how the customer is using the memory, and back to Mark and Mark's comments about how can we differentiate and create tighter relationships. As an example, the automotive guys, as they look to their experience post-Japan earthquake and Taiwan floods, they want to understand the downstream players. There's an increasingly tight tie to the automotive end users and understanding what they're looking for and how we can adapt our systems is a great way to continue to add value.
To borrow on a comment that Mike made as well, there's sort of a new dimension of mooching that we see coming in the future because the SoC manufacturers, a lot of them are taking a high-volume segment and then redeploying it for lower volume, higher value embedded segments. A PC graphic subsystem gets adapted to be an automotive display solution. A little further down the road, supercomputing architectures are going to get adapted and shrunk because they got to fit in a rearview mirror for advanced driver automation and safety solutions. Much more broadly, the mobile apps processors, like we were talking about, show up in a whole host of embedded applications, certainly including infotainment in automotive.
To the extent that we can get that much further upstream, be thinking about not only what does the graphic subsystem guy want, but what is he going to want when he converts it to the automotive display, again, there's that much more opportunity to influence and tune that subsystem and add value so that we can not only grow the business, but hopefully grow the value of the business going forward as well. With that, I think we're going to turn things over to Q&A.
Back on slide, I believe it was 38, Glenn, you showed the different market segments.
Yeah.
Clearly, those are forecasts which always end up wrong. My question for you is, which one of those categories do you think you will be wrong on the low side, meaning it will exceed your expectations? Conversely, which one do you think has greatest risk of falling short of the forecast?
I think the enterprise SSD, that top line, is the one I personally think is significantly underestimated. It's going to take a long time to get going. The innovation and disruption that we're seeing in the data centers, I think we've barely scratched the surface. We've got all these high-performance flash bits trapped behind this legacy hard drive interface, and we're getting away from that, and people are starting to see the value. I think that one, I would be willing to bet, is going to see higher growth. I think where we might see a little bit of pressure is on that traditional segment that Mark Adams talked about, the stuff that's related to film and music storage, the USB, the SD cards.
I think that one is one that's because for the most part, a lot of that storage is ending up in some of these portable devices in the cloud.
Relative to the enterprise, given that there are very significant economic advantages-
Yeah
to the server company and to the server farms, is it possible that that will end up being bigger than the consumer SSD?
Wow. I would love that. From a dollars perspective, I think that's possible. Bits might be a different answer. The prevalence, the ubiquity of people want to carry a lot of content with them. I think that's going to continue for a long time. From a bit capacity, I think the portable computing client stuff will-
Yeah
exceed that.
We'll just add that these data centers are performance-driven. There's been some public discussion about the kinds of metrics that the large data center operators look at specifically for response time. Frankly, that just drives a lot of silicon, a lot of memory silicon, both in NAND and DRAM and potentially others in the future. It's a very dynamic space.
I'm hoping that they'll allow me one more question for Tom. The embedded design wins ended up, in every case, exceeding what your goal was listed as, if we understood the diamonds correctly. What is it that you all are doing that is allowing you to exceed the expectation or your internal goals?
Yeah. I think it's, probably in simplest terms, it's focus. Again, a couple of people have commented on the shift from 10 years ago, when Micron was a PC DRAM company for, I don't know, 75% of its bits. The mindset of going after hundreds and hundreds of designs and the work that's required from salespeople and FAEs and EU marketing is just very different when you're going after this much more granular approach. I think it's just been one of training, setting the right goals, the right reinforcement. It's a lot of fairly mundane blocking and tackling, because when we can get the product in front of the influencer, be that a chipset manufacturer or an OEM or an ODM, we've got pretty good product. It's getting the team focused in the right way.
I think it's been that effort and that focus that's allowed us to be successful on that front.
Thank you.
Just a quick one on the server segment. You obviously rightfully said that we're moving more to generic servers with clients specking them directly. In that context, once you qualify, let's say, by Intel for Romley, what is the most important for you in terms of generating sales? Is it engagement with Google, Facebook, et cetera, or is it to work with a Quanta or Wistron to make sure you're actually there where the manufacturing happens?
Great question. Two notables. The first is to your point on enabling and qualifying. The big issue is still, frankly, high-quality bits. These applications do require high-quality bits. We are over-shipping relative to our market size in that space. It's been a good space for us relative to the other guys. We have been supply constrained in that sense, that's part of the reason why we're excited about new capacity. I will tell you, though, that also with part of this data center bring up, there's a change in the model where the orders come in a little bit with less predictability. The average size of the business is growing by leaps and bounds. It's a little bit clumpier than it used to be. What that means is, frankly, good supply chain operation and being ready to supply that business as it comes in.
A lot of these customers will call up and say, "Look, it turns out we're going for it on this next data center, so if you can supply X number of DIMMs in the next three weeks, you've got all the business, and we don't want to mix and match." You either get all of it or you get none of it. That has taken some supply chain retooling that frankly, we think we've done a pretty good job at, but the business has changed.
Okay, guys, we're going to go ahead and take a 15-minute break now so everybody can pause for a few minutes. We'll come back in the room here at 9:45 and get things going again.
Great.
Thanks.
Take me higher. You take me higher. You take me higher. You're the real thing. Yeah, the real thing. You're the real thing. Even better than the real thing. Even better than the real thing. Even better than the real thing. In the morning you go running for the man who stole your water. Then you fight till he is done it, but they catch you at the border. The mourners are all crying, and they drag you by your feet. The hangman isn't hanging, and they put you on the street. Yeah, you go back, Jack. Do it again. Wheels turning round and round. You go back, Jack. Do it again. When you know she's no high climber, then you find your only friend. In a room with your two-timer, and you're sure you're near the end.
You love a little woman, and she brings you only sorrow. All the time you know she's smiling, you'll be on your knees tomorrow. Yeah, you go back, Jack. Do it again. Wheels turning round and round. You go back, Jack. Do it again. Now you swear and kick at Vegas, that you're not a gambling man. You find yourself in Vegas, with a handle in your hand. Your black cards could make you money, you hide them when you're able. In the land of milk and honey, you must put them on the table. Yeah, you go back, Jack. Do it again. Wheels turning round and round. You go back, Jack. Do it again.
Come on. Baby, don't you wanna go? Come on. Baby, don't you wanna go? Back to that same old place. Sweet home Chicago. Come on. Baby, don't you wanna go? Hide Hey. Baby, don't you wanna go? Back to that same old place. Sweet home Chicago. Well, one and one is two. Six and two is eight. Come on, baby, don't you play with my heart. Hide Hey. Baby, don't you wanna go? Back to that same old place. Sweet home Chicago. Come on. Baby, don't you wanna go? Come on. Baby, don't you wanna go? Back to that same old place. Sweet home Chicago. Six and three is nine. Nine and nine is 18. Tip that brother, baby Elmo. See what I mean. Hide Hey. Baby, don't you wanna go? Back to that same old place. Sweet home Chicago. Come on. Baby, don't you wanna go? Come on.
Baby, don't you wanna go? Back to that same old place. My sweet home Chicago.
Okay, everybody, if we can get everybody seated, we need to get started again. Go ahead and take your seat, and the next section up will be Scott DeBoer. I think you guys have a lot of good questions for him.
Brian, can we move?
Scott, you can sit in the middle here.
Oh, okay.
You're the main event, man.
Is this working?
Yes.
We'll get started. My two counterparts here already have been introduced today. I'm Scott DeBoer. I am in charge of technology development for the company. I'm trying to focus on a set of questions, from the focus of these questions, they're pretty relevant. We get asked these questions quite a lot. I think we can get through the answers on some of these, Brian and Glen will both have some thoughts on how these are impacting our product line. Hopefully, we can get over pretty quickly to some other things. The first question is really related to a technology wall, and that's.
A discussion point around the industry right now. From our perspective, there isn't really something that is a specific technology wall. There's a slowing in the pace of, a potential slowing at least, in the pace of the cost reduction for both NAND and DRAM. I put one graph here that shows at least a perspective of what we think this may look like going forward. In terms of a wall, from our perspective, we have clear visibility into 2-plus more nodes of both of our primary products, NAND, DRAM, from a technology level. Not that there aren't significant challenges to make both those happen. We believe we know how we're going to do that. What this graph does indicate, where we show the change in the pace of the cost reduction over time, is really the trend.
Actually, it started more with NAND over the last couple of years, maybe surprisingly when we think about it. When we look at our technology nodes, historically, we've done significant big shrinks on a periodic basis to lead to this cost reduction. If you look at the NAND technology graph, and you look at where today is, and you see the line coming off of it, in reality, over the past more than one year, our competition has not shrunk in both directions anymore. We're not doing a shrink of the magnitude that's been done historically. Up to 20 nanometer, Micron has still done that. Our 20 nanometer product is a 20 nanometer by 20 nanometer square.
When the industry talks about this 1x node, which is 19 nanometers, that's really only 19 in one direction, and it's 20-something, more than 20, in the other direction. Our 20 nanometer die is extremely competitive. It's actually the smallest die, and the only 1 28 gig, as Glen talked about before. It's comparable to what a 1x node was going to be. When we look at the next node, our 15, 16 nanometer type node, that will again be what we believe is the smallest die in the industry. It'll come in along the curve somewhere between the solid line and the dotted line.
In the NAND space, the interesting thing there is Vertical NAND really brings a different dynamic to where the cost reduction path goes, I'm not really forecasting other than to say that it falls someplace closer to the dotted line. Then we see a scaling path on that also. From a technology wall point of view, that's the first question. Really on the second piece of that, on the pace of cost reduction, the pace of these nodes, depending on how big of a shrink a company chooses to do, the pace can be moved a little bit. If you take a big, aggressive shrink on the technology, that's the thing that is extending out. Different people right now are making decisions to do small, 10%, 15% shrinks and try to pull them in time-wise.
If you look at something compared to a historical pace, we believe that the technology node migration is going to slow. Okay, then those are the main two, I think one of the things that I was going to do here is just ask Glen and Brian both to talk to how the strategy of this node migration maybe impacts both of their respective areas and where we sit right now.
Sure. I can kick off. I think one of the notables is that as DRAM continues to scale, it is fair to say that there is some of the non-idealities that you start to see in the silicon do pop up, and this is true of not just Micron, but everybody out there. At least on the DRAM side, it is giving rise to concepts where the DRAM can be more intelligently managed, and it is akin to thinking of how NAND is managed today with controller silicon. Part of HMC, part of the appeal, and you will start to see this in other segments as well, is the use of management blocks inside of a controller or inside of the host that try to hide these non-idealities. It is more than just error correction.
There is other techniques to do that hide the fact that we are dealing with a small number of molecules and atoms here. If you can do that intelligently, which, by the way, speaks to the value of working closely with the customer and the enabler, you can overcome a number of those idealities or non-idealities.
Yeah, same for NAND. I think the more difficult it gets, the better it gets for us in a lot of ways. There is a lot of innovative things that we can do within our controllers to address some of the scaling challenges that we have. When we do end up making the transition to 3D that Scott talked about, that is going to put us in an even stronger position to hit the ground running. We can either deal with new issues that that new technology brings with it, or if we get a little bit of margin back, we can use that to increase the performance and maybe get a little bit closer to that curve that Scott was mentioning.
The next couple questions here, I have been asked these quite a lot also, are around our view on EUV first off, and specifically for DRAM and NAND and how is that different. Does the timeline of EUV specifically impact the technology migration? Generally on this topic, we view EUV as a technology that is really cost performance based. We do not have If I talk about the two-plus nodes that we look out, we do not have a position there where we say, "If EUV does not work, we cannot make those nodes work." What we do know is that our technology can be more cost-effective, but obviously our competitor's technology can be more cost-effective also, if EUV comes up and meets the performance spec where we can enable it for certain levels. It is more of a cost question for us.
We're working closely with the supply chain on EUV, which is a short supply chain. We're working closely with them. We have partnerships. We have time and engineers working at IMEC in Belgium on an EUV tool there. We're engaged on this. We're just watching it closely, working with, obviously, ASML as we go along to pick the right time for us when it's cost-effective. I'm sure that may lead to some other questions. The second big topic there is timing for 450, and this is another topic where we're carefully watching it. We're not out there portraying ourselves as an industry leader on pushing 450 forward, on pre-funding it, on really trying to push the ball. What we are doing is, again, watching it carefully to make a judgment on the right time for our significant engagement.
I think a lot of you who've followed us for a long time, we've talked in the past about not being as aggressive as potentially we should have on the 300-millimeter conversion, and I think we've learned from that. Although we certainly didn't want to be too aggressive on 300 millimeters. That didn't really pay off for all the memory companies who did that either. It's another judgment call. We're going to watch it carefully, and we're going to make investments at the right time for Micron there. Okay, let me move this slide forward. On general barriers for Moore's Law that are impacting us and are relevant to both the question and the first part of the topic, there are absolutely physical and electrical challenges and limitations of the material.
Those things, while they're not necessarily a wall, they're things that change the performance of the devices and things that we have to look for unique solutions to go try to work around and improve. An example of that, when we talk about scaling and challenges in NAND, we look at this 1y node that we're doing, and certainly that's going to come in. It's going to be a cost-effective node for us. It's a node that we're looking at and maintaining a leadership position on. At the same time, we have the majority of our development resources and focus on Vertical NAND, and we think that's where the big win in this game is, and it's going to be different for different companies because there's different approaches for how Vertical NAND's being done across the industry, and obviously, we think our approach is the best one.
We have a lot of confidence in that path going forward, you could call Planar NAND scaling a physical limitation, but when we look at our scaling path, we found a way around it. Equipment capability, we talked about EUV and it not necessarily being a barrier, but the things that are challenges on equipment capability are more around control systems for the process. Our challenges on equipment capability are down to atomic-level control requirements and how you maintain that across a huge volume of wafers in multiple fabs and make it the same every place. This is a major challenge across many equipment types, etch, clean, film deposition. All those things are strongly challenged with just fundamental capability for what our requirements are today and going forward.
The last one that we spend a lot of time on now that maybe was more straightforward 10 years ago, for sure, is the balance between how much complexity we have to add at time to make a node actually cost-effective in the future. We spend a lot more time now really understanding how much complexity we can add before the node is really reaching a point where the returns aren't where it pays back. That's, I think, a challenge, but at the same time, by planning it that way, we feel much more confident about the nodes that we're picking in the future. This two-plus nodes into the future, we believe we have a strong path to those, really either with or without EUV. Okay. Those were the big questions on that one.
The next one is really around emerging memory, and these guys will have some more to talk about there. From a first piece here, the dynamic that's coming up now is these new memory technologies being capable of enabling new kinds of system solutions. They have different kinds of properties. They have a different kind of cost structure, which is somewhere in between NAND and DRAM. They're not going to be better than NAND. They provide opportunities, and maybe both Glen and Brian can talk a little bit about some of those things that they're thinking of with the new memory technology coming in, then I'll talk a little bit more specifically about a couple of them.
Yeah. I'll start on this one. This is pretty exciting from my perspective because there's a nice progression or sequence of events. Scott and his team have a very nice near-term scaling path for us on 2D NAND. The introduction of 3D NAND and these emerging technologies, they all sort of build on one another, and I think it plays nicely with our progression up the value chain. Daniel, I think, had a question earlier about, I think you called them our puzzle pieces and which ones didn't we have yet. I would phrase it a little differently. I actually think we have all the puzzle pieces today to do the vertical integration. I think the puzzle's not big enough yet. There's just so much innovation out there, our development bandwidth isn't there.
One of those important puzzle pieces is the controller technologies, the IP surrounding that, the stuff that we're learning on the NAND technologies today. It applies not only to the next generation of NAND technologies, but also these emerging technologies. We are experimenting with, for example, our Phase-change memory technology that's used on the mobile side. In the lab, we can take those devices, put them in an SSD, and show some customers some interesting things with it we can do. We're already getting innovation ideas for those next generation of system appliances that are going to dovetail with the technology roadmap that Scott and his team have laid out before us.
That's it, Glenn. We just add on the DRAM side, there's really no perfect memory out there. It's a continuum. It's a big space. There are different memories that can be better for certain applications, better price points than others. Really, you have to look at this as a continuum and the memory as a subsystem. That, again, speaks to what Glenn mentioned on having the controller point in there specifically to manage those non-idealities and make sure that you're getting the best out of what each memory technology is really good at.
Okay. I think we're maybe a little bit behind on that. We've covered several parts of this topic. I think the big thing relative to these kind of new memory technologies coming in is the application space that they enable and being able to think about actually modifying the architecture of the memory system. When you bring in memories that are significantly faster than NAND, still non-volatile, and at a cost point somewhere between NAND and DRAM, you can start thinking about different system architectures, and that's probably another topic for Glenn to mention a bit.
I think that that's interesting thing is I think some of the emerging technologies that we're looking at, they're not necessarily replacement technologies in the classic sense of the world, where I think that the NAND scaling path that Scott and his team have before us is going to deliver the lowest cost per bit non-volatile memory technology for a long, long time. I think where you're going to see is some of these emerging technologies emerge and create new levels of memory hierarchy in system-level applications.
Okay. I don't know if you have anything to add to that.
I would add as well on the DRAM side that as some of the scaling starts to slow down, some of these technologies, like Through-Silicon Via, like assembly technologies, lower power transistors, lower voltage logic transistors, these become more important both to the application as well as the overall system. That's a pretty good piece of our focus right now on the DRAM side. Things may be slowing down on just the traditional bit shrinking, but there's so much more that has to happen when you have all of this memory in that application. You got to get the form factor right, you got to get the package right, and you got to keep the power down.
Okay. I think really for the last couple of questions that we got beforehand, both these guys have kind of touched on the role of memory in the future through the previous session and now. I think we could probably take questions.
450 millimeter is pretty far out. Do you think there's a need for more 300 millimeter high volume fabs, and is the U.S. a likely place for it?
Well, that's going to depend on the way the market evolves, I think. At the moment, and as Mark mentioned, I think earlier also, we don't see the need, and I don't think anybody in here would say there's a big need for a new DRAM high volume fab right now. As time goes on and you look at the market projections across the non-volatile space and some of the specific mobile applications, I don't think it's out of the question that the capacity grows for some of the emerging memory options and for certainly support for the volume required for the enterprise drives, though. In the timing point of view, that would absolutely still be a 300 millimeter based on 2017, 2018 kind of pilot line timing for 450 right now.
Do you see the U.S. as a likely place for a new memory fab?
Um-
By a U.S.-based company?
Well, I think it's kind of speculative on what the dynamics two or three years from now when that kind of thing would be contemplated would be. I think it's hard to judge what exactly would make sense from a financial point of view two or three years from now when a greenfield fab's going to be built.
Yeah, pretty speculative questions. Obviously, we can only speak for ourselves, and as Mark and Mark indicated, we're happy with our capacity plans going forward. We have a well-honed-out strategy of how we work these cycles to the extent that new capacity out there is required. We look at the total cost of ownership of the fab and would make a decision at that time on the best place to be.
I thought we were going to talk about the election in November there for a second.
Yeah. I have a question on EUV. You already talked about EUV being cost-effective before you can adapt it and the DRAM scaling. What is the throughput you think would make sense, would be cost-effective for Micron? Also, how long you could scale beyond 20 nanometer for DRAM, you think you would need EUV? Also, is it possible to kind of give a comparable chart saying, okay, EUV is 60 wafers per hour, it is better than double patterning or triple patterning or quadruple patterning, something kind of that?
Sure. Obviously, this is a concept that we study a lot internally, and I think to the first part of the question, some rough numbers from us are probably at 75 wafers per hour, it starts becoming a bit interesting. It really needs to get to 115, 125 wafers an hour for it to be very clearly cost
effective relative to double patterning, for sure. I'll get a little more technically deep here just for a second, okay? If you look at what really is driving the roadmap, it's not pure line edge resolution, it's overlay. It's pattern placement relative to other patterns. At the moment, there is no piece of the roadmap on immersion versus EUV that says EUV is going to be better in terms of that pattern placement. When you're scaling down below 16 nanometers and looking at that, it's all about how well you can control where the pattern's actually placed. We can make lines extremely small. Putting them in the right place is the urgent thing, and that's got to be addressed between EUV and immersion going forward, and that's the technology limiter, actually.
Thank you.
To that extent, if EUV were to fail or never materialize, what are the contingency or what are alternatives to continue with the cost curve? Second question has to do with the silicon through via. I'm assuming that the throughput for TSV is single-digit. What would need happen, is it an equipment vendor or is it customers that would need to make sensor changes, and what are the changes that would need to happen to help improve the throughput and economics of TSV?
The two different spaces there. On EUV, for these 2+ nodes that I'm talking about, it's a pure cost question, and the actual, I guess, if you'd say, interesting point about the cost piece of that is the technology we're using for pattern doubling is extendable. If we have to do more pattern doubling, you're not buying all the stuff that you already had over again. From a cost point of view, some of this stuff is fully depreciated assets out there at that point in time, and you're just buying the next piece to expand the amount of pitch doubling that you have to do. Relative to buying brand new scanners for a node like we've done in the past, I think the economics still aren't too bad, as long as you keep building on the same technology.
Say you have to quadruple patterning instead of pattern double. By the time you get to that point, the pattern doubling infrastructure's already in place, and you just have to add more of it to get to the quadrupling. It's actually more expensive to buy brand new scanners at that point. From a complexity point of view and how long it takes to get a wafer out of the fab, obviously that increases the amount of time the wafers spend in the fab. On the TSV point, the infrastructure's actually made a lot of progress over the last couple of years, but it's still 3DI and TSV technology are both still areas of big challenge for the equipment industry and for us in terms of building a more cost-effective process. Especially in areas like metrology for 3D interconnect, those type of things, the technology's still very immature.
It's made a lot of progress over the last year and a half, even. I think it's on a pretty good pace to continue to meet the challenges as we get to a point where we're really in a high-volume position with a product like HMC. I think we're pretty confident that between what we're doing and what the equipment industry's doing, it'll be in a good cost position. We're well past single wafers, though. Even today, the TSVs are actually relatively easy. Those are just big, deep contacts that in the DRAM industry, we got pretty good at getting things bonded together and everything. There's some challenges there, but even today, the throughput is nascent, but there are reasonable throughputs out there. At least I have Brian convinced that the process technology is a yeah.
Had to lend in on that one, Scott.
Quick question. Can you give us an update on where you stand on your ReRAM?
We have a lot of internal focus on Resistive RAM right now, and I think as I mentioned at our winter analyst conference, we have a partnership with Sony that we're working on Resistive RAM technology. We view that as one of those emerging projects that can come in the 2014, 2015 timeline, where it takes a different space. It's not a NAND replacement or a DRAM replacement, but it's a technology that we're confident is going to have the best performance of any of the Resistive RAM options out in the world. It's one of the major projects. I don't know if you're going to the R&D fab tour today, but it's one of the major projects in the R&D fab that we've built over the last year.
Hi, I have a question for Glenn. Historically, you've been able to drive demand, big growth in NAND by price declines. As Moore's Law slows, how do you maintain that same demand, the growth, or do you see that slowing as well?
Well, I think a big part of that strategy is climbing the value chain. I mean, a big way that we can deal with the scaling limitations. One of Scott's team members says is currently we're almost on a first-name basis with the electrons in some of the cells that we have. I mean, they're that few and far between. We're doing a lot of work in our controller technology to help deal with that. Between what we're doing there, not just at the NAND component level, but when you're delivering a system or an appliance, the degrees of freedom that you have to deal with those kind of issues, a whole new set of knobs become available to you at that point. I think that's our first response.
Other than that, though, I think Scott and his team have some pretty good things up their sleeve that are going to help keep this path going for a long time.
A very quick one on the ecosystem for TSV and 3D packaging, referring back to Mark's comments before on the collaborating with ecosystem and foundry. How do you see that on your side in terms of the investment Micron has to do on the R&D side and capacity, versus how much actually the logic foundries will do hopefully in close partnership with you? How should we think about who's going to drive it going forward?
You want me to take that one?
Well, I can take it from an infrastructure-
Go ahead. Yeah
point of view from a capital equipment, but-
Go ahead
yeah, I'd like you to take the other part of it. One thing is if you compare strategies relative to how the Through-Silicon Via technology is actually built, there's an option to build it late in the flow or earlier in the flow. Micron builds it in the middle part, front-end part of the fab part of the flow. From a TSV point of view, our strategy is not to outsource that piece of it. As we go forward and we look at volume and interaction with other people, and I'll let Brian comment a little bit more on this part of it, but from an infrastructure point of view, as we ramp this up, it's a question that is somewhat still open as we figure out exactly how we're going to most efficiently deploy the technology.
Yeah, it's a great question because there's a lot of interest out there, both with HMC, but also with, as Mark indicated, solutions where you can take high-density DRAM and put it directly next to an SoC or even on top of an SoC. Frankly, it does call for close partnerships with the foundry directly. It's also worth saying that that's much more of an ASIC-type model for the SoC enabler. It takes very tight partnership to get the memory enabled correctly with the particular SoC designer. In terms of who does what in the actual manufacturing flow, we're well advanced in some of those discussions and at the partnership level. Can I just tell you that there's some obvious ways things fall. We do the DRAM stacking. That's an obvious.
When you talk about getting that placed next to the SoC on some kind of a substrate, for instance, there's some obvious ways to do those questions just based on what's good for that supply chain. Some good partnerships with the foundries solve most of those issues.
Okay. I don't know where Ivan went.
Where did Ivan go?
He's out in the hallways. Okay. Yeah, I think we're done with this section and moving on to the next one.
Thank you.
Okay. You want this on?
Keep it.
Okay. We just had the white shirt team, now we got the blue shirt team, huh? First question we've got is, discuss your capital structure, including the use of capital leases and other instruments. Well, we have a pretty simple, and I'd submit, healthy capital structure. We use two principal forms of financing in Micron today. One is capital leases, which are secured by our CapEx investments around the globe, and the second one is our publicly traded convertible debt. When we think about our capital structure, we're focused on several variables. One is the minimization of cost of capital. Another one is the profile of debt repayments that we try to manage over time, and also minimizing our covenants associated with those financing instruments.
If you look at our convertible bonds, we use both structure as well as anti-dilutive capped calls to minimize cost of capital and to manage the payment schedules. On our leases, we use our property, plant, and equipment assets to get very favorable terms. We find these give us the best cost of capital and best structure for the overall company. I would add that we just last month secured a three-year, $255 million AR-backed revolving credit line at very favorable terms and with no covenants on the borrowers. It's available to us now for short-term cash cushion or short, medium-term capital needs as we go forward. I'm sure you've all noticed that in current markets, the straight debt markets are pretty available right now and quite affordable.
We at Micron have a practice, I mentioned, to minimize covenants given the volatility of our market environments, we have not historically engaged in straight debt kind of financing and probably would not do so unless they were essentially covenant-free. That's roughly the structure we have in our overall financing at Micron.
I'll move on to the next question. What is the maintenance CapEx, and what do you need just for shrinks without adding new capacity? How much bit growth does this imply for DRAM and NAND? Mark, I might let you kick that one off.
Okay.
There's a slide up here, which is historical directional data relative to CapEx spending. You can see that there's sort of a dotted long-term trend line on there, as well as some peaks. The peaks, obviously, associated with when you add greenfield capacity, you're going to spend more capital because the capital dollars for greenfield capacity are obviously much more significant than when you're adding bits relative to technology migration. Keep that in mind. Relative to those numbers themselves, by the way, just some sort of general markers for you. I think I would say DRAM for greenfield capacity is in this sort of $50 million for 1K per month around for greenfield capacity, and NAND probably closer to $40 million for 1K per month.
When you talk about a go-forward basis for scaling technology, the CapEx associated with that is lower on a percentage basis than it has been historically for your installed base. I think we've talked about all the reasons for that. One is technology's migrating a little bit more slowly, longer between the nodes. The other is lithography, as Scott just talked about, is really stagnant right now. As we move to new technology nodes, we're not replacing existing lithography solutions with new lithography solutions. We're adding small amounts of incremental capacity to handle slight additions to the overall number of more difficult layers.
The net effect of those trends is that the long-term capital associated with upgrading capacity, be it NAND or DRAM, is more of a 5%-10% of greenfield capacity addition number than it is something that's maybe closer to 15% that we've seen historically. Of course, it always varies depending on what your install base is at Micron's 30, moving to Micron's 20 or Elpida's 30, moving to Elpida's 25, and the size of all those things. Think in terms of a range of 5%-10% of the greenfield cost of the installed base to keep that current. You can work through the math. I gave you sort of the greenfield numbers to think about.
What this means is that just keeping up with technology today and staying at the leading edge is not as demanding a treadmill as it has been historically by a significant increment. On top of that, obviously, we have CapEx that's associated with. Something to keep in mind here is, we talked about fungibility. Taking DRAM capacity and moving it from state-of-the-art DRAM capacity to state-of-the-art NAND capacity actually turns out to be a lot cheaper than taking state-of-the-art DRAM capacity and migrating it to the next DRAM node. Within that 5%-10% range, it actually can be on the low end of the range when you're talking about transitions from one technology into the non-volatile area. There's a bunch of other things we spend capital on, obviously, in the business.
Maybe, Ron, you want to chat a little bit about some of those?
Sure. Just to quickly break down our CapEx structure. In a typical annual CapEx budget, we run in the range of $300 million-$500 million of our CapEx is what I characterize as maintenance CapEx. It's R&D capital, IT infrastructure, all the other miscellaneous capital applications, maintenance capital, et cetera. For tech node transitions, we typically run, and it varies a lot, partly for the reason Mark mentioned, that DRAM NAND mix matters, and in fact, that slope of that line for Micron is partly affected by increasing mix of NAND over time. In general, for tech transitions, on average, it runs around $700 million-$1 billion per year, current run rates for tech node transitions. That's the way to think about it. I think another part of the question was, how much does this apply for growth that you get from tech node transitions?
Rough numbers, DRAM, we get low 30% kind of bit growth from our tech node transitions, and you saw the projections from the BU guys that DRAM is in the mid to high 30s range, and that's why the view about not maybe needing a lot more wafer capacity comes into play. On the NAND side, the bit growth from tech transitions runs around high 30s%, and projections for NAND bit growth in the market are in the high 40s kind of range. It would infer probably some more wafer capacity over time. That's a rough way to think about our structure. Obviously, our capital budgets the last couple of years had another item, as Mark pointed out, and that's additional capacity expansion, notably on the NAND side of the business. Okay. The next one is a complex question.
How do you plan to improve your capital efficiency and ROIC? I've got a slide here. First of all, we're obviously not where we want to be for the long run in terms of financial performance. I thought it'd be useful to provide a comparative here. This graph shows free cash flow for Micron in blue and some other relevant pure play comparisons over the last 12 months. This is operating cash flow minus CapEx. Even in the challenge years, Micron does reasonably well. You can see it's performed significantly better than Hynix. Even though DRAM was a more challenged business, this last year was very comparable to SanDisk, which, if you compare on operating cash flow basis, Micron actually did better. Obviously, we have more CapEx because we have all the fabs in our consolidated structure.
At any rate, just a point of comparison for you. In terms of what we want to do going forward to improve things, obviously, want to continue to improve the operating performance of all of our operations, as Mark Adams commented on earlier. I would just add that we strategically focus on joint venture leverage and mentioning a couple of key ones for us that have been brought up several times. The Intel joint venture, which was recently expanded and extended, gives us significant leverage both in terms of manufacturing scale as well as R&D leverage. That's a valuable addition of capacity for Micron as we purchased that this last year and have more trade bits that we can move into the market at trade margins. Another one is in Inotera, which we bought in 2008. Their performance is significantly improving in the last 12 months.
As you can see on the graph there, they also have positive free cash flow in the last year. Their output performance and yields are competitive with our other fabs today. Good progress there. Obviously, you've heard comments about premium product mix and our multiple technologies and the leverage we get from those activities, and a lot of comments about how we expect to improve on that going forward with an expanded footprint. That's clearly a driver of improved performance as we look forward. I would mention, I guess one last thing, that is that Micron has traditionally leveraged its capital capabilities and its scale through acquisitions. If you look at one of the more recent ones, I already mentioned in the Intel acquisition of capacity, but Inotera, that was just over two or so years ago, we bought April 2010.
In the last couple of years, we have generated $3 billion in incremental revenue. Just from the NOR part of that acquisition alone, we've generated $1 billion in free cash flow, and that compares to a net purchase price for the entire acquisition of about $550 million. Obviously, NPV on that is in the neighborhood of $300 million at our cost of capital. We focus on opportunistically investing in acquisitions as part of our strategy. You heard from the BUs about how that really leverages our business and expands our capabilities. Looking forward to the Elpida structure, typically, as Mark mentioned earlier, we find that if you take the full purchase price of the acquisition, plus the cost to convert to the merged technology, whether it be DRAM or NAND, is about half the cost of acquiring greenfield capacity.
That's one important baseline as we think about our strategy and leveraging acquisitions. Another piece, and this is obviously in the very formative stages, but our initial assessments would indicate that the cost to convert Elpida capacity to a merged DRAM roadmap or Micron's NAND roadmap is roughly equivalent to what we'd incur converting Micron's capacity to a next tech node. It's within the range of our estimates right now. Important data point is we think that it'll be great economic leverage for us going forward. In summary, I'd say we've got a lot of opportunities to leverage our return going forward.
Yeah. Very thorough. Not a lot to add there. I will say that, I have read from time to time reports about the Elpida situation and potentially large amounts of capital required in order to make that capacity competitive. I'd just like to reiterate what Ron just said, which is, even if we were to take all that capacity and move it to the next DRAM node, it is not significantly different than what we would be doing with our own internal capacity. Obviously, we've talked with you earlier today about no matter where in the network, it may or may not occur and whatever the timing is, clearly over time, we're going to need more NAND and less DRAM, which is cheaper.
Yeah. Good. Next question, please discuss your access to capital situation in light of planned investments. Our capital markets, from our vantage point, are liquid and healthy and certainly open to Micron today. In the last fiscal year, we raised $1.6 billion in financing, $1 billion in a publicly traded convert, and about $600 million of various forms of lease financing, including leasing up some of the IMFS assets after the acquisition of IMFS, or that's IM Flash Singapore, I should say, if you don't know our acronyms, from Intel. Another thing I'd mention is we proactively addressed our 1.875% convert that's coming due in 2014. As you probably know, we've exchanged and repurchased some of that in the past. We're continuing to look opportunistically at taking other structural actions with the remaining approximately $900 million of that convert.
As we go forward, we may or may not make some moves in terms of restructuring or exchanging part of that. Another element in terms of our capital situation in light of planned investments is we've been very transparent with the rating agencies, and in fact, they've expanded our involvement there. We've had S&P ratings for a while. We recently added Moody's rating. They both rate us BB-, S&P with a negative outlook and Moody's with a stable outlook. We also recently added R&I, the largest Japanese rating agency, who rates us BB+ and stable. We're beginning to increase our connections there as well. I guess the last thing I'd mention is that, in light of planned investment activities, we have typically had access to low cost of capital in support of our acquisition activities.
We did that in 2008 with the Inotera acquisition and got some low capital as part of the support of that arrangement. Likewise, with the purchase of capacity from Intel, we got favorable financing arrangement. As we go forward and look at Elpida structure, we have some favorable financing structures associated with that as we look forward to that acquisition. I don't know if you had anything else to add to that.
No. Thanks, Ron. Go ahead.
Okay. The fifth question is, how will the Elpida deal impact credit metrics? Sort of a segue from my earlier comment. The short answer is that the Elpida acquisition is viewed generally as a very positive catalyst for Micron by the credit rating agencies. They view, on a pro forma basis, the combined Micron plus Elpida, as you've seen, will be the number two pure play memory supplier, that we have got significant scale, which has been an observation from prior ratings reports, that was an issue for Micron. It would also provide a stronger technology roadmap and significant leverage across our total infrastructure, as well as some of the other comments you've heard about being able to move more volume into premium markets such as server and mobile and integrated solutions with mobile MCPs, et cetera. Lots of opportunity viewed there.
Also the fact that Micron's back-end assembly and test is a big synergy as we look at the combined capacity, because it's a competitive advantage for Micron, and we'll be able to leverage that across the Elpida volume as we go forward. Lots of positive catalysts from the rating perspective. Do you have anything else you want to add?
Nope. You're doing great.
I guess I'd just add one thing, that is that Micron's still committed to having a very healthy capital structure. If you look at the Elpida financing construct, it's interest free, it's spread out over 7+ years in terms of the repayment schedule, and it doesn't have covenants. It fits nicely in terms of keeping our capital structure in good shape. Also, our total capital structure across all of Micron is covenant-free. We have focused very hard on ensuring that's the case. We've also focused on broadening our financing base, we have a lot more financing around the globe, consistent with our increasing global footprint. Finally, we've worked a lot on cash management to make sure it's all accessible and also fungible to move it where we need it, all of which are the case in our structure today.
In addition to that, we focus a lot on liquidity management. I think you've already heard that our capital investment strategy is somewhat discretionary. We can back down capital investments if we need to in weak environments. Also, you can see, if you look back through history, Micron has consistently, every quarter, produced positive operating cash flow, even in the most difficult market times. We continue to focus on liquidity and ensuring that we have the right balance sheet that we need to manage our business going forward.
Did you want to mention anything about recent credit activity you undertook?
Go ahead.
Recent deal in Singapore. What was it, the AR?
Well, I mentioned.
Oh, you did earlier. Okay.
I mentioned that in the outlook. That was the credit revolver I mentioned earlier. Yeah. Okay. Let's see. The last question is, what are the mechanics to account for Elpida and Micron's book value and future depreciation? Basically, under current accounting rules, the net asset values are fair valued in an acquisition. We have not done this work yet, but it will be future work to be done. They're fair valued. If the purchase price is less than the fair value, then you recognize a gain at closing, similar to what happened in Numonyx, and you put the fair value on your balance sheet, and then that is depreciated over time if appropriate, if it's a depreciable item. We anticipate, similar to Numonyx, where we had a fairly significant gain, that this transaction, we will also have a gain.
We just haven't done the homework on that to figure out the range. With that, I'll open it up to questions.
Yes, sir. I've been talking about fungible capacity all day, going back to your capital intensity, how much would it cost to migrate a 30 nanometers DRAM to 1x or 1y nanometer NAND? I'm assuming that that's the future of NAND migration a year from now.
It's in the range I talked about, it's in the low end of the range. 5%-10% of Greenfield installation cost is the right number to think about in terms of taking a leading-edge DRAM fab and converting it to NAND. That same range also covers taking an existing leading-edge DRAM fab and converting it to the next DRAM technology node.
Think about the bit capacity or incremental bit capacity added when you make that conversion.
You have to take a look at the 1y, I guess, is what we would be talking about, a NAND node. The bit per square inch of silicon there is obviously dramatically different than the 3X or the 2X DRAM, bits per wafer you'd be converting it from. Because it's on the order of an order of magnitude or more, you almost discount what you had in terms of DRAM bits and just think in terms of you have that much new NAND capacity created.
Can you guys take a second and just remind us what your CapEx budget is, and then if you could offer some color onto when you'll make purchase commitments or when you'll place the bookings for that CapEx across the next four quarters?
The next fiscal year CapEx budget is about in the range of $1.8 billion, $1.6 billion-$1.8 billion, and it's a little bit front-end loaded in terms of the first half. In terms of purchase commitments, we're already in our fiscal year, and some of that's already occurred. That's the rough range we're dealing with.
Okay.
Hi. If I could just question about the NAND, the bit growth without wafers, you said the high 30s. It seems to be quite a lot lower than historical and quite a lot lower than I expected and what I've been hearing from other players. Can you clarify that and why is it so much lower than historical?
Sure. Well, again, I think the trend is going to be longer between nodes. It is true that many people are talking about making a more significant step from a roughly 30-nanometer node down to 20, skipping a 25, and the timing of that is actually getting pushed out. You have to look at not necessarily what are they going to do in their next increment, but what are they going to do on a time-weighted average as they move forward. I actually believe that the numbers over the longer haul is actually lower than that, not higher than that.
Again, it's the challenges are relative to scaling of DRAM, which are around capacitor aspect ratio and how you even drill that hole and how do you store enough electrons on it at the low voltages we're operating at. All those types of just very difficult scaling issues moving forward are going to drive that economics.
As you come closer to the Elpida acquisition, I wanted to find out what is Micron's real strategy on the mobile DRAM side? Are you going to still pursue your product offering, or how is it going to change the landscape for you guys? Thank you.
We've been pretty public about saying we think Elpida's got a good low-power DRAM offering. We also now have, as Mike Rayfield was talking about earlier on, we have a pretty good 30-nanometer low-power DRAM offering ourselves now, so we have some optionality there. Elpida's got a good follow-up roadmap with their 25-nanometer node. Clearly, given where we are and the uncertainty relative to the exact timing, they will continue with their existing product roadmap and product and technology migration prior to close, and we have this joint development program in place that allows us to make sure that that offering post-close is as cost-effective as possible and take advantage of some of the back-end efficiencies Ron was talking about, et cetera. We will run, for a period post-close, a non- low-power DRAM portfolio and a Micron low-power DRAM portfolio, just because it's inevitable.
Because while we have a high confidence in close, we don't have 100% confidence yet. Anyway, between now and then, we have to deliver products to our customers and couple them with our own internal NAND portfolio. Longer term, yes, we'll have a merged roadmap, and that could happen as early as 20 nanometer, if we are nimble. Their process flow, for those of you that like to dig into this kind of thing and look at cross-sections and stuff, is not that different than ours. It's amazing how, and this is sort of the funnel we're all being forced through that's also driving this slowing and the rate of technology migration. If you go back 10 years and you looked at everyone's designs, they were all very different relative to integrated process technology.
We're all being driven down a common funnel here as this thing gets tougher and tougher. When you look at what's going on at Elpida from a process technology perspective relative to what's going on at Micron, the roadmaps are already reasonably well-aligned, and that's a big piece of why Ron says early indications are that to convert an Elpida fab to the next technology node are not that different than what they would be for our own internal roadmap. To the extent we collapse those two, it becomes even more true. Hopefully, I answered your question. If there's a follow-up, happy to take it.
Yeah. There's a report in the press which says that Rexchip was not able to pay its banks on a timely basis, and they were looking to try to extend those loans, right? Given that you're paying close to $300 million for equity for someone who doesn't have money to pay his debt, is there a risk that Rexchip files for bankruptcy? What does that mean for Elpida's stake in Rexchip and your potential stake in Rexchip, and how does that impact your valuation on Elpida?
Let me first make sure I clarify that Micron's right and obligation to purchase Rexchip shares only happens at a closing of the Elpida transaction. Prior to that, if the train comes off the tracks, well, all bets are off. I'll let you talk about Rexchip's situation here, Ron.
Yeah. I may need to give you a one-minute piece of background in case you're not all up to speed on this. One of the interesting structures in this sponsor agreement is that we are actually providing pre-closed financial support for Elpida. Support means we will help them, which is kind of interesting in a pre-close with antitrust issues, but this is a situation where we're actually helping a competitor get financed. Why do we do that? Because we want to make sure, as Mark said, they're on the roadmap that we want them on. There's DIP financing and there's CapEx financing that was provided for.
It so happens that that DIP financing is already in place, debtor-in-possession financing, and the CapEx financing is in place or being put in place, and ideally, it'll be done by Elpida itself with some support from Micron, if necessary, to be determined, but it would be Elpida-level financing is the intent. As part of that arrangement, for a number of reasons, I should also mention, we can't by Japanese law, actually do direct financing into Japan, it has to be done in other forms of support from Micron. Also as part of the arrangement, it was specified that our support for Elpida would have to also transmit, they would have to provide support into Rexchip if and when it was needed. I think both companies, Rexchip and Elpida, are doing what they can to keep their financing intact.
We are actually interacting with their financial teams on a very regular basis, almost daily, which is unusual, but because of these financing interconnects. Our anticipation is that Elpida is working with backstop support from us to get necessary financing to Rexchip, and those approaches are being evaluated. Rexchip's doing all they need to do to try to get their financing taken care of, and that's probably where you heard a press report, but I wouldn't draw any grand conclusions from it. I think it's all being worked by the teams.
Let me just add one more comment on that, which is, I'm not sure how public Rexchip's finances are and probably shouldn't comment too much on them. Rexchip itself is a pretty healthy company in terms of got a lot of assets and not much debt. To the extent you're reading something about some issue relative to a bank repayment, it's likely a relatively small liquidity issue as opposed to a financial viability or issue where you would say, "Hey, I don't want to put my equity into this thing.
Sure.
Okay. Just coming back to some of the comments you made earlier on the two tracks for mobile DRAM, low-power DRAM for Elpida. If I understand you correctly, initially, you said that after the close or at the close of the deal with Elpida, you would start converting everything to Micron Technology. If I hear you correctly, it's probably more likely that Hiroshima will do mobile DRAM on Elpida technology for probably a full node after that, then everybody comes together in the middle of 2014. I guess in the meantime, you probably will convert Rexchip to your own process. Am I reading correctly what you said, or am I extrapolating?
No, I don't think I said any of that. I've been careful not to say exactly what we're going to do, or at least I've tried to be. It is for sure the case that Elpida and/or Rexchip, because really we think of them as this is a deal that will close simultaneously, and we look at that capacity as one big chunk of capacity. Between Elpida and Rexchip, they're going to be providing low-power DRAM in the marketplace for quite a while, and where that might happen, I can't tell you. As Ron said, we are today supporting Elpida, with backstop financing to enable capital migration along their own process technology roadmap, which will support their existing low-power DRAM ramp, as well as next-generation node low-power DRAM ramp.
That doesn't mean that post-close, we wouldn't take some piece of the capacity and immediately start moving it towards something else that's either Micron or a NAND or more of our merged roadmap. I've been very non-specific about that.
Sorry to have been specific then, I was trying. With your significant operational background, what I don't quite understand is, let's say you decide to run a low-power Elpida process until mid-2014, for argument's sake, would you actually run a different process in the same fab? I've actually never seen this in the past. It's probably why I'm struggling a little bit to imagine this kind of hybrid process flow creature, I guess, for you also.
I'm not sure I understand the question. Let me try and address it this way. Elpida has an existing technology node that they're running low-power DRAM on. They have a next node that they're going to run low-power DRAM on, and that next node is their process technology node. We are doing work with them today. We have entered into a joint development program that will allow them to implement certain pieces of Micron intellectual property that will make that next technology node, and those next low-power DRAM products, more cost-effective than they otherwise would have been. We're not going to get into the nuts and bolts prior to close of trying to modify their process technology roadmap, but we certainly are going to and are working with them already to help make sure that we hit the ground running at close with the most cost-effective possible solution.
Is that helpful? Last question, guys.
Thanks. There have been some reports that would indicate or say that you guys overpaid or are overpaying for Elpida. When you look at the acquisition cost, all in, including the node transitions to get onto the Micron technology, relative to historical transactions, it seems that the price you're paying is higher. In addition to that, the DRAM market has not only deteriorated but deteriorated somewhat significantly since you originally announced the acquisition. How do you respond to the thought that it appears like you might be paying more than you need to, frankly?
Let me answer that first, and then maybe, Ron, you want to add some additional comments. First of all, it is undoubtedly true that the market is much weaker now than it was when we signed the deal. For those out there that maybe think that they're going to intervene in this process and get a better offer or something, that's just Looney Tunes. That will not be happening. Having said that, we still think there's compelling value in this deal, in terms of the value of the assets, and we've talked all about why we believe those assets are valuable to us from a go-forward basis. Albeit, we don't think it's as good a deal as it was when we struck it. We also think there's a lot of other value elements that are valuable, frankly, to Micron, but maybe not to others.
There's a lot of value elements we bring that actually make the assets valuable that wouldn't be there otherwise. For instance, the capacity is a lot more valuable to a Micron than it would be to, really, anyone else in the world that's not a memory supplier, because they don't have a roadmap or an ability to help Elpida in the interim develop a more compelling forward-looking roadmap, et cetera. We're uniquely positioned in terms of being able to provide more value to the creditors when we make this acquisition. We're also uniquely positioned in terms of being able to leverage what Elpida has to the benefit of our shareholders. That's why it makes sense to us, and that's why we're still interested. Are we going to renegotiate the price? We have no intention to renegotiate the price of the acquisition today.
It's a bankruptcy process, as you might imagine. It's very complicated. We'll have more information when we get to close. As long as Elpida continues to execute to its interim business plan and the wheels don't come off the world economy, we still think this is something that's going to make sense for Micron. It's because of all the value elements we talked about, not just the assets. There is a lot there that hasn't necessarily been there in the past when you just pick up a pure capacity play, like a Manassas fab from Toshiba or whatever else. Ron, did you want to add anything?
I think it was good coverage. I think if you look at it's very valuable to Micron at the prices we're looking at right now for all the assets we're considering. When I made the comment about the accounting, that we actually go through a process of fair valuing every single element and then comparing it to the purchase price, and that's why I gave the indication that even in today's environment. I think it's quite competitive with prior acquisitions. As Mark mentioned, we're buying a lot more than just fab capacity. Their technology capabilities, their customer connections, their product portfolio, there's a lot of value there. It's a good arrangement.
Is that we're all done? All right. Let me just thank you all again for coming. Hopefully, you like this format. Obviously, you'll get a lot of opportunity to give us feedback and whether we should try it again or ping-pong it or whatever, we're open to all your suggestions on that. Our goal, obviously, was to make this as interactive as possible and try and provide the best possible information to you, and hopefully, you got that out of the day today. Thank you all for coming. It's been a great pleasure to see you all again, and we'll look forward to seeing you again in the spring. Thanks.