Well, good morning, everyone. Thank you for coming to Micron's Analyst Day. It's a real pleasure to be here. We try and get over this part of the world on a regular basis, and it's been a little longer than we would've hoped since the last time, but we're very happy to be here today to share some information about Micron Technology and our path forward. I'm going to just kick off here with some high-level overviews relative to our view of the memory industry and Micron's investment priorities, what we're focused on a go-forward basis. We've got a number of our executive staff here to share some information with you today. Following me will be Mark Adams, with a little bit more detail on our operational priorities and what he's focused on in terms of driving efficiency, new products, and customer satisfaction moving forward.
Scott DeBoer from our technology development organization's here to share a little bit about our view of technology roadmaps going forward and some of the important things Micron's working on. We've got Brian Shirley to talk about all the various market segments Micron's addressing and what our priorities in those areas are. Introducing for the first time today to you in the investment community, Darren Thomas, who's the new leader of our storage business unit. We get through all that. Ron Foster, our CFO, will come up and talk about how Micron's planning on allocating its capital moving forward and some of our financial performance and plans going forward. I will have a relatively brief presentation up front. I will hold Q&A for myself until after all the various executives have an opportunity to talk.
I'll be back at the end of the presentation to do a little wrap-up and answer any questions you have for me. Without further ado, if you look at this slide here relative to DRAM supply and our view of it on a go-forward basis, you can see a continued projection of relatively muted DRAM supply growth going forward. If you look at 2014, that number has been adjusted up a little bit relative to where we had it before. Our view is that that's going to come in about 30%, and that's a little higher than we had previously indicated we thought supply growth was going to be. The net result of that is that we think the go-forward supply growth over the next 6-18 months is going to be a little bit lower.
Again, that 30% is probably weighted slightly to the front half of 2014 as opposed to the back half of 2014, with some good performance by really most of the competitors in the marketplace in terms of bit growth in the first half. As we look going forward, you see this compound annual growth rate for DRAM running in roughly the 21% range. By the time we get to the end of the day, and you see what we think is going on in all the various market segments, I think you'll come to appreciate that's a very reasonable picture relative to overall market health. You look at the right side of the graph here, and you look at where the growth is going to occur, obviously, mobile growth will continue to be very strong. Our view of enterprise is that it will also be very strong.
Two big components here, obviously, the cloud and the on-premise bit growth in the enterprise, both of those big contributors. We've taken a relatively conservative view as to how big those growth numbers can be, Brian Shirley will comment more on that on a go-forward basis. The overall story here is strong growth in both enterprise and mobile, muted demand, leading to overall, I think, favorable market conditions for DRAM on a go-forward basis. When you switch and take a look at now what's going on from a NAND supply perspective, again, you can see the CAGR on a go-forward basis, 31% and declining as we move through time.
Really a view that most of the growth in NAND bits is going to come from technology migrations first to the 1X and 1Y nodes, then on beyond that into 3D, with a capital intensity that drives a reasonable supply growth on a go-forward basis. The story here is really all around what's SSD penetration going to be like. There is elasticity in a couple of these segments at least, that's fairly significant. We don't see how this gets out of whack in any significant way for any sustainable period on a go-forward basis. We see a really very healthy demand in both the SSD arena and in mobile to drive whatever supply is eventually supplied to the marketplace. The story I think that we should think about is a memory industry that's really in a very beneficial and virtuous cycle today.
We have had significant consolidation across both all the market spaces, the suppliers in the market today all have sufficient scale in order to drive an effective and efficient business model going forward. Our view is that we'll continue to see what we've seen over the last couple of years here, which is really a return-focused approach to investment by the incumbents in the market, one that drives a supply that's in line with customer and demand to meet customers' needs, but not market share driven. We see limited new capacity in that environment coming into the marketplace, with the slowing technology migrations that are really a factor of what's accomplishable at a reasonable price given the technology challenges ahead, we see relatively in line to an undersupplied market on a go-forward basis.
We see pretty explosive demand across diversifying market spaces as we start thinking about not only what's the growth in mobile and in enterprise, but what's the growth in all the various Internet of connected things, in the automotive space, in the industrial space. Really a very virtuous cycle in terms of demand growth that's broadly distributed across a broadening set of end applications, in some cases, a broadening set of customers of significance. Some of the other guys, Brian Shirley, will come up and spend a little bit more time on that, as will Darren Thomas. The net result, I think we're pretty excited about being in the memory industry today.
I've been in this industry for about 30 years, and while it's always been challenging, rarely have I seen the opportunity for sustained growth and the opportunity to drive value-added differentiated solutions in the way that we can today in an overall market environment that's very healthy. As we move forward, we couldn't be more excited about the opportunity that's afforded to Micron to go out and service our customers. If you think about us in the marketplace today, we're a truly global company. We have great customer relationships and great engineering and technology development distributed around the world. We have capacity. We have technology to be effective and efficient, and really, a very positive situation for Micron. Our focus in that environment is really on three things.
First of all, we want to make sure that we're doing a good job for our customers by having the right technology available, not only from a product enablement perspective, but also in manufacturing capacity, to deliver the solutions that they need when they need them in a cost-effective way. Number one on our priority when we think about what does Micron have to do going forward, we have to continue to develop great technology and great products and memory solutions that we can deliver to the customer in a cost-effective manner. That involves investment in both the manufacturing capacity, not to create new capacity, but to create the capacity that has the performance characteristics that the customers demand for their end products.
We have to continue to invest in system-level capability, value-added solutions, so that as this market, over time, continues to de-commoditize, we have the right products based on the right customer intimacy to meet the customer needs. Micron today has been very focused on adding incremental capabilities throughout the team at the management level, all the way down to the entry-level engineers with respect to our ability to deliver memory system and memory subsystem solutions. We do that both organically and on a go-forward basis. We'll think about where inorganic growth makes sense. At the end of the day, we want to be a memory subsystem and system solution provider. We want to deliver those solutions that really differentiate our products relative to the others in the marketplace. We believe that's the path forward to success.
Finally, we'll do all that with a view to the shareholders. Our company really has historically been lean and mean. We want to stay that way. We want to spend the shareholders' money very carefully to achieve these objectives. When we have extra left over, as we anticipate in this market environment we will do, we want to look for ways to return that excess capital to shareholders as we fulfill those first two priorities. The relative priority of these three boxes will move around as we move through different points in time. When you think about what Micron has done over the last year, we've been very proactive about returning excess capital to shareholders, and Ron will give you some more information on that here when he comes up and talks a little bit later.
That will continue to be our thought process as we move through time. What does that mean relative to how Micron is going to fund capital as we move into 2015? We want to give you a forecast relative to our capital spending to support the business in 2015, and we think it's going to be in the $3.6 billion-$4 billion range. These numbers are always tough to pin down exactly in terms of where in the range we might fall, and we want to maintain flexibility to adjust the range relative to market conditions. Having said that, we think the number is going to be something like this or in this range in order to support those priorities I just talked about. It's important to note that the focus here is not on growing bits or adding capacity or gaining market share.
The focus for Micron in 2015 is to make sure we have the right technology in place and we have the right products in place and the right capacity to manufacture those products in order to support the customer end needs at the right performance points, power points, to drive their businesses forward and to do it in a cost-effective way relative to competition. When you look at these three big buckets of what Micron is going to spend money on in 2015, there's a technology and product enablement bucket, there's a bucket for non-volatile memory deployment and for DRAM deployment. Think in terms of those numbers being roughly 30% for the top bucket, roughly 20% for the middle bucket, and roughly 50% in that DRAM category at the bottom. Those numbers are not precise.
We shaded the colors on purpose because we're going to be market reactive as we look at what's going on in the marketplace, but those are ballpark what you should think about. In the technology and product enablement space, there really are a number of things we need to do to service these memory systems and subsystems that I talked about. That includes things like new memory technologies, cost-effective ways of building advanced hybrid package solutions to support the memory systems and subsystems that we want to do, as well as things like the Hybrid Memory Cube, where we have advanced high-performance memory solutions for enterprise and networking. We want to continue to make sure we're investing in the technologies of the future. Outside of NAND and DRAM in this bucket. We've got capacity, low-level capacity, and early pilot line activity around future memories.
Scott DeBoer will talk a little later about what some of those technologies look like and the market segments that they might address. Finally, we've got a bucket that facilitates all the other activity that has to go on in the company to make sure that we can deliver increasing volumes of SSDs, or increasing mix of more complicated products to the end marketplace. In the non-volatile market, we've got the completion of the build-out of our 16-nanometer NAND planar technology as we continue to migrate from 20-nanometer to 16-nanometer. Roughly two-thirds of the bucket there is the first installation in capacity relative to a 3D NAND conversion that'll take place over a number of years on a go-forward basis.
To get that process off and moving, there is capital spend in 2015 that puts the initial increment of 3D NAND in place so that the technology learning and the early socket enablement and customer deliveries can take place in the second half of 2015. Finally, in DRAM, there is the completion of 25-nanometer conversion that's been ongoing throughout 2014, as well as early production ramp of 20-nanometer technologies across Micron's product portfolios. The thing to keep in mind, again, is as we move through the year, we'll continue to monitor market conditions. Our approach here is really around market return, not around market share, and making sure that we're putting the right capabilities in place to deliver value-added products in a cost-effective way.
I'm going to stop here and turn it over to Mark Adams to talk a little bit more about what's going on operationally, and I'll be happy to come back later on and take any of your questions. Thank you very much.
Thank you, Mark, and good morning. Appreciate you joining us today. I wanted to pick up on the theme that Mark had talked about, is how we're looking about the business going forward and how it's different than we may have looked at it in the past. I'm not sure I need to get up and give you a big presentation on the dynamics and the big trends of driving growth in technology today. What's interesting about this is that if you look back in the memory business maybe a decade ago, where memory was in terms of the bulk of the capacity in the industry, it was pretty much of an industry-standard product portfolio and a very narrow set of applications.
When you think about today's big technology drivers, the trends in technology, it's really the catalyst for why we think the memory business is a different business going forward. You think about something as simple as where networking is evolving to and the impact on growth in terms of infrastructure enablement for mobile and corporate enterprise applications. You think about embedded and machine-to-machine type applications and just the massive scale-out and growth in terms of smartphones and mobile and the application base there. Certainly cloud computing, changing the architecture and how the world is connected, and over the last couple of years, the growth in big data applications. What's unique about this for us in the memory business and for Micron as we look at our business is a dramatically different way of doing business for us.
When we look at serving our customers in these end-market applications, we're engaging with defining and innovating solutions that help our customers on a unique one-to-one basis. There is no industry-standard part in subsystem solutions that we're developing that spans across all these. Each one of these categories has different levels of technology and not just the silicon, things around the silicon that allow us, again, to empower our customers' different solution approach. From our perspective, these trends in technology are really driving a fundamentally different opportunity for Micron going forward. How does that relate to Mark's mentioning of a returns-based approach? Well, when we think about how we're going to invest in 2015 and beyond, we're looking at things that have a dramatic impact on the return side of our business.
We're no longer investing for, quote, "market share" or just to get into a capacity battle. What we're trying to do with our business is to advance either from a cost position or we are going to invest in capabilities that allow us to differentiate and actually improve our margin opportunity. When I think about things like technology development and deployment, in the past, you would do this as you added capacity. Our focus is not necessarily driven by capacity, but cost. When we think about our business going forward in terms of core silicon and process enablement, we think about how do we enable ourselves to be a cost leader at the silicon level. The other side of it is we're investing for tomorrow.
Mark talked about new memory technologies, and I'm not going to steal too much of what Scott DeBoer will get up and talk about in our roadmap in future years. We're continuing to invest, again, on driving capabilities to differentiate and drive margin, not driven by capacity. Beyond process technology at the silicon level, we're trying to drive our capabilities in manufacturing and not just front-end manufacturing, but all the way through back-end test and assembly. When I put the slide up earlier, it talked about the big data trends in terms of the technology industry. You can imagine that each of these businesses has unique customer requirements, unique supply chain requirements, product differences in terms of how we build the products and what the solutions look like. Again, that's fundamentally different than a DRAM module company from 10 years ago.
Our business, as we think about manufacturing, and as we think about all the way out to supplying our customers with their products, is a different supply chain model all the way back from how we build it to how we ship it. We're trying to invest in our company to be world-class, not just at the technology level, but how we drive our capabilities out to the customer. Lastly is Mark talked about subsystem and systems-level solutions. The memory business and how it's evolving is allowing us to drive the core technology we develop and drive these into unique solutions that power our customers. When we think about solutions in each one of those market segments that I identified in terms of trends, when we think about mobile's not a core piece of silicon that we just sell to the market.
Each one of our customers we're engaged with in developing products and technologies and roadmaps that drive not just one piece, but multiple pieces of our product portfolio and our technology portfolio to solve a customer's problem. That's true with storage. That's true with just about any piece of our business as we go forward, and we're going to continue to invest, again, not necessarily driven by capacity, but capabilities to differentiate our products and to innovate with our customers to drive a different business model in the memory business. If I take each one of these individually, Mark talked a little bit about where we are today in terms of executing on today's investment.
In DRAM, we will continue to drive our 25-nanometer DRAM conversion on a global basis, and that requires a lot of execution on the team and capital in 2015 to finish that roll-out and will allow us to drive better cost performance. On the NAND side, likewise, we'll continue to drive our MLC 16-nanometer products into the marketplace and drive more of a differentiated approach in terms of some of the unique products we have in the NAND business. What I talked about earlier about solutions, it's not just DRAM, and it's not just NAND. Our capabilities as we look at controllers and firmware and even how we package and stack silicon to each individual end market capability are things we're driving today.
If you think about eMCP for mobile applications or eMMC, where we take a controller and take our silicon and bundle that into a product solution, we have a unique opportunity to drive not just a commodity approach, but a differentiated approach, and that's the things we're investing in today. Beyond those investments, we have a lot going on in the company as far as increasing our cost capability and our margin capability. On the cost side, we're going to be continuing to roll out and drive improvements in our DRAM business. By the end of this year, we'll be sampling our 20-nanometer product, and we look to put that into manufacturing ramp in our 20-nanometer DRAM technology during calendar year 2015.
On the NAND front, Scott DeBoer will talk a little bit about our enablement of 3D NAND, and we think we have a great story there and a big winner in terms of allowing us to bring innovative technology to the market, as well as we will be rolling out TLC technology for low-cost applications, which we think will allow us to strengthen our NAND business. Finally, Micron has been an innovator on the packaging and silicon interface capability in terms of a number of different applications. Many of you will know that we've been a driver of the Hybrid Memory Cube opportunity in the industry, and with that, dandelion technology that allows us to succeed in that type of business. When you think about technology and our investments at Micron, a lot of it's around, again, cost-driven benefits, margin enhancements, and capabilities.
A returns-based approach to that as opposed to what we used to think about in terms of investing for capacity and market share. All of our investments in these technology areas are really driven to improve our financial performance and improve our capabilities to deliver innovative solutions. How will that show up? If we do as we say we will do and execute, you will see us innovating in a whole different set of applications going to the market. For example, in the mobile segment, you'll see us innovate from low-power DRAM all the way out to taking, again, core silicon and hiding that behind a controller and allowing us to meet customer capabilities and requirements for mobile applications.
You look about in terms of enterprise DRAM, Brian Shirley will talk about that shortly, you think about taking stacked packages of DRAM and interfacing that with a Hybrid Memory Cube solution to differentiate and allow better bandwidth performance to the networking and server space. In NAND, as we think about long-term success in NAND, we think about how we take NAND core technology, again, using the three-bit per cell technology and controller technology firmware and how we deliver innovative solutions for storage and other market applications. You look at embedded today in the automobile, we think that NAND will play a specific role in providing unique solutions in that market.
You'll hear Darren Thomas later today talk about enterprise storage and what we think about our capabilities around not just the core NAND silicon, but our firmware, our controller, all the way out to the full end systems capability we need to be successful in that business. Those are the type of things that will allow us to differentiate and innovate with our customers, and those are the things you'll see us invest in. Not capacity, capabilities, and cost-driven. In the area of manufacturing, we are continuing to strive to be the world-class leader in manufacturing in the memory space. We continue to look at ways that we can invest our performance within the factories, but also along the supply chain. A couple good examples of that are in our Elpida factory, the former Elpida factory in Micron Memory Japan, in Hiroshima.
We're continuing to look at ways we can optimize the throughput and improve the overall fab performance and align with other factories in the Micron network. On the island in Singapore, we're taking a look at how we can optimize all of our NAND resources there to drive the economies of scale, the cost benefits, and actually the capabilities to drive not just a healthy 16-nanometer execution, but also how we think about further technology roll-outs in the NAND non-volatile state space. Things like clean room optimization. How do we look at our factories and equipment utilization? These are areas that require investment for us to become world-class in the performance and results we measure ourselves upon.
One of the areas that we need to execute really well on as this memory market diversifies in end market segments, we need to be able to be flexible in terms of our product mix. One good example of that is in the mobile-to-PC market dynamic. When you think about the capacity that we allocate to both of those markets, we have to be flexible to ride the opportunities that each of the market segments presents and have the capabilities to move where the best returns will be over both the near term and long term. We have continually done that in 2014. The rest of our business diversifies, we'll be focused on being more flexible, more nimble, and driving the economies of scale around product mix and the flexibility of making use of our capacity for a stronger return.
We've had a pretty strong year on the manufacturing side, we've reduced our cycle times on the front end about 30% in DRAM and about 40% in NAND. We continue to think there's opportunity for us to drive better economics around our manufacturing performance. These are areas, again, that some of this requires investment in capabilities, both in terms of equipment utilization, as well as in terms of how we look at our measurement systems around manufacturing. We think it's very important as I talk about flexibility, but also to be able to be world-class in terms of cycle times, in terms of throughput, in terms of quality, that we're going to be investing in this area.
Again, not increasing capacity, but as we look at the opportunity for efficiency and driving better financial performance and being flexible enough to adopt to the new opportunities in the memory space. The challenge I talk about product mix is, along with these new diversified markets, as we look at the better margin opportunities and innovative solutions in these new segments, while it's fantastic in terms of what it means for the memory business and the health of the memory business, but we have to be able to react to each of those opportunities. You can look back and see 10 years ago, the concentration of our business in what we would say, more commoditized businesses. Well, the opportunity for us is to drive these solutions into different end market segments.
How we manage the capabilities and how we have to invest in driving the flexibility around driving our diversified end market product mixes will help us be much more effective in inventory management and customer satisfaction, and being able to be a differentiated memory supplier. I've talked to you so far about technology and how we need to invest in either cost-driven process migration or margin enhancement capabilities in areas such as controllers and packaging and firmware. I've also talked about our capabilities around being a world-class manufacturing operation, all the way from the front end in terms of what we do in terms of managing products through the fab and out to our customers. The other unique opportunity for us, which we think is a strong foundation for where Micron is going in the future, is our ability to deliver systems-level solutions.
When you think of NAND and DRAM at the core of what we do, we will need to continually invest in capabilities around the silicon to develop solutions. These solutions offer us the ability to solve our customers' problems, meet their demand, and drive better margins for the company long term. Our capacity, when you think about DRAM and NAND, is a precious resource. How we innovate with that capacity and what we do with that capacity to the end markets is really the focus of our teams. When you think about our enablement of silicon, and whether it be something like an eMCP or eMMC for a mobile application, all the way up to an enterprise storage-class device, where we're driving enterprise storage applications to large corporate enterprise applications. We have the ability to deliver end-system solutions fully to our customers.
That is a much different opportunity than you could look back and say 10 years ago in the memory business. As we think about our business going forward, we will continue to invest not only in technology, not only in our operational capabilities around manufacturing and product mix management and inventory and supply chain. We are going to increase our investment and our ability to deliver innovative solutions, which requires not just the investment in silicon, but technology and capabilities around the silicon to allow us to differentiate for our customers. As Mark said, we couldn't be more excited about the opportunity to serve these end markets. We think there's tremendous growth in these end markets. The investment profile that we're discussing with you today, as we think about our business, is really focused on how do we take advantage of these opportunities?
How do we drive ourselves to be world-class in terms of cost? How do we invest in capabilities, whether it be, again, controllers or firmware or packaging or software, that allow us to differentiate our products in a higher value segment and a higher value capability in terms of product roadmaps at the systems level. With that, I'd like to stop and happy to take any questions you might have around the information here or anything Mark presented.
Mark, just a couple of questions around CapEx. First, for this fiscal year ending this month, are you going to hit that $3 billion kind of number you've been talking about? Because it's a relatively big spend this quarter. Then more importantly, when you look at the 3 segments that Mark Durcan put up for next year, that sort of technology segment being 30% of next year's spend, how has that been historically as a percent of spend? Can you give us a sense of what that means for your wafer start growth relative to the CapEx number you're throwing out for the next fiscal year?
Sure. Remind me the first part of your question.
This year's CapEx.
This year's CapEx. Well, we don't normally forecast that. It's safe to say that we're not making any updates today, and I think it's safe to say that we're continuing to invest in our business. Relative to your other question around 2015 and the technology in the 30% category of technology improvement, historically it's been lower. Really what's driving that is, as I mentioned through the theme of my discussion, is it's not just about fixing and innovating and driving process capabilities, but it's the capabilities above the silicon that's a newer aspect of the memory business and one that's very important and critical for us to invest in going forward. One way we look at our business is how much of our memory will be behind the controller going forward.
What that means is controller capability and firmware capability and test and assembly capability of end systems and quality capability around driving those systems to end markets. It has been lower, and I think, actually, this is going to be something that is critical to Micron's success as we think about differentiating our products. Those end market segments, if you go back to this, each of these market segments offers dramatically different behavior than the 2004 bar chart here in the commodity business. You think about enterprise storage and what that requirement means and what the product capabilities need to be in terms of winning in that business. You think about embedded, and you think about server and mobile. They're all unique in a lot of ways. We have to invest in those capabilities.
We've talked about in the past where the memory business, in terms of process lithography, is slowing down. These capabilities become more important as we add value.
Great. Thank you, Mark. Could you go to the page 16 to revisit your cycle timing chart?
Sure.
Well, I'm hearing continuously from the chip makers when they implement next-generation technologies, manufacturing period getting longer and longer. Your chart seems to be in opposite way. For instance, for me, my memory, I remember 10 years ago, DRAM makers saying it will take only maybe 40 days to manufacture, to fabricate the wafers for the DRAM. These days, they are saying maybe 50 days or 60 days. Your chart clearly indicates maybe shorter cycle time.
Sure.
How the Micron can achieve that?
Philosophically, this is a relative chart. That's why we didn't put any timelines, per se. On a relative chart, we're closing the gap on our competitors in cycle time. This year, for example, as I said, relative to our competitors, we went from number 3 in cycle time to number 2 in NAND. We're driving a relative performance in terms of cycle time performance, and that probably proves out more in the future. Our gains this year were not relative. Our gains were relative cuts in terms of our starting base, in terms of cycle times to where we are today. As you can see here in the 2nd bullet, we do talk about complexity drives longer cycle times. It doesn't mean, though, for us, it won't be as important a focus to be the best of world-class at what we do.
Any rough idea how much the wafer capacity may decline every year going forward during the course of implementing the new, more advanced technology, such as 20 node or 1X node?
I don't think we're in a position where we want to give a percentage number on that, although we certainly recognize the dynamic you're referring to in terms of the impact on wafer growth, which is why we're pretty conservative on overall growth rate in terms of both the DRAM and the NAND business. Thank you.
Yeah, good morning. I was wondering if you could go back and reference Mark Durcan's remarks about the shift in bit growth in the DRAM market. Over the past several months, Micron has been reducing its forecasts for DRAM bit growth and now we're increasing them, and perhaps you could explain some of the dynamics there.
Some of that. The relative bit growth, some of it has to do with our transition and integration of Elpida and how we're managing our fab network. The other part of it has to do with our increasing migration from 30-nanometer to 25-nanometer. You're seeing natural process bit growth. I think we need to be clear that we don't anticipate growing more than the industry, and we're not investing in greenfield technology or capabilities or capacity. What we are focused on is the capabilities to lower our cost, and even as we do that, we're looking from a returns perspective. Some of this bit growth you're seeing is natural migration from a process standpoint, from the 30-nanometer to 25-nanometer.
When you couple that with capabilities of improving cycle time and what have you're seeing some modest improvement in bit growth, but not driven by large greenfield investment.
If I could have a follow-on.
Sure.
We've had a situation of unusual price stability in the DRAM market, and yet the industry-wide, including all three supplier bit growth figures, have just ratcheted up fairly dramatically in the past week. I was just curious if you could address that.
We don't look at it like that. We don't think those numbers have ratcheted up. On top of that, maybe to address an item that came out last week, we thought the announcement last week by one of our competitors was a validation of what we've been saying all along, which is no one's going to go out and light up a whole brand-new factory. This is an increment into the industry, and the fact that it's actually in place and in a limited environment shows that there's no new massive DRAM capacity coming online. If you saw Mark's numbers here, we think this is embedded in those numbers, and we're pretty comfortable that the supply-demand balance stays healthy. We're very optimistic about that. We think it's a validation that there's no major new capacity coming on in DRAM.
We took more of a positive look out of that. The market will absorb this, and the growth rate in DRAM relative to what you're going to hear about DRAM demand, we think we're pretty comfortable with it.
Mark, I have two questions. One, just to follow up on DRAM demand. If you could just help us understand if anything's changed in the last 6 months in terms of how you think about the DRAM demand profile that gives you comfort around the low 30s type of bit growth this year from a supply standpoint. The second question is about technology investments in general. I think some of the longer-term investors are interested in understanding how you benchmark your technology investments compared to competition to make sure that you're not under-investing in any portion of the business.
Okay. I'll come back. The DRAM demand question, generally, we continue to be very bullish on the DRAM business. You've seen Intel, in the past, come out with strong endorsement of the PC market being better than expected. I think the number that's generally accepted is their forecast for 2014 was about a 5% decline, and what we're seeing in computing, in desktop, notebook, is something closer to flat to maybe down a %. In general, across all of our segments, the demand continues to be robust. Mobile DRAM is still very strong demand. Networking continues to be strong, server, embedded, automotive. Just about across the board, we're seeing very strong demand. If you look at inventories in the channel, inventories not just in the channel part of DRAM, but at the OEM level, inventories remain pretty tight going into the back half of the year.
This is a period where there normally is very heavy accumulation as people plan for the holiday season. We're not seeing healthy inventories reflecting what we think is strong demand and strong balance in terms of supply and demand in the market. As it relates to how we benchmark ourselves and to the extent that we're measuring ourselves well enough to know that we're not under-investing, we look at the items that I presented, too. For example, how we're looking at timing our 25-nanometer and 20-nanometer investments and also our 3D NAND enablement and some of the investments around controller and packaging.
When we look at that, we actually, obviously, are very focused on a returns model, an ROIC model that Ron will talk to, but it gets back to how we think we can capture the right level of gross margin down to the operating level and make sure we're getting a good return for our shareholders. What I would tell you is that it's not necessarily benched just against purely a competitive slate. It's what we can do versus what our baseline is. Some of these timing issues are choices we're making to drive those results for us long-term.
The point we're making today really is that while none of it's necessarily driven by new capacity focus, a lot of it's driven by this return approach, which is allowing us to think about our business and what you would do to make a better economic choice. As you serve your customers, can you do things different to drive a better financial model for your company and yet serve the customers and create innovation? That's when you think about, I talked about a little about process, but a lot about capabilities and manufacturing efficiencies, because those are the things that will return better financial model for the company, operating profit, but also allow us to differentiate downstream.
Thanks. To understand the range of CapEx from next year, 3.6 to 4, is it a more fixed and variable number than the past in terms of the 3.6 kind of has to happen regardless of revenue to some extent because it's driven by technology versus capacity? My second question is, when you say flex between DRAM, LPDRAM, and NAND, do you mean initial or being able to convert after the fact?
Let me answer the second part. Maybe I misspoke. What I meant on that, the flex is really within a technology. We have some capabilities on the NAND and DRAM side, but not a lot. It's more within a technology capabilities of how we might do a mobile DRAM to compute DRAM and how we can change wafer starts within a 9 to 12-week period. If we see, again, this is a hypothetical If we were to see an impact on demand in one of those segments, while the other one continued to be pretty strong in demand, we're building the capability and the flexibility to move that capacity in a nimble way so we can drive better results in the near term. That's not just true in that one example, but it's true across all of our capacity.
It also will be true in NAND, when you think about client SSDs and enterprise SSDs, and think about mobile NAND and those applications in embedded, we have to be able to move our capacity and flex it around because some of the price and some of the cost of being diversified in markets are, you have to be more capable from a supply chain and end systems capability to serve those customers. That's really what we're talking about in terms of those investments, is it wasn't really from DRAM to NAND, it's more within a technology, how can we react to customer and market signals to drive better profits, but also drive better customer service in terms of our ability to serve the markets that need the capacity most?
Excellent clarification. The puts and takes between the three, six, and the four, and whether it's more fixed versus variable CapEx on a relative basis, given it's more technology versus-
It's a great question. I'm not sure I can give you an acute answer to, is it more fixed or variable in the past? I will say that, as we look at it, the range that Mark gave, we feel pretty good about the returns piece of that range. The range that Mark presented, we feel it's got great upside to drive better results to company.
Sorry.
Sure.
Mark, you talked about more customized product going forward and more diverse end applications. We've already seen over the last couple of years, obviously, the contract market becoming larger and the spot market becoming much smaller. I'd like to understand, as there's going to be even more customized product, does the spot market totally go away? That's the first part of the question. The second is, how do you deal with inventory management? Because you go away from being a commodity market where if one customer doesn't get their production forecast right, you can always sell it to somebody else. That isn't the case anymore. How do you make sure that all these production forecasts coming in from all these customers across a whole range of different products don't end up giving you inventory problems internally?
Well, I think you've hit on an area that we're very focused on in terms of each one of these businesses, and the way we've structured ourselves, are really different businesses. If you look at what we call CNBU, computing and networking, that's a lot different in terms of the behavior and in terms of the applications, the end market, and the supply chain, and how we service our customers than compared to server and embedded. One's got higher turns, larger volume, one's a broader set of customers and slower turns with healthier margins, as an example. We're investing in our capability to actually have multiple supply chains within our operations, and that's really a key part of our execution. Each one of our businesses have to drive that level of performance.
I will say something about inventory before I get to the second part of your question is, it also makes inventory a little bit different. We've had a pretty high priority in terms of driving our inventory to an optimal level. We're not in a liquidate inventory mentality in this new business of memory. People ask us a lot about what would happen if this happened or that happened. We are looking at, again, the financial methodologies to drive our business to serve each of these customer segments. What's nice about these customer segments that comes along with the challenge there that you identified in the inventory pieces, the customer stickiness and the customer engagement are different than a commodity business. In PC DRAM in years past, you could take a part from customer A and give it to customer B. That's not necessarily true in mobile.
That's certainly not true in embedded. That's not true necessarily in networking. It's not true as you go down the enterprise scale in terms of big data and cloud and enterprise storage. The perishability in this inventory is different, so our customer engagements are different. They're more one-to-one. We're going in with our storage investment unit team and talking to a customer about how to solve their problem and how to define a solution for just them. The inventory discussion is more about how we manage demand forecasts and their ability to forecast their business, and we work with the customer on that. It's not one where we feel compelled that we're just going to move those bits elsewhere because we'll work with the customer to drive the solutions.
The short answer is, a bit more of a supply chain by grouping, either the business units or the end market categories that we're driving that capability in the company. That requires information systems, that requires management systems to help us understand and get a good sense of how we're performing to goals and making sure the customer's aware of that. It also tells you that our relationships with customers are a lot different. We often get asked on calls, how is the discussion going with the customer? How has it changed? Quite frankly, the customer has to take more ownership with us as we think about how we build a business together, and those solutions are really partnerships, because they're building their products based on our capabilities.
You think about end products in the future, whether it be for the data center or for enterprise storage, these customers are working with us to find an architecture that's really proprietary to them in most cases. That's the growing trend in our business. As we manage inventory, it's more about looking at their forecasts and our business ability to project what they will actually take from us from a demand perspective. Then they have to take ownership in working the business model with us, and that's a different perspective than we have a commodity set of inventory products that we have to move because we're concerned about the pricing.
We think the pricing dynamic that drove that in the past is muted, and we think going forward, each of these business segments represents an opportunity for differentiation and also to manage our business relationships one-to-one with our customers, which allow us to drive, again, better gross margins for us and a better business model going long term. Okay. I think that's all I have time for now. I think we're going to now pass it over to Scott DeBoer, and Scott's going to give you a little bit more deep dive on where we're investing on a technology level.
Hey, good morning. I'm going to start out today just talking a little bit about some of the technology highlights that we've had over the past six months, especially since the last analyst conference, and line it up to our investment priorities and some of the general strategy of where we're going from a technology development point of view. If we look at our DRAM technology highlights through this period of time, we're very, very pleased with how the technology integration has gone between the teams from Elpida and from Micron, and how we've combined those to really make strong progress on not only 20-nanometer technology and the rollout of 25-nanometer, but also on the nodes beyond that. Our progress on 20-nanometer yield in Hiroshima has been excellent through this period of time.
In addition to that, we've been able to focus our resources in Boise on advanced nodes on the 1X and the 1Y nodes. I'll talk in a little bit more detail about those on an upcoming slide. On the NAND side, we've had strong success and nice recognition for the innovation of our 16-nanometer technology rollout. Had some nice awards around the capability of that product. In addition to that, we're very confident in our position on 3D NAND. I'll give you some update, a little more details today on our 3D NAND direction. We believe it's on track for production in mid-2015. As Mark had referenced, it'll be moving into manufacturing fabs towards the end of this year.
On the package technology front, the big focus of investment for us around memory system enablement and the rollout of the Hybrid Memory Cube, but also several other 3D package technologies that I'll talk about today, continue to go well through this period of time. Most of the focus in this year has been around customer enablement of these unique 3D package technologies, with really a setup for the manufacturing to start out in 2015 in a bigger way. On the new memory technology front, we've got several technologies that we've been focused on. In the next couple of years, we've got targets to bring different technologies for early manufacturing introduction. The first one we're targeting for next year. I'll go a little bit deeper into DRAM specifically right now.
Our 20-nanometer technology transition, as Mark and Mark have indicated previously, is really focused around differentiated product enablement. Our transition to 20-nanometer is absolutely targeted at that. Of course, cost improvement is always a piece of these transitions. The cost improvement, in this case, indicated on the graph below, which is probably hard to see in the back. If you look at a high-density product like an 8-gigabit product, really sits in the sweet spot of the die size for manufacturing costs. We look at our DDR4 and our LPDDR4 products, at 8-gigabit, and they really come in and are enabled by the 20-nanometer node.
Couple other interesting things on this particular graph is it also highlights, there's kind of a lower bound to where it's optimal. This will lead into a little of the discussion about the products that we tend to leave on lower densities, leaving them on legacy nodes and cover those for different parts of our business. In addition to that, we look out to where 16-gigabit products really get to be enabled, which is really starting at the 1X timeframe, but really requires the 1Y node to enable those. In 2015, we'll be rolling out 20-nanometer. Obviously, we're developing it and rolling it out in Hiroshima. Then we'll be rolling out one other fab. We'll have two fabs running volume on 20-nanometer in 2015.
The development team's focus right now really is around clarifying the path to the enablement of two nodes beyond this 20-nanometer node. We have some good early suggestions on the path there. When we look at DRAM and focus on what we're really getting for the technology nodes as we're rolling them out, it's a much more complicated situation than it's been in the past. As you look at the graph on the left of year-over-year gigabit per wafer growth, it's an industry graph, and the projection of the trend there going 30% and going below, driven by really this complexity increase, especially beyond the 25-nanometer node. The graph on the top is just an indication of the driver of that, but it's just a proxy for that. It's the litho levels as you're going through nodes.
The litho levels, of course, bring along process complexity, that were even referenced in a question a little while ago, for all the process areas. It's really a building of complexity as we're going on. The yellow line is indicating the wafer outs per area. Just indicating that the output reduction through time, as we go to these advanced nodes. The net of this really is that process node transitions are becoming less effective, and it comes back to why we're doing them in the first place at this point to focus on these capability enhancements. It also brings out that legacy nodes, when we look at Increasingly less commoditized nature of this business, our legacy nodes become more important, and how we manage what volume on which nodes becomes a different business proposition than it was in the past.
Really, the focus on this then all leading to, we've picked our certain products for 20-nanometer, we go focus on the enablement of those for system-level products. Switching over to the NAND side. Again, we have a complicated transition in the NAND world that most of you are aware of, from planar to 3D NAND. On Micron's planar roadmap, we have optimized the cost and the performance of our 16-nanometer, both in terms of MLC and TLC products. That is our transition point. It's our final planar node, and I'll talk a little bit about why that is on the next slide. We're very pleased with the 16-nanometer product. The yield ramp on this technology is the fastest that we've had in Micron's history. The quality of this technology is enabling to products that we'll have for many years.
When we look at our 3D NAND technology, which is illustrated on some of the pictures here, our first product will come out. When we talk about our generation one product, it'll be a 32-tier, 256-gigabit die. First die will be MLC, but the technology's really designed around TLC capability and enabling high-performance solid-state drives. One of the unique things about this technology is the architecture, we believe, not only provides outstanding performance from our data to date, it also provides us with what will be the highest gigabit per area density in the industry. When we look at the transition of NAND technology from planar to vertical, it's a story similar to DRAM in that it's a space and cost consideration with a few kind of significant differences.
When we look at the cost of the node transitions, historically, for NAND, we optimized our 16-nanometer technology to be an extremely low-cost point. We picked a node that was appropriate for low capital cost and high performance. If you look at, again, the same situation on the wafer outs, relatively no wafer outs hit on 16-nanometer from 20-nanometer, the way that we took it or chose it. When we go to 3D NAND on a 32-tier technology, you do have a substantial 1.7x to 2, somewhere in that range, reduction in wafer outs per area. You have to contemplate all that when you're looking at this node transition. Now, when we looked at, for Micron, putting in an additional planar node, which is certainly capable of doing, for Micron, it wound up being a relatively high capital cost again.
Because of the way the NAND transition goes, the capital put into another planar shrink is not very useful to the 3D. We view that as not efficient use of capital, and that really is what drove us to go from 16 straight to 3D. We think that the cost crossover, if you look at the bottom, for 3D NAND relative to our planar 3-bit per cell technology, will occur in late 2015. We think that at that point, we get to a more historical-type cost reduction with the rollout of 3D NAND from that point on. Okay. A couple of the other technology enablement areas that have been referenced. The 3D packaging on the left side of this graph, representing kind of the value contribution of packaging in general, packaging technology.
Packaging technology now has a much wider connotation of the things that are included in package technology. Certainly, on the NAND side, it goes to very high density, high die count stacks to enable high-performance NAND technology. On mobile DRAM, very unique three-dimensional packaging to enable different kinds of form factors. On the high-density compute, we've talked many times about our Hybrid Memory Cube and the excitement around that in the industry, really enabled by a combination of system-level management in combination with raw process technology around 3D interconnects and through-silicon via technology. A lot of excitement on that side and a big focus area for us where we're adding capability resources. On the new memory side, we're very focused around the cost and performance gap between DRAM and NAND.
In looking at opportunities for different memory technology that can fit into this space of cost and performance in different ways. We believe that there's a real opportunity, especially when you look at system memory products, for using some of the new technologies we've been developing that provide, maybe, non-volatility at higher speeds than a NAND is capable of, but at a much better cost structure than DRAM. Our focus is around that, and we think there's some exciting stuff to come over the next year in that area. This is the last slide, kind of tying off overall direction. One thing that's just visually to look at on the calendar year 2015, there's an awful lot of new technology elements coming in a big way and all kind of lining up vertically.
We have some big and interesting challenges through this period of time and great opportunity as we roll out multiple new technologies through the same calendar year. Our focus through this period of time is improving our DRAM technology position, and as I mentioned, we've had some great success with the combination of our teams in Japan and Boise to really drive a faster introduction cadence and new product enablement through this period of time. On the NAND side, obviously the 3D NAND is a big opportunity, and we're focused on enabling that and getting that out in the market in 2015 with a high-performance product. On the technology packaging side, lots of things, and Brian's going to talk some more about some of the system-level opportunities that these three-dimensional package technologies are enabling. Some great things there.
Finally, the intention to initiate a new memory technology in the next year into the market. Ready for any questions.
Thanks. Hey, Scott. How does the NAND controller challenge change when you go from 16-nanometer to 3D? Just help us understand what was kind of the issue with going to TLC in a faster timeframe on 16, and how are you able to deal with that with the migration to 3D?
I'll start off on a bit of that, and I know Brian will have some to add to that and even Darren. One of the things on 3D NAND capability that is really been focused on TLC from the very beginning. Our enablement efforts around the controllers for that have. It's not like we're starting now trying to get them ready. They've been a focus over multiple years. In addition, Brian will talk about the level of resources we've been adding to the controller area, to make sure that we're positioned well with the right capabilities to roll out significant number of controllers in parallel, as opposed to some of the kind of slower nature in the past, that we're a little bit slow on some of our TLC parts. I know Brian will add more to that as well.
I'm sorry, there was two parts to your question, right? Okay.
Okay. Thank you. I wanted to ask on your package and test where you're putting a lot of investment, I guess the 2 stages. First one on 20-nanometer, where you're at in moving a converged process and also transitioning to Micron process-
with your own tests on the mobile DRAM.
Sure.
Further out on the 3D packaging, when you think that becomes a meaningful part of your mix-
Okay
to move the structure to 3D.
Sure. From a process technology point of view, when you were talking about converge, you're talking about converge between compute and mobile or between Elpida and Micron?
Both.
Yeah. We have converged our process technology into one platform at 20-nanometer. We brought parts from each and aligned and moved forward on 20-nanometer with one set of process technology. Mobile versus compute, there are subtle differences, and the mobile process is slightly different and will continue to be different, both for DRAM and likely for NAND. I think that part's gone well. On the test infrastructure, again, there's subtleties that are different between compute and mobile tests, but we're definitely coalescing to one test flow as well.
A bit of a longer-term question when you're focused on new technologies. Right now, we've got a fantastic supply-demand situation because it's very, very hard to scale for all the reasons you've outlined, and that has prevented anybody really having confidence to put a lot of CapEx into new builds. When we move to a new technology, could that potentially change? One theory I've heard is if we move to something like ReRAM or PRAM or something like that, NAND and DRAM merge into one product, people get confidence to invest again, and then we're back on Moore's Law again, and then we go back to the DRAM industry of five, 10 years ago. Obviously, we all hope that doesn't happen, but what's the risk that it could potentially happen?
I think that's an exceptionally low risk. I think when we look at things like, well, first off, the complexity is often underestimated in going to ReRAM or Spin-Torque or different types of memory technology. There's a huge amount of complexity behind that, and actually, an awful lot of the memory technology from either NAND or DRAM is very similar and has to be brought in. The other piece around new memory technology is it is very system application specific. It's going to require management, in ways that are similar to NAND or more advanced, and it's going to require some very careful consideration there. We've got a view that these new memory technologies will come along, but they're not really displacing NAND, for example.
Now, there's perspectives, I understand, in the industry, but when we look at new memory in general, if you look at all the publications and different things on Resistive RAM, there's really not anyone that is actually competitive with a three bit per cell vertical NAND in terms of a cost per bit. We really think it's an additive business opportunity. It's not a one plus one still equals two. This is a two and a half out of the emerging memory opportunities.
Understanding that 3D NAND has a cost advantage, but I've heard from other folks that there's also a performance advantage, particularly for enterprise. First of all, I wanted to find out if that was correct, if there's metrics for that. If that is the case, is there a risk that you're a little bit behind other folks, that some designs might move because of that performance advantage, or are you comfortable that that doesn't happen?
I think we're very comfortable with the position that we've got right now. There is some low volume enablement in the industry right now. I think the timing of our products coming in and going out to customers for enablement is very competitive in reality. The time of our volume ramp up is also very competitive. I think despite a lot of publications and noise, there's 3D products in the market today. We understand that. We also understand the competitive advantage, which we obviously haven't articulated perfectly yet, of where our products are going to come out and have performance relative to the ones that are out there today. To the first part of your question, absolutely, we have a performance advantage on the 3D NAND relative to planar.
There's different enablement pieces, and even to the question earlier, there's controller technology that has to come along with it. There are fundamental performance advantages from the 3D NAND technology, just at a basic physics level that enable better performance.
Scott, a couple questions. If you look, you guys have been significantly under-earning your peers in the NAND market for the last 12-18 months, I guess I'm trying to get a better understanding of how much of that is a technology shortfall on your part versus the parts of the market you're in today. How much of it's a technology issue versus a market issue? Can you give us an update on where you are with TLC? Have you finally-
Sure
gotten over that hump?
If I ask the technology guy that question, then ask the business guy that question.
My second question, quickly, how do you juggle the ROI from 16 to planar in the litho bay? Because you'll have a lot of immersion tools for 16 that you won't be able to readily use at 3D. How do you balance that ROI equation through that transition?
Okay. To start with, I'll say that from a base technology point of view, Micron's been very competitive over a number of years on the pure component technology piece. I think we're confident in that now, and we're confident in it going forward. Darren is going to talk more, and Brian will too, about how we look at the NAND business and what our strategy is going forward to make sure that the problem which you framed up there isn't the problem going forward. The second piece is on our 16-nanometer technology. First off, Micron is a big global place, if we do have an extra scanner, we can put it someplace else. DRAM does have growth in that. As far as tool utilization overall, we don't have any excess tools with this transition. There is still immersion required for 3D NAND.
It's just not in the formation of the cells themselves, but you still have the circuitry and different things that drive immersion levels. Net of it is there really isn't excess equipment from it. Okay.
Thanks, Scott. We're going to take a quick break now. Let's keep it pretty brief, five, 10 minutes. There are some refreshments outside, et cetera. For the webcast as well, it'll be about a 5-to-10-minute break, we'll be back on. One more quick thing for those of you in the room.
[Presentation]
All right, everyone. Let's see. If we can get everyone to take their seats, we're going to go ahead and get started here. Hey. What's that? All right. One more time, if we can get everyone to take their seats, we'll go ahead and try to keep this rolling on time. All right. Let's go ahead and get rolling here. For those of you that I haven't met, let's see. Okay, good, we're rolling here. For those of you that I haven't met, my name's Brian Shirley. I oversee the business units in design and product engineering for Micron. What I'm going to share with you today, first of all, a couple graphs on aggregate DRAM and NAND demand. You saw the view of supply. We're going to go into a little bit more detail on the supply piece of it.
I'm going to go through a few of the key segments that we're excited about. What I'm going to do, I'm going to go ahead and cover the memory and the storage pieces for a number of those segments. When it comes to enterprise storage and client SSD storage, we're going to defer that for Darren Thomas, who will come up after me, and I'll have a chance to better introduce Darren for you as well. Let's go ahead and tear into it. First of all, this is our view of what's happening on the DRAM demand side of the house. We think this is a relatively compelling view. In a lot of ways, it's actually still fairly conservative, but what this works out to, this is a four-year CAGR, 2014 through 2018, working out to 25% demand growth.
Really, the key drivers, I don't think any surprise, but the combo of mobile and tablet really having become a significant piece of the market. Obviously, server storage and networking becoming the other piece, and client as well. Those three pieces, in some sense, working as a virtuous circle here, both on the demand creation side, the storage, how we move the bits, et cetera. Starting from the top, handset. Again, we think a relatively conservative view. What this 26% number has baked in is roughly a 10% unit CAGR in DRAM content per phone, moving from roughly the one gigabyte range up to essentially the two gigabyte range by the end of 2018. Tablet, since the last time we met, we have pulled down this growth rate slightly.
It's about a 20% unit CAGR on the tablet side, DRAM growth per tablet going from roughly one gigabyte to the three gigabyte range. It's fair to say that as we pulled down the tablet number somewhat, we do look at the cell phone, tablet, and client PC ultrathin market as a bit of a continuum. Indeed, since the last time we met, the PC number has come up slightly. That's not so much a reflection of unit demand on the PC side increasing. It's actually a bit of a flat model, sitting in total at about 320 million units year in, year out through this timeframe. Under the surface of that, there's obviously a whole bunch of movement between desktop and traditional notebook, generally moving over to ultrathin in large numbers by 2018.
Finally, we get to server storage and networking, where things are clipping in at a relatively torrid 41% growth rate. I'll talk more about that coming up. Just a phenomenal growth rate inside of server, driven really by unit growth on the cloud side and content per box growth inside of on-premise server. On the NAND side, a four-year CAGR here of 38% on the demand side. Again, we think this paints a compelling view, but in a lot of ways is relatively conservative. Handset coming in at 27%. You can think about that. I gave you the unit numbers baked into phone and tablet, but you can think about the content per phone roughly going from on the order of about nine gigabytes per unit of storage today, up to roughly the 26 gigabytes by the end of 2018.
On the tablet side, you're looking at something closer to roughly 17 gigabytes on average today, going up to about 60 gigabytes by 2018. The other notable here is obviously enterprise SSD, huge growth rates, and that's driven, something Darren will be talking more about, as enterprise SSDs truly push more from what we would consider the initial caching opportunities and more into what we would call the hot edge. Client SSD, I think something here, a relatively conservative estimate of 33% CAGR. What this would imply by the end of 2018, client SSD units on the order of 150 million at about a 390-gigabyte average. Frankly, in our minds, there's a tipping point, sometime in this timeframe. These numbers would not yet show that kind of a major tipping point. We do think that there is upside to these numbers.
Big picture, something that, outside of the demand trends, we're probably more excited about, is what's happening to the actual solutions inside of each of these segments. That's driven by a few trends here. First of all, as most of you know, a huge push inside of the application to get more performance out of the processor. The way you do that is you get the memory closer to the processor. Secondly, power is becoming critical in every single one of these applications. That helps to give a further push of processing and memory closer together. Frankly, in something we're seeing in just about every one of these applications, the line between memory and storage starting to blur.
As we go through each of these segments, you'll notice a number of opportunities where we are already starting to ship DRAM and NAND in the same package. There are opportunities where they may be in the same package, in the same module form factor. That speaks to some of that trend. It also speaks to the viability of some of the new technologies that Scott DeBoer is looking at. What we're doing about that, first of all of our R&D investment is really contemplated by these application drivers. Certainly, how we look at DRAM development, 25-nanometer and 20-nanometer, both from a low power as well as a high-performance standpoint. How we look at our planar and vertical NAND developments, how we're looking at alternative memory developments, but also some of the assembly technologies that Scott talked about.
We have made significant investments, not just on the R&D side, but also on the production tooling side for this next generation of form factor, where frankly, the assembly technology is just as critical, if not more so, to this end application. The other piece of it is more in this memory solutions core, and it's really the solutions development piece. Something we've done, and I think we've done a fairly good job on the DRAM side, and we're making the investments on the NAND side, is actually bringing in a large level of application expertise, specifically on the architecture side, as well as the go-to-market side. What that does, it helps us to get closer to the end application and get this memory specialized, customized, if you will, in a way that, frankly, it's not the same commoditized high-volume JEDEC shipment part from 10 years ago.
The other investment we're making is significant resources of controller and firmware resources inside of the company. You've seen us develop in-house controllers for HMC. You've seen us do it for our first-generation PCIE device, which in a lot of ways is still a performance leader. You will see increased investment across the board, Darren will talk about some of the investments we've already made, but a significant level inside the company organically, and not averse at looking at other investment opportunities as well. The final piece of that is really this field applications piece, getting labs, the go-to-market engine out closer to the customers. That helps us take a lot of these core technologies and core developments and get them tuned for the end customer application, and enabled, which frankly is fairly critical.
In automotive, in NAND storage opportunities in the industrial market, even in mobile, there's a large number of opportunities of tuning the memory solution directly for that particular customer, and that's something that we have made investments in as well. What that leads to Frankly, proven success for Micron in the networking server and AIM side, both from a breadth of portfolio as well as an overall share of market in these industries. I'll speak more to the individual products coming up. It also speaks to several opportunities to ramp the next generation of products sooner, both those that require TSVs, and as Darren will speak to, really the next generation of enterprise and client-based SSDs. Going into each of these segments in a little bit more detail. Enterprise here, this includes both server and networking memory, by the way.
What a good segment this has been for Micron. One figure, we're going to be looking at content per server, really driven by the opportunities of in-memory databases, by real-time analytics, being driven from roughly a 64-gigabyte point today to 180 gigabytes per box by 2017. That's right around the corner. That's pretty big growth. Really, this server CAGR is driven, again, both by cloud computing as well as on-premise applications. Cloud computing, probably a little bit more of a server unit growth, and on-premise server really being driven by the amount of DRAM per server for this kind of real-time analytics. Networking side, here you have LTE, which is really in the initial innings.
If you think in terms today of LTE having something on the order of 380 million subscriptions worldwide today, generally driven by about three large countries, our view is that by 2017, you're going to be looking at about 2.3 billion LTE connections across the entire world, obviously. That drives a huge amount of networking gear. Here you have the other prerogative of the next generation of routers and switches, really 100G Ethernet switching, which requires very high-performance DRAM. I'm not going to go through all of the product lines. Suffice to say, mainstream DDR3 and DDR4, we have pushed both on 25 and 20 nanometer for eight gigabit solutions here. What that does, it allows us to take the sweet spot module density from roughly 16 gigabyte in the server space today up to 32 gigabyte next year.
Hybrid memory solutions, we've talked an awful lot about HMC in the past. Hopefully, you saw the press release that Micron and Intel released a few weeks ago. The next generation Xeon co-processor, which, by the way, is helping with the HPC space inside of a number of upper-end systems. The memory solution of choice is a variant of HMC that Micron has worked with Intel to design. We're pretty proud about that. We think it helps to validate the HMC concept as really the go-to memory solution for these upper-end solutions. I would also say that the next generation of network packet buffering, our view is there's really no other way to get it done besides HMC, and we're working with a whole lot of customers to help do that. Non-volatile DIMM, these are standard DRAM DIMMs with a layer of flash behind it.
This is getting very good uptake for us. It's really driven by enabling some level of persistent memory, but with the ability to still have DRAM on those modules, so you're not giving up the DRAM layer itself. Data loss prevention, cache tiering, a number of companies finding this very useful inside of their systems. RLDRAM 3, obviously, this has turned into the memory of choice inside of today's routers and switches. In-package memory, I will tell you that this is under development right now. This is something that we're very bullish on. There are pieces of this that are informed by HMC, pieces of it that are informed by the next generation of assembly technology. In-package memory across the space is something you will start to see really in the next few years. The Automata Processor that we introduced last time.
When you talk about big data, you think in terms of what the velocity of the data is and how structured it is. When it comes to unstructured, high-velocity data, you need some kind of a pattern recognition engine, and the best way we've found to get that done is with a DRAM technology like Automata Processor. This is in 2015 revenue. This is a long-term play, but it's something that we're pretty bullish about. Inside of mobile, what a good segment this has turned into for Micron. We were not large players in this, obviously, three to five years ago. We are, frankly, across server, across PC, across mobile, we are oversubscribed today in terms of bits. We could move a lot more to any of these segments.
I will tell you that in addition to a number of the application drivers you see up here, really, what we find most interesting is the growth happening in the mid- to mid-upper end smartphone. Really, some of the trends that you're seeing out of China. Obviously, you've seen the numbers from Xiaomi, from Coolpad, et cetera. What's interesting about that particular piece of the market is that it's pulling on LP memory and NAND inside of the same package. With a controller and firmware. That's something that inside of Micron, we feel uniquely able to go and push out there. This is a trend that's been very, very good for us. Frankly, you get back as little as a year ago, this was not well contemplated by the market.
Obviously, we are still pushing an awful lot of memory out the door for PoP memory, LP inside of higher-end smartphones. You'll see us in a big way with 16-nanometer NAND in the eMMC space, but this eMCP, that combines DRAM and NAND and the controller, is something that we think we bring some good advantages to having all of those technologies in-house. The embedded segment, as most of you know, we have a business unit dedicated to this. It's really a combination of the automotive market, which, by the way, is seeing good growth in memory, driven by infotainment, by ADAS systems, by digital displays. You have the industrial, medical, military market. Really, in some ways, this Internet of Things, machine-to-machine communication. You have the connected home market, DTV, set-top box.
All of these systems are driving good growth, and it's something that we feel uniquely able to deliver on, both across DRAM and NAND, and frankly, even NOR. There's a level of memory innovation happening here. Frankly, this space pulling in solutions generated inside of mobile, for instance, eMMC opportunities, and helping us, we think, to really round out a nice legacy portfolio in a way that we're uniquely capable of doing. There's a lot of competitive dynamics in the NOR space as people are looking out there for who is a vendor they can count to be there for the long term. If you see what Micron is capable of doing by putting NOR inside of 300-millimeter fabrication, along with legacy NAND and legacy DRAM. This space is not calling for the latest and greatest technology. It doesn't need eight-gigabit DRAMs necessarily.
The connected home does inside of DTV. Inside of the IMM market, these are not large density devices necessarily. What they do need is good legacy support, something that we've helped to deliver here, called the Product Longevity Program, PLP. This guarantees to these customers that we will be in long-term production with these products for five, 10 years, and that's a very, very valuable thing inside of this space, when you talk about the industrial market, the automotive market. It's really helped our embedded business unit become one of the more profitable business units inside of the company as well as inside of the industry, something we're pretty proud about. Finally, even in the client memory space, good innovation happening even here. We think about this both in terms of ultra-thin growth, memory soldered down.
There is a big push, obviously, with the Haswell processor introduced this year. It will be introduced on the server side next year, driving LP memory, soon to be some level of DDR4 memory. This is a space that's really been transformed in terms of something that the boxes are going out with higher content per box, as well as a little bit more focus on how to get the right level of performance and power down. Thinking about the graphics side, obviously, we're showing a 6% CAGR here. Interestingly, if you had included 2013, that CAGR would've popped up to about 19%, and the reason for that is that what you see in the graphics space is, I call it a very clumpy uptake as the next generation of game consoles comes into play.
Graphics, in general, are doing very, very well with the next generation of memory consoles out there today. We play in every single one of them, both with our GDDR5 portfolio as well as our standard DDR3 high-speed portfolio, and has turned into a very solid market for us on the high-end performance side. That's how we look at the market. Some trends that we think are very, very favorable here long term, driving really, again, from the application side, specifically into the needs of the application. Really a perspective, in my mind, that memory is specializing, it's customizing here, almost more of an ASIC-like market. We believe those are good trends for all of the things that we can do uniquely here, both across DRAM, NAND, controllers, and assembly technologies. Frankly, puts Micron in a good position to win in the memory industry long term.
With that, I'd like to go ahead and open it up to questions.
Great. Thanks. Hey, Brian. Could you just start off with some comments on near-term pricing dynamics, both in DRAM and NAND? Maybe also mention if you're seeing incremental evidence that Micron's getting paid for memory solutions versus commodities.
Sure. Very good questions. In terms of short-term dynamics, I would just tell you that, as you're probably hearing about in the channel market, generally, pricing both in DRAM, certainly on the PC side, the PC channel market, as well as what we would call the white box server market, pricing is up recently. We are seeing that trend propagating into other markets as well. In terms of NAND, you asked about NAND as well. We are seeing an uptick in this market as well, probably best evidenced by what's happening in the channel. I would tell you that from a day in, day out demand perspective, again, we are oversubscribed. We could, across really our big high volume spaces, mobile client and enterprise, we could move a lot more volume in any single one of those spaces easily. Pleased from that perspective.
In terms of the solution side, again, on the DRAM side, I'd say we have seen evidence of that solution value add, probably most evidenced by some of the networking solutions that we do today. NAND, as we've told you, this has been an area of investment for us. We do feel like this is a great opportunity, and you'll hear more about that from Darren. We have been making those investments significantly over the last couple of years. We've got some work to do, and we think both in terms of SSDs as well as eMMC, as well as really these combined DRAM and NAND solutions, there is just a ton of opportunity there.
Great. A follow-up on Hybrid Memory Cube.
Yeah
and through-silicon via readiness in general.
Yeah.
It seems like Micron's focus for HMC is mainly on the enterprise side. Is that a fair interpretation? Second, do you see TSV getting introduced in other areas like mobile in the next 12-18 months?
Great question. Really, HMC, I think it's probably proper to think of HMC as a solution best utilized by the enterprise side, in what I would call the short to medium term. We are believers that on the client side, that would include mobile, pieces of HMC, there's a lot there. That includes some of the packaging technologies that go into the stack, the DRAM stack itself, the technologies that go into the memory management, the technologies that go into the error correction. We do think that pieces of those, over time, make their way out to the client market. Will it be something that we call HMC? In the next 18 months, probably not. HMC, the product itself, is a very upper-end product for servers as well as networking. The core components that go into HMC absolutely will be making it out to the client spaces.
Next 18 months, I think you'll see maybe limited uptake of that. Think in terms of really kind of the two-to-three-year timeframe. Yeah.
Hi, Brian. You mentioned that you are developing the in-house controller for HMC and PCIe.
Yeah.
I guess it's for SSD, right?
That's right.
It seems like you are leveraging third-party controller companies like Marvell and Phison.
Sure
in the eMMC and the SSD. First of all, I guess, how do you define the in-house? If it is a-
Yeah
hardware from third party and put it in your firmware, is that considered to be an in-house solution? Long term, what's the pros and cons for the in-house versus third party? Can you separate into SSD and eMMC controller outsourcing strategy?
Yeah, great question. For clarity, we do use a combination of both in-house and third-party controllers today. In-house, we have developed our in-house controllers both for HMC and our PCIe drive. Then we use, on occasion, external controllers as well. Frankly, on a go-forward basis, we're fairly pragmatic. We look for opportunities where we will invest the right level of resources to make sure in spaces, in products where we think it is best for Micron to develop our own controllers, be it for advanced management reasons, for IT reasons, that we have the capability to do so. There will be other times-- NAND, from my perspective, we could ship every NAND device right now behind some kind of a controller. I don't think it's necessarily realistic for Micron to build 20 different flavors of controllers ourselves.
What we're interested in doing is making sure in the most critical high-performance applications, as well as those where we need to make sure from a management perspective, we can get that solution shipping quickly. We have the capability to do that with in-house controllers if need be. We may use external controllers as well for some of the other applications. This is something we stay fairly pragmatic about. We have developed our own good internal controller base to ensure that we're ready for whatever the interface or conversely, the NAND management side of the picture needs to be. We'll be ready with a controller one way or another.
Very much. Second question, very quick on near-term dynamic as well, because Apple commented that for the September mobile DRAM price is coming down.
You also mentioned that PC DRAM price is coming up. What is your mix shift into a cycle? Are you going to do more PC DRAM in 3Q, and what does it mean to your blended margin in ASP?
I'm not going to speak to any particular customer. What I would tell you about these markets in aggregate is that we are pleased with the near-term ASP trends in all of these markets. To Mark Adams' point, we do have a certain amount of flexibility of balancing across our network for optimum profitability. We are certainly in a position where we make sure that as we balance that profitability, we're keeping an eye on the long-term landscape as well. I would just tell you that we are oversubscribed today, so pleased with the trends. I think we're set. With that, what I'd like to do is introduce as the next speaker, Darren Thomas. A number of you have had a chance to meet Mike Rayfield. Mike is here with us today.
We think a great example of where we have taken our mobile business unit and placed it under the leadership of a mobile expert that came from the industry and truly understands the care abouts inside of the mobile space. Recently, this last March, we brought Darren Thomas to the company. Darren comes to us most recently from Dell, and prior to that, from Compaq. Over about a 24-year career, 15 years at Compaq and then nine years at Dell, Darren had the opportunity to actually run the enterprise storage businesses of these firms, learning firsthand what these operations care about as they push hard into the true enterprise storage systems space. It's been a pleasure to have Darren with the company.
He's going to tell you a little bit more about client and enterprise SSDs, I think really doing a great job bringing to Micron a true end-user perspective of what's needed to make these products successful. With that, Darren Thomas.
Well, it's good to be here. As Brian noted, that I come from a different end of the industry. I'm not from the chip and the silicon side. As Mark Adams put it, I'm from what's above the silicon. The silicon is very important, as many of your questions have pointed to, what's above matters a lot also. Matter of fact, it may, in some cases, even matter more to what's going on. I'd like to point to the industry and just talk about what's happening now. Not necessarily trends. Trends would be more things that are going to happen very quickly. Just looking at this industry. Today, 90%, probably 99% of where NAND is being used is speed. It's being used because it's over 1,000 times faster than a hard drive in most cases.
The IOPS of the NAND that we sell, even in the client space where we tune them down to be more cost-effective, even at that space, we're talking nearly 1,000 times more speed. That's the market today. That's not all NAND is capable of. I talk about redefining the future of storage. NAND, we've made it look like a hard drive. We've made it act and perform like an SSD. There are startup companies out there right now that have taken 44 terabytes and put it in 1U. A soldier can jump out of an airplane with the entire database of the military on his back at that capacity. That's what's possible now, and that's what we're talking about. There are companies that are taking 500 terabytes and putting it in 1U, probably within the next year. 500 terabytes, a petabyte.
Just to put that in perspective, EMC, our good friend and a partner of ours, to currently sell that capacity, takes 168U. That's four and a half racks full of disks to make what fits in 1U and uses 500 watts. Uses literally the power of a light bulb to operate that. Those are transformative kind of things, that hasn't even hit the industry yet, I'm just talking about size and power. There's also things like resistance to vibration and shock. NAND doesn't have that problem. Spinning drives do. There are data centers that are popping up. We had a customer come out to Micron the other day, he talked about a lights out data center. When he says lights out, he means he's going to shut and lock the door. No human could go in that building for one year.
That's his service model. It's not that he's trying to see how dark he can make the room. He's trying to save money on people going in and changing things and touching things and upgrading things. He's basically saying, "I built a data center. It has enough redundancy in it. I'm not going to touch it for a year." He has two major requirements. There can be no fans in the device. It has to operate at room temperature, and there can be no spinning drives in the device because spinning drives fail too frequently. SSDs operate today with very little effort on the part of us, very little extra effort. They operate about 120 times more reliable than a hard drive. That's significant. It's beginning to question whether you need a hard drive or a drive at all on a hot plug device.
Does even hot plug make sense when the failure rate goes from roughly 6% a year to 0.1% a year? You're starting to change the way people think about data centers. They can turn the lights out, pull the people out, and run the system at a significantly lower cost. You can tell I talk like somebody who's from this other end of the market, and I credit Micron. It was very smart. I know Mike Rayfield comes from the same background I do, but it's these kind of trends that we need to understand and comprehend because it may take us a year or two to make the controller that goes on the NAND, that took us two years to design, that goes in an SSD that took me one year to design. I'm talking four years in advance, I better know what's going on.
These are the trends that we're talking about that we face today and what we have an opportunity to do. I want to talk just a little bit about client. We get the question, what are they going to do to get the margins up? Well, I'll tell you right now, we didn't have enough investment in client. Client's an important part of the space of this industry. A lot of volume goes through the client world. We had the technical capability to do one drive. They weren't great people, but there just wasn't enough of them. Doing one drive every 14 months will not meet the requirements of this industry. As we sit here today, since I've been here, I've been at Micron about five months, we've quadrupled this capability. We did it with internal hiring.
We did it with partnering, we've basically made it so we can make four drives at the same time. There are three drive capabilities in this space that we want to do. We're basically now running at the speed we should run at. The first thing is we have to have more drives, we have to have more capability in this space. What you'll notice is, even with the prior capability, 98% increase in the quarter-over-quarter revenues, 84% in gigabits. That's great, wait till I can get four drives at a time hitting this market. You can see right now we've got a few things already coming. It says we're sampling the M600. We're not sampling it next quarter. We're sampling it right now. The M600 is the new 16-nanometer drive in the client.
We already shipped the 16-nanometer drive here, the MX100. Very first ones to ship it. Within a quarter, we'll be shipping a client version of the same drive. We're moving very quickly now, getting these products into market. This is what it's going to take to recover our business. It won't be right away. This client business requires more than just two drives. It requires a significant set of winning of the OEM customers, the Dells, the HPs, the Lenovos. We have to participate in this market. We're on track. We've got the growth rates going now. We've got the products coming out. We've got a TLC-enabled product we'll be sampling in 2Q of 2015. We've got the world's first 16-nanometer drive out today, and it has won significant awards. You saw the one that Scott had on his.
We've won about seven other awards for the device itself. Very fast-growing market. All of this does not assume what's really going on in the market. No one's ready to really predict when this market's going to tumble over. When is the true client space just going to say, hard drives are out, SSDs are in? We're seeing it at the premium clients, the notebooks, the very thin notebooks, the premium clients. We're very near that tipping point. It's almost 100%. One of the very interesting things is this mobile experience of SSDs. There isn't anybody who doesn't know you can't turn on a tablet in seconds. You turn on your laptop, it takes up to 90 seconds. You put an SSD in that laptop and turn it on, and it takes about five or seven seconds. We were just talking in the back.
It could be faster if Microsoft knew there was going to be an SSD in every notebook. It could actually be quicker than that. Software capability. We're approaching a time where you can't do this with a hard drive. We can actually take an SSD. If you're going to be on a plane and your battery's running low, you're going to be able to send a signal through the operating system to make your battery last longer. It'll turn the SSD down even slower. Hard drives can't do this. You can't slow them down. They run at the speed. They're either on or they're off. The devices we have, we can run them at any speed less than its full speed that we want. That capability is coming. Here's a very interesting one. Hard drives are being designed out.
The next generation PCIe with NVMe, there is no hard drive interface for that. That's just an SSD interface. This market is coming very quickly. I think it's very hard to predict exactly when this will tip. I think this kind of growth rate is conservative. Enterprise, we have a similar challenge, probably a little bit bigger, because in the enterprise space, nobody buys an enterprise drive after looking at it for a month or two. You're talking about long design cycles. The design cycles now, instead of being a year, might be a year and a half. The controllers we're talking about. Brian mentioned we've got several controllers under development. It's in this space. We have a PCIe SSD today. We're on its next generation. This next generation will increase the capacity significantly.
I'm not going to tell you now because I want to save something for the announcement. I'm not talking about 2X, and I'm not even talking about 4X. We're talking about a significant increase in the size, the density of these devices. We also have another controller under development because, see this SAS space? It's huge. The SAS space has really taken off because we're now approaching 12 Gig SAS. Six Gig SAS wasn't fast enough for SSDs. When customers tried that's when the PCIe market took off. This market took off because this one wasn't fast enough. Now you have 12 Gig SAS out. It's double the bus speed, and you see a lot of interest by all the storage companies. You see all the all-flash array companies. They're doing this on 12 Gig SAS buses.
The SAS market is really taking off, and Micron, we have an all-Micron design, and I heard the question earlier, so it's our silicon, it's our firmware, it's our NAND, it's everything Micron. The advantage of this is we will be able to make a very high-performing device. We've obviously got some good customer partners helping us with this. It's one of the beautiful things about this industry. Brian mentioned that when we sell to the likes of Dell and EMC, they get very involved in helping us define the way these devices look two and three years in advance. It's a great opportunity, and we're taking advantage of that. Looking at this, there's a couple of very different trends going on here. First, I'm sure there's nobody in the room who hasn't heard that all active data's moving to flash.
The trouble with that is only 10% of the data is active, right? 10%, 15%. I've heard the number as high as 20. Most experts kind of agree on it's between 10% and 15% of your data's active. The rest of your data's not active. Now we have really good software that takes the inactive data and moves it somewhere else. We have this tiering software in the industry. You have this inactive data that's going to cold and cloud. Doesn't sound like SSDs, does it? Doesn't sound like flash. That's really wrong because there's not a cloud company out there that doesn't want to get your data quickly so that they can upload your data and get you online. You would never do it if you said, "Look, I'll buy your online cloud system" if it took you three months to transfer the data.
They need to do the transfers. The front ends of clouds are still going to be flash, but there's something else important happening. We've talked about big data analytics. Well, big data analytics means that customers can now mine older data for important information. When they do that, the belief is about 30% of the cold data is going to go back to active. Some significant amount of data goes to active. The nice thing about this market is, we can move data down to cold and move it back up to hot, and a significant piece is going to stay hot because of big data analytics. The last thing I mentioned, lights out data centers. Probably the most costly thing about a data center is the human resource that goes into it. Certainly, that's true in the U.S. and in Europe.
May be a little less true in Asia. The human cost of data centers is significant, and people are looking at now the technology exists to run data centers longer at a time without human intervention. My last slide, I'd like to point out, obviously, I came from this side of the market. It's very important. I think, throughout my career, I've always wanted to have more control of this part of the market. You can see these 3D NAND and planar technology capabilities. I think ownership of NAND is advantageous. I'm not going to say you have to have it, but I think it's a very advantageous thing because I can go to Scott DeBoer, and I can tell him what customers want here, and he can make it happen here. There's only a handful of companies that can do that.
It's one of the reasons why I came to Micron, because this is the virtuous circle. These customers ultimately define what these people buy. These people define this and this. If you understand the end user customer, you will understand the future. Micron's invested in this. This is who Micron is now. I'm happy to take questions.
Thanks. For the storage business unit, what kind of metrics are you targeting toward? Is it growth rate, share? How do you think about what you're trying to achieve, first of all? Second of all, the tipping point for SSDs, is it a $ per gigabyte delta that does that, or is there some other factors that may create that inflection point?
Yeah, two great questions. The first one, I think of market share, because if you have market share, you can do a lot of things. I don't think we're willing to trade market share for margin. There's an extreme in both directions. Ultimately, what we drive for is to grow this business and to have a significant position in the market. Since we are one of very few companies who actually does create NAND, it kind of gives me a target for how much market is my fair share. I'm a longtime believer in owning my unfair share, but my fair share, I know what that number is. It's very easy for me to see I'm not there now, and that's the first thing I want to get to.
I want to make it clear, I'm not going to trade margins for that because we do have this virtuous circle. From cradle to grave, we own from the silicon to the solution, we're going to manage this very well, market share matters. The second question? Yeah. It's very hard. I've asked this question to customers, it's funny, almost everybody just assumes it's like one to one, when it gets down. I know for a fact from experience, that if I get close, the Seagate and the WD are going to drop. What's going to happen is I'm going to take away their higher margin businesses. I'm going to take away their higher profitability businesses, what they're going to be stuck with is the $0.03 a gigabyte stuff. That's where they're going.
I'm not in a race for $0.03 a gig. I think what happens is there comes a point in time and the total dollar amount makes sense, I'll give you an example of that. When we ship the MX100, you could buy a 256 drive for under $100. Right at $100, you could buy a 256 drive. We sold out in three days. That drive, by the way, was almost four and a half times the cost of a hard drive at that same price. For $100, you could get rid of all your failures in your laptop. You could really remove that risk in your laptop. A lot of people, for $100, they're going to do that. Matter of fact, I was one of them. I bought the 512 gig one.
You see that I don't think it's an x factor as much as it is when people are buying laptops, the average price is $700-$1,000, your difference can't be the SSD is $400. When the SSD gets down to $100-$150, it starts flipping over. I think at the low end of the market, it might be close to two x, two and a half x. When you get to the enterprise space for all these other reasons, power and cooling, space, size, reliability, total cost of ownership, it might flip at three x. We're not there yet, we're close. We are so close. It's amazing. Great question. Catch you at lunch.
Okay.
Thanks. Hi. I've got a question. For enterprise, what is the business model of choice? It looks like you've shown on slide 42, a number of customers, including EMC. That's more traditional kind of SSDs to storage solution providers such as EMC. Would Micron consider an opportunity to potentially move up the value stack and provide potentially storage solutions themselves, potentially capture more value? Perhaps, another way of disintermediation is working through the big scale guys like Amazon, Google, et cetera, Facebook, et cetera.
Yeah. It's a great observation. You notice that I broke this into three colors here. Those are actually three completely different customer sets in the enterprise space. The PCIe one is the end user customers. They are the ones that make that call on which PCIe SSD they buy. The SAS one is the storage companies like EMC, and the SATA ones could either be the server vendors or could be the big cloud guys. We're going after all of those. I think buried in your question was, would we consider going upscale? We're going to consider everything. We're not obviously willing to commit to any of that, but we're not afraid to look in all the areas where we can add value and where customers would buy.
I would tell you that a lot of the cloud guys are already skipping all the intermediaries and coming straight to us and buying either wafers or SSDs directly. That disintermediation's already occurred. That was my last question. Thank you. I think I'm introducing Ron Foster, who's going to bring us home. You have a clicker? Right there.
Well, thank you all for coming. I'm going to cover some summary financial highlights for you. You've heard a lot about the business, and I thought it'd be useful just to give you some financial perspective. First of all, in terms of Micron performance, give you a view of that, and where we've come in the last year. In terms of investment priorities, you've heard from Mark and Mark, both about our priorities as well as from the business unit heads. I'm going to give you a little bit more perspective on capital, and expand on what Mark Durcan said earlier. Then, spend a few minutes on capital allocation and our balance sheet structure. As I'm sure you know, we've made a number of moves in terms of our capital structure in recent months. First, at Micron performance.
We have a strong operating and financial model, and the important thing to keep in mind is our financial model is in place to support the priorities of the business. That's very important to us. If you look at two parameters, a capital efficient business model, we get a high return on capital with our shared partnership investments, such as our relationships with Intel and Nanya Formosa, and low-cost asset acquisitions, the most recent one being the Elpida acquisition that we consummated about a year ago. We also have lower fixed costs coming from this, which generates a higher income and return on assets for our business. You can also see that we have low cash taxes and low operating expenses structurally in our business. The second area is a flexible financial model, and comments have been made about this already.
Our JV relationships and our strategic partnerships help us share cost and market risk as a business. That was important in enabling us to get through our transitions that we've gone through over the last few years, and you see all the moves we've made on the left side of this diagram here to build our business. It's also been commented that our flexible capacity can be shifted to meet market needs. I think that's been covered pretty well. We do invest incrementally to make sure that we've got that flexibility so that we can respond to the dynamics of the market. You've heard a lot about diversified products and customer mix, which gives us additional flexibility so that we can move to the higher return markets as they present themselves and as we are able to access them with new products.
In terms of balance sheet flexibility, I'll cover more on that in a bit. We also have greater access to our global capital markets. We are an international company spread around the world, and that also enables us to access capital markets around the world. There's also been some comments and questions about this, but our capital expenditures can be modulated based on market conditions and return on investment criteria that I'll expand on. First of all, in terms of one performance measure, this shows fiscal Q3 2013 Micron performance a year ago and our latest reported fiscal quarter Q3 2014. Revenue up about 72%. Obviously, we had the Elpida acquisition that came in after that, which contributed significantly, as well as strong market performance. Gross margin moved up about 10 points to 34%.
Notably, net income went up about 16 points to 23% year-over-year, and that income performance is competitive with the best performers in the memory space today at a net income level. Our asset turns are also doing quite well, and Mark Adams mentioned some of the things we've done to improve asset turns. We're running an ROA right now of about 25% against a cost of capital that runs in the 9%-10% range. Again, very good value add contribution. Another way to think about Micron performance, if you look in terms of scale on the left side of this diagram here, whether you look at it at market cap, revenue levels, or net income, Micron is at the upper end of our semiconductor peers in the top five or so in terms of our size now.
From a valuation perspective on the right-hand side, whether you use enterprise value to EBITDA or P/E ratio, whatever, we tend to be towards the lower end of the scale, despite the fact that the stock price has moved up a lot in recent months. Turning now to investment priorities. A lot of comments have already been made about our business investment priorities. I wanted to just expand a little bit on Mark's comments about CapEx. Our long-term capital intensity as a company is declining. This is a Micron graph. If you look at the period 2001 to 2008, our CapEx as a percent of sales ran about 41%. In the last five years, it's been running about 19% or so over that time period.
2007 and more recent 2011 were related to NAND wafer capacity expansion, new fabs in Lehi, Utah, and our joint venture with Intel, and more recently, expansion of our Singapore NAND fab and the build-out of that in the 2011 timeframe. Overall, you can see the capital intensity is moving down over time. If you compare that to industry averages, industry CapEx as a % of sales average runs 8 to 12 points higher over these corresponding time periods than Micron's. That gives you a rough idea of how Micron has been able to perform on CapEx to sales. As we look forward, Scott DeBoer mentioned, and others as well, that it's getting harder to migrate tech nodes as we go forward. They are actually getting more capital intensive, but they also have a longer life for the reasons that were described.
When you look at that combination, we still expect capital intensity to continue to decline in the industry and for Micron as we go forward. Turning now to capital allocation. There's four elements in this diagram here I want to cover. Again, I want to reiterate that our top priority in our financial architecture is to support the business priorities. Part of the way that Micron has got to where it is today is a result of having the capability and flexibility and efficient capital model such that we could support the business as they move forward. We have very robust processes to evaluate return on assets and ensure that our steps and moves that we make as a company at every level, top strategic level down to detailed product level, have a high return and are well above our cost of capital.
On the CapEx side, as I already mentioned, we're guiding $3.6 billion-$4 billion in 2015 timeframe. We are in the same general range we've been running, 20%-25%. As we go forward, as I mentioned, we expect that the CapEx intensity as a % of sales will continue to go down in the future. We're targeting on the balance sheet side, if you look at the right side of this diagram, a minimum cash balance that we have to support our business and to be conservative on our balance sheet structure so we have the flexibility when opportunities are in place. This uses a target range of 12 months of our operating expenses to cover that, as well as our current debt maturities. Today, that put us about $3.4 billion as a target cash balance that we'd like to have on our balance sheet.
We can, of course, evaluate that over time, it can vary based upon those parameters, but that's a range that we want to target at this point. In the lower quadrant, continuing access to low cost of capital is obviously very important to our business. We are a capital-intensive business. Although it is declining over time, it's still a very significant component. To keep low cost of capital, we have a target leverage ratio that we want to try to maintain. We want to get to investment-grade crossover-type credit as we go forward. Part of the element of that is a debt-to-EBITDA ratio below 1.5, we want to stay in that kind of range. We're below that level a little bit today.
Low cost of capital comes from our global footprint, where we can get access around the globe, and we will, in fact, move to capital sources as opportunities arise. I'll expand on that in a bit. Our joint ventures and strategic partnerships actually help us with capital access and cost of capital through our very valuable partners in many respects. If you look at our cash balance, and this is a pro forma fiscal Q3 2014 adjusted for the high-yield bond that we just executed and the retirement of notes that we also recently noticed are our B-notes and the Series B notes that we are in the process or actually repurchased already. It's a pro forma fiscal Q3. Our total cash balance is about $5.3 billion on that pro forma basis.
$2.5 billion of that is in our Japan and Taiwan subsidiaries. These are the subsidiaries we acquired in the Elpida acquisition. It is in place there to support both our installment payments in the sponsor agreement as well as CapEx required in those locations. We've also converted, at the beginning of this calendar year, to a cost-plus model. It was an important part of our sponsorship arrangement, where it has a more predictable cash flow, albeit reduced, into those locations, but also reduces the risk to the Japanese subsidiary in terms of supporting their CapEx and installment payments. Most of that cash was put in place prior to the implementation of our cost-plus model that we initiated in December of this last year.
If you look at other Micron cash, $2.8 billion, add up as a $5.3 billion total that I just gave you in that pro forma. About over half of that is in the U.S., to give you a rough order of magnitude there. Obviously, the cash outside the U.S. is in place to support our operations around the globe and technology migrations that have been described previously. If you build it up in terms of our target cash I just showed you to cover 12 months of operating expenses and current debt maturities, that leaves us with a global cash cushion around the world of about $1.9 billion. There we go. It's a little delayed response there. Some things that have been going on in terms of our access to capital. First of all, as I mentioned, we're targeting long-term investment-grade crossover credit profile.
We were recently upgraded to BB and Ba2 by Standard & Poor's and Moody's, we've begun migrating from convertible notes and underlying the equity premiums associated with those convertible notes towards a straight debt option. This lowers our long-term weighted average cost of capital. You can see in the graph here the bond indexes for B through BBB-rated bonds, they're at historically low levels. We've capitalized on that opportunity, first in February with a $600 million inaugural high-yield offering. We just consummated in July a $1.15 billion high-yield offering that came in 10-and-a-half-year terms, 5.5% rate with investment-grade covenants. I mentioned we support our business with our financial strategy. It's extremely important that we don't have onerous financial covenants restricting the ability to operate our business.
We've been able to make very good strategic moves throughout market cycles historically. The investment-grade covenant package is important to us. We were able to get that along with slightly better than BB terms on this latest high-yield bond. In terms of dilution management, we've also been focused a lot in recent months. This breaks down by quarter, the metered approach we've had in place to our convertible dilution management. We're managing our convertible dilution and trying to work that down within the target ranges for debt and cash that I just mentioned. You can see that in each quarter, in open windows, we have exercised moves in terms of taking out some of our convertible bonds. We have moved in the highlighted yellow there with the high yields to help with the funding to support some of that.
We have exercised every call option we've had as they became available to us in the flexible structures we put in our converts, in the event that the stock price would move up and we needed to move on them. To see the results of that in terms of dilution management, we've spent about 83% of our free cash flow year-to-date on convert repurchases, that's about $2.1 billion through Q3 2014. In total, we've reduced our share dilution at a $32 stock price by about 102 million shares as a result of all those moves, including the benefit of the capped calls that we put in place with our converts. Finally, in summary, I just want to give you a summary view of we're focused on long-term value creation.
It's important to us that our financial structure support the investments that we need to make for operating efficiency. I mentioned the CapEx structure we have in place. Our CapEx is certainly flexible, as I commented on. We have the ability to adjust our CapEx over time as returns warrant. I can tell you that the current CapEx plan, as Mark mentioned, has a projected very high return on invested capital. That's why we're moving on it the way we're moving on it. As market conditions change or the environment's different, we can, of course, modulate that as required as we go forward. Investing for value-added solutions in the business expansion of our markets that you just heard about from several of the speakers. In the returns on capital framework, we've done several things. One is we've established our target capital structure.
I've shown you the cash and liquidity goals, leverage goals. We're aiming towards long-term investment-grade crossover credit profile in those metrics and how we perform as a company. We're continuing to migrate our debt to straight debt as the cost of capital on that debt has been very favorable in the marketplace. We want to get in the low- to mid-single-digit range in terms of dilution exposure from our convertible notes. Secondly, we've been focusing on dilution management, as I just showed you, and we'll continue to do so until our targets are achieved. We've done a lot in 2014, and in 2015, we'll continue to focus on dilution management, again, with convert repurchases as appropriate and/or stock actions to help us with dilution while we continue to reshape our balance sheet.
Annually, we will review our return of capital policy as we maintain our target capital structure, and excess cash will be available for dilution management, share repurchases, and/or dividends as we go forward. With that, it's my last summary slide. I'll open up for any remaining questions you have after several question sessions. I do need to show you, this is the reconciliation schedules that are also available on the website, how I got to some of the numbers I just showed you on the non-GAAP recons.
Hi. Given the previous presentations and opportunities that have been highlighted in the NAND storage enterprise market as well as embedded, are you comfortable with your market share in NAND? Going forward, if that's not the case, how should we look about it, and what are the strategic alternatives you look to expand the NAND market share, if you do?
Are we comfortable with the market share in NAND? I think Darren already summarized that reasonably well, that market share is important, and we're not happy necessarily with where we are. Are you talking about overall NAND in terms of our scale of our business? Oh, okay. I think one way you need to look at scale of a memory company is the aggregate. As you know, we already made some shifting between DRAM and NAND in Singapore, strategically, and we can adjust that. We measure scale across the whole memory enterprise. As a matter of fact, I think it's highly valuable that we have all three platforms, DRAM, NAND, NOR, to utilize, as you've already heard how memory technologies are migrating, and you're even getting hybrid technologies between them. Having the scale and crossing all the memory platforms is highly valuable.
We can adjust strategically if we need to over time, we think we're in a pretty good place both in terms of DRAM and NAND and the recent moves we made with our structure.
Coming back to the earlier question about more customized product, I'd love to understand how it impacts two areas in the financials. The first is working capital. Does it require more working capital going forward or not? The second is the ability to forecast revenues. Does it become easier or more difficult? You've got a more one-on-one relationship with customers, but as you know, customers in a time of tightness will often overbook or else put out a bigger production forecast than they actually are looking for to secure supply. I'm just wondering how you factor that into your own kind of budget.
Well, as Mark Adams already mentioned, we are in several different market segments and market types today. It's not like we're moving into an area we don't understand. As a matter of fact, we are regularly managing, notably since the Numonyx acquisition and also in the mobile market space, markets that are somewhat more customized than our historical commodity DRAM marketplace, for example, or commodity NAND. Even today, in our business units, we are engaged in evaluating working capital requirements and managing projections on revenue, et cetera, that are structurally different I would say that for the more customized product segments, it's been a learning experience for Micron, but we've been going through that learning experience already for four years now. I think we've learned a lot in terms of managing customized products, both in terms of forecasting and managing that in a customer relationship perspective.
As was also commented, it does require a different type of customer relationship, and I think you typically build those out in those product areas where that matters a lot. In terms of working capital, we have to be more certain of our projections that we make on customized products, but there's things we can even do there. For example, we can stop it at partway through a process, what we call a die ID level or that sort of thing, and we have alternate paths we can go on and actually hold it at that point, and still, because of faster cycle time Mark talked about, we can get it to the customer within timelines that are needed. We are actually designing all those capabilities into our business or the various business units, depending upon the characteristics that we're dealing with.
Thanks, Ron. Thanks, Ivan. This question is regarding your JV strategy. It seems now your JV is generating a higher gross margin than yours. Are you going to negotiate a price during the renewal of contract? If not, what would be the reasons?
To which JV are you referring to?
Inotera.
That was a joke. I can tell you about the mechanism we set up. I can't tell you about what's going to happen, obviously. We have an annual process of renegotiation of the arrangement with Inotera. The first thing I'd mention is that it's a highly valuable partnership. We put it in place at a time when we were actually making big moves with both an Elpida acquisition and Inotera doubling the capacity, taking virtually all their output. We needed the financial flexibility, which we had to do with our balance sheet and our financial structuring to be able to accomplish all that at the same time. It's a hugely valuable partnership to us also because, for example, we get low cost to capital access in that location for our partner, Inotera, and that's highly valuable to us financially.
When you look at the financial value equation, first of all, you look at risk and return. It's been highly valuable. Secondly, cost of capital and asset utilization, it's been phenomenal. One of the things I regularly do for the management team is go through and show the return on asset calculations for our joint venture relationship, and they're very competitive with our business averages when you adjust for all those variables. The first point I'd want to make is that it's a highly valuable partnership to us. We have a great relationship with our partner, and we do have an annual plan to review and revisit terms and conditions, which you'll have to just stay tuned on that one, but that's how it works.
Thanks. Ron, would you just summarize for us some of the key assumptions you're making behind the CapEx guidance you gave us for fiscal 2015? One example might be expected capacity additions in DRAM and similarly in NAND from the industry.
You're referring to industry expectations. Let me back up to maybe a prior question that was asked because I think it contextualizes what you're asking. We have significant ability to modulate our CapEx. There was a question asked earlier about, because it's technology-related, does that mean you have to spend it? The fact is that we can cadence and modulate the time phasing of CapEx expenditures, both in terms of rollout as well as how much we convert for a period of time. You also heard that we phase fabs as we roll them out. All that can come into play in terms of how much hits a fiscal year in terms of CapEx. Right now, we have a very rigorous and robust process to look at return on invested capital on every element of our capital plan.
What I can tell you, and others have already commented, we have right now a great return projection for everything we're doing, and that's why they're there. Obviously, if that equation changes, we even have a triggering process to keep track of whether something changes, evaluate it every month, and can alter course. That can mean a deceleration or an acceleration, depending upon the economics. I can tell you right now, what Mark was referring to is we see a great ROI on the capital we're putting in place. To your question about the industry, that's tied to the supply and demand assumptions that we just went through previously.
Okay, thanks, Ron. Just a couple of follow-up questions regarding your CapEx number. I think that your DRAM CapEx and NAND CapEx, not much increase year-over-year basis next year. However, the solution area, when I look at your chart for page nine, I can see the sort of new CapEx increase for technology and product enablement. Would you provide us a call? Is it sort of a more controller development or in-house controller manufacturing or any other software development? When I look at the SanDisk in the SSD area, they've been very active in the M&A. Are you planning to do more active in M&A for the solution areas? This is my question. My last question is the tax.
The magic number is very low tax, how long and how you can maintain maybe only the 10% effective tax rate even five years from now? Thank you.
Okay. A lot of questions there. In terms of the capital question, well, let me hit the tax one first, just get that out of the way. If you look at our cash tax rates today, they're low single-digit rates. On a book tax basis, that's a little bit higher because we have DTAs in the Elpida acquisition, et cetera, that come into play there, you have a higher book tax rate, but cash taxes are in low single-digit rates. To your question about how long does that last, we have NOLs of significant magnitude in the United States, and we also acquired them in Japan. Our tax structure with our Singapore principal structure in place enables us to utilize those NOLs over a longer period of time. It's years, even at fairly high cash flow rates as you go forward.
I can't give you a precise number because it's a function of profitability. We do have a structure set up that'll optimize those low cash tax rates for a period of time. Your first question was.
guidance for next year.
Yeah. How much that other category.
Other than DRAM. Sure.
Yeah.
Because
Yeah.
Technology of the
Sure.
I wonder why you think about the or higher.
Yeah. The question is about the other bucket, which Mark mentioned was about 30% of the CapEx spend next year. First of all, as already been commented, we've got a lot of new technologies we're developing, working on. That includes our R&D technology CapEx, and we are expanding in a number of new areas in terms of our future memory technologies. It also includes infrastructure investment and other categories you mentioned, such as CapEx associated with support of the solution space and back-end processes, which are somewhat varying as well. That's the reason for the expansion in that category. As I mentioned, we can modulate these CapEx categories based upon market conditions, but that's our scoping of that of it today.
No, that was the second part of your question, is that, again, just to be clear what has been said previously, this is all tech node migrations, which can be modulated over time as we go forward. We can also roll them out in a different phasing if needed. Our current estimate is that we have a very good return for this level of CapEx, and given the market supply-demand equation that's been characterized to you. We can adjust it over time, but right now, that's our view, is that that's the best ROI for Micron moving into the market environment we're looking at.
Thank you.
Is that it? Okay. I'll now turn it back to Mark Durcan for wrap-up.
All right. Thanks, Ron. I just want to wrap up here quickly and then take a few additional questions on anything we've covered today. Before I do, we've talked to you about the industry structure, and we really do believe that we're in an environment that's very beneficial on a go-forward basis, given not only the consolidation that's occurred, but also the dynamics in the overall memory industry vis-à-vis technology migration and product and end-market diversification. We're very focused on making sure we make well-rationalized decisions relative to our spending that really drive cost-efficient manufacturing and value-added products for the marketplace. There's been a number of questions around market share that have come and gone here. I want to make it absolutely clear, Micron feels like we got plenty of scale and we got plenty of market share.
When Darren talks about market share and value-added businesses, he's talking about revenue market share. Really what he's saying is, "I want to make sure I'm getting Micron bits into the most value-added segments." That's true across the whole market. We don't need more share. We just need to make sure we're getting our bits into the most value-added segments, and we continue to be very focused on that. We also want to make sure we stay efficient. This is a tough business. It's always been a tough business, and we'll make more money for our shareholders if we stay operationally lean and mean, and that's what we're going to do. Mark talked about cycle time and all the virtuous things relative to yield learning and flexibility for market demands that come with that. We're also focused on our whole supply chain.
How do we get a more cost-effective back end into some of the legacy Elpida products? How do we get the most effective distribution mechanisms for our products? Whatever it is, across the whole gamut of inputs to our supply chain, we're very focused on operational improvements. We're not just a component company. We think of ourselves as a memory systems company, memory subsystems company, and we're developing leading-edge technologies to support that, whether it's NAND or DRAM, or whether it's emerging memory. I did hear some questions around ReRAM, and where does that fit, and how does that all play out? The early deployment of some of these things is, of course, it's going to be in value-added sockets.
When we talk about investments we're making from an R&D perspective to support the future, that not only supports our future growth and relevance to our customers, but it also enables us to deliver more value-added solutions because we have a broader spectrum of end products. Today, if you think about who's got the broadest product portfolio in the memory world, it's Micron, bar none. We want to make sure that we're there in the future with some of these advanced memory products as well. We're the most global company with the best customer relationships, and we're leveraging those across all the market segments. We're not focused on any one segment. We are very deliberate in making sure that we maintain the opportunity to play in all those segments.
We don't care about market share in any one, but we want to make sure that we're positioned in all of them, so that over time, as the markets evolve, we remain a viable and valued supplier to all those customers in all those segments. Finally, as Ron talked to you about, we think we've got a differentiated financial model and business model that has served us well and will continue to drive those types of efficiencies moving forward. We're very focused on the memory space, but we're also very focused on how we drive return for our shareholders. I think you can expect that we'll continue to have that thought process and mindset as we move forward.
I did want to do a little bit of redirect on the capital that I talked about up front, because I heard a number of questions in the back of the room around some of the things I said. First of all, this whole category of flexibility and precision. How precise are these numbers and how flexible are they? They are not precise in any given timeframe because we spend capital when we accept tools, and we have tools coming in, and we never know exactly how those numbers are going to play out. Relative to fiscal 2014, we told you on the last earnings call would be toward the upper end of the range of what we had forecast for the whole year. That is still true. We will fall in that upper end of that bucket.
The precision in that, I cannot tell you exactly which tools we accept, but that is sort of the spend rate that we are targeting for 2014. This range I just gave you for this year, again, we do not know precisely what that is going to be, but it is all very well-justified on an ROI basis. In particular, this top bucket, these things that you look at, and it is sort of my other category, that is emerging memory technology to support value-added products for the future. Think of those not as something that is impacting supply in any big way, but not insignificant investment in the future and in future value-added products, whether that is potentially early ReRAM to produce best-in-class storage solutions or to support new memory architectures out into the future.
There is a chunk of that is around back-end efficiency, whether it is some of the packaging things that Scott DeBoer was talking about, or whether it is the enhanced test flow that drive lower costs across the mobile DRAM product portfolio, that there were some questions to Mark Adams about. Whether it is emerging memory or back end, there are things in that value-added bucket, which is bigger than it has been historically, that clearly have a return for Micron, and that is why we are making those investments. I think there were some questions around supply also, in particular, I think around, are we in any way concerned that the supply growth in 2014 is coming in a little higher than we had originally projected? What does that mean for the future and for the assumptions that are embedded in our capital plan going forward?
What I would say is, we actually think it is positive. What it means is that the 20-nanometer nodes that people have been rolling out, the thing that is driving the bit growth up a little bit higher is that that technology has gone relatively well into manufacturing at some of our competitors. By the way, that is true for Micron as well. We are very happy with the way our 20-nanometer technology is coming together and the way that is starting to fold into the manufacturing fab. Gives us confidence that the capital spend I told you, at least that bottom bucket, that 50% that is going into DRAM, still makes sense. We will look at those numbers and adjust them if market conditions change or if we learn new things about our yield learning, but we feel pretty good about that.
We feel pretty good that the growth that's happening in the second and third quarters relative to DRAM supply being slightly higher than we thought, we think that actually is very positive because it says, yes, the memory market's robust and can easily absorb all that supply and still have a positive market environment. Yes, that means there's less supply growth coming out into the future because that transition has already taken place more quickly than people anticipated. Let me stop there and answer any additional questions that are out there in the room. John.
Thanks, Mark. Mark, Micron was very early on the thesis around industry consolidation, and you were also very active in consolidating the industry. Now, as you guys talk about systems, subsystems, value add, you're saying all the right keywords. The undertone is: don't think of us as just a cyclical commodity. Yet when I look at your margin profile, gross margins of 34%. If you look at your R&D as a percent of revenue, it's 8%, 9%. Those to me don't indicate non-commodity value-added businesses.
You guys have gone out on a limb now and given us a revenue forecast on a quarterly basis. I'm kind of curious, as you think about pulling this whole strategy together, how should we think about the longer-term margins-
Yeah
of this business as sort of a benchmark of success as to whether or not you've been able to transform this from more than just a cyclical commodity?
First of all, if you think about gross margins, I think I'm out there as saying I think this business can sustain much better gross margins going forward. I'm not going to jump on the hook as to what that is exactly and when. I do think there's room to improve gross margins, and in particular for Micron, I think there's room to improve gross margins. Having said that, there are some things that structurally drive to lower gross margin and some of the things around the Inotera relationship and the Intel relationship that Ron alluded to. Yes, I think we can improve all of that. We have grown our R&D spend roughly 30% here in the last year and a half. I'm very happy to do that.
By the way, we can do that cost-effectively because we have much larger scale than we used to. When I think about what Micron is able to do today relative to driving advanced memory solutions relative to what we could afford historically, we're in much better shape than we've been historically. I'm not opposed to spending more money. I also want to be efficient. I'm not going to go out there and waste the shareholders' money on R&D spend that I can't be very, very comfortable is going to generate a return. What I'd rather do is go create joint venture relationships with Intel for emerging memory solutions or with Sony for emerging memory solutions, or with IBM for emerging memory solutions, and try and get more done with less. That's what we've done historically, and we'll continue to do some of that.
I am prepared to spend more to make sure we're delivering those solutions effectively, whether it's advanced components or systems. Did I cover everything there? Okay. Yep. In the front here. Oh, sorry, all the way in the back.
Thanks. Couple of questions, a bit more longer-term in nature. First of all, some of the people who talked before you talked about memory getting closer and closer to the processor. Now we have in-package memory. A little bit further out, how do you think that evolves? Are the IDM companies, the logic companies, starting to invest in memory? Some of your customers are talking about monolithic process itself.
Is it going to be a joint venture road like you're exploring with Intel?
Yeah.
That's part one of the question. Second, especially for high-volume applications, now memory is probably the second most expensive, memory plus storage combined, after the screen for a smartphone or a tablet. At what point do some of your customers, especially the customers who have very deep pockets, think about putting in some more capital to work either through a joint venture with yourself, you have entered into those kind of partnerships in the past, or maybe somebody else?
Yeah.
How does the dynamic work?
Okay, great. Thanks. Let me handle the first part first. Relative to this whole memory processing convergence, I'm a big believer that that is the future of the electronics industry. I think of it as this is me because I'm a memory guy, but also because I think it's true. I think the future is a memory-centric computing architecture as opposed to a processor-centric architecture. That's just because of the diversity of the end applications that we're serving today and the diversity of the processing requirements in all those end applications. I think what you're going to see is, yes, this is important, but the square inches of silicon in memory continue to grow, and the processing requirements relative to the memory requirements in all those applications have a very different balance on a go-forward basis than they have.
What I think you'll see is you'll see Micron continue to work on things like Automata Processors and other ways of getting processing into the memory, but you'll also see more and more bits of processing that don't require the absolute leading-edge logic process get embedded into the memory itself as opposed to the other way around. I think that there will be some level of onboard memory, but it's not SRAM in the future with processors. I think the bigger trend is more in the other direction for all the internet of connected things that'll be out there in the world. The second question around, I think broadly I could describe it as new entrants into the marketplace, whether it's customers or new entrants from wherever they come from. It's very, very difficult.
I think getting into the memory business is probably the toughest place for anybody to get into because the technology requirements are so high, the R&D spend is so significant. There are intellectual property barriers. There's a very deep capital spend, very capital-intensive to create new capacity today. It's one of the reasons I'm so happy that Micron has this very solid installed base that we've acquired over the years very cost-effectively now. For a new entrant, you've got to be willing to lose a lot of money. Even well-heeled customers, I think, are unlikely to head down that path without thinking it through very, very carefully. One more in the back, and then we'll come up front. Sorry.
Hi. Thanks, Mark. For the cash return, Ron stated now that Micron has fixed its target capital structure, and considering that you now have $1.9 billion additional kind of cash cushion in addition to the target capital structure, target cash you need. I'm just wondering, Ron obviously said about continued focus on dilution management into fiscal 2015. I'm wondering if that might include something a little bit more aggressive, such as actually buying back stock or even dividends.
I think, first of all, you should understand that as we've gone through this process of capital return over the last year, the capital structure of the company has already changed. As we move forward, we'll continue to think about those different ways we can potentially return capital in the future. Our thinking on that will evolve over time, and so while historically we've been very focused on the converts, obviously as we buy those back, the calculus will get different, as we think about the future and future ways of returning capital potentially. Having said that, relative to the existing cash cushion, that cash obviously is distributed all over the world.
We want to make sure we maintain flexibility, as Ron alluded to make all the moves we need to make in our business as we navigate through what's a pretty interesting time in the memory business, with lots of opportunity to invest for the benefit of our shareholders as well. We're not in a rush to go out and spend our cash cushion down to zero. We're going to continue to be measured, methodical, and prudent in terms of what we do with our cash and how we return it over time. In the front now, finally. Probably the last one here.
Okay, great. Thank you very much. Maybe my question is over the competitive landscape, because everybody's saying the memory chip industry is well-consolidated. But for me, once I count the number of the NAND chip makers, Samsung, Hynix, Toshiba, Samsung, Micron, partially Intel there. But the question is, as long as Samsung really, really paves the way for the 3D area, they already seem to be one year ahead of Micron in terms of 3D. So once you have some technology bottleneck dealing with 3D from next year versus Samsung already build up the capacity in China, and then they already proved the one-year track record for now, but not really impressive so far anyway. The question is, what are you going to do with the NAND business if Samsung becomes more powerful in 3D next year?
Meanwhile, what will be your contingent plan if the 3D does not work, unfortunately, in your fabs? This is maybe the first question. Secondly, when I look at the DRAM, we've been hearing that maybe ASML or others will deliver the EUV. But I've never seen EUV in the DRAM manufacturing side. But you are already talking about the 1X node for the next couple of years. But what would be your contingent plan if the EUV does not work? Maybe quadruple patterning possible maybe? Yeah. Thank you.
All right. First of all, relative to 3D NAND and the path forward there, I am very, very comfortable that the technology Micron has is not behind where Samsung is. Our technology, we are very excited about, and we believe that it will be relevant to the marketplace, as relevant if not more relevant than anything any of our competitors are going to deliver. We feel pretty good about that. Do we have backup plans or contingency plans? Well, we'll be flexible on our CapEx spend. We'll continue to evaluate our assumptions going in and make sure that those still make sense. Micron has a very deep technology capability in the memory space, and we also have a lot of good emerging memory technologies under development.
While those are not replacements for NAND flash, as we alluded to earlier, those are more storage class memories that sit between NAND and DRAM and may eventually coexist with both of those or replace some level of DRAM as opposed to NAND. We also have a lot of different activity going in those different types of technology that can also be value-added solutions for the future. Sorry, the second question was again?
Regarding the EUV.
EUV. Micron is not dependent on EUV for any part of its technology roadmap going forward. We would love to see EUV be cost-effective for deployment in the memory space. It's not there today. When it is, we'll be happy to deploy it, but the barrier gets higher and higher. Even today, if EUV were to start working, in almost every case, we would have to use it in a double patterning scenario. It's not automatic that you cut out a lot of costs, even if you could deploy it immediately. We do have other techniques that will allow us to continue scaling without EUV, but we'd still like to have it available.
Having said that, the other thing that's difficult for EUV is that as time goes on and we install additional multiple patterning technology in order to move to 20-nanometer and 16-nanometer and to deploy advanced emerging memory technologies, that capital's already spent. The cost equation gets different relative to deployment of EUV. As time moves on, it gets more difficult. We're still very hopeful we'll have it in our portfolio in the future. I want to wrap up. Thank you, everybody, for coming today. We think we've got a very bright future. We're very comfortable with the market conditions, and we're very comfortable with our internal capabilities and abilities to deliver advanced memory solutions to our customers moving forward. Thank you very much for attending.