Micron Technology, Inc. (MU) Earnings Call Transcript & Summary

Ladies and gentlemen, thank you for standing by and welcome to the Mobile Business Unit Update, Winning in the 5G Era Conference Call. [Operator Instructions] Please be advised that today's conference is being recorded. [Operator Instructions] I would now like to hand the conference over to your speaker today, Mr. Farhan Ahmad, Head of Investor Relations at Micron. Thank you. Please go ahead, sir.

Hello. I'm Farhan Ahmad, Head of Investor Relations at Micron. Thanks for joining us for today's presentation and Q&A session. On the call with me today is Raj Talluri, Senior Vice President and General Manager of Micron's Mobile Business Unit. Dave Zinsner, our CFO, will also join Raj during Q&A. We request that you keep your questions focused on topics related to today's presentation. As a reminder, the matters we will be discussing today include forward-looking statements. These forward-looking statements are subject to risks and uncertainties that may cause actual results to differ materially from the statements made today. We refer you to the documents we filed with the SEC, specifically our most recent Form 10-K and Form 10-Q, for a discussion of risks that may affect our future results. Although we believe that the expectations reflected in the forward-looking statements are reasonable, we cannot guarantee future results, levels of activity, performance or achievements. We are under no duty to update any of the forward-looking statements after today's date to conform these statements to actual results. I'll now hand it off to Raj.

Thanks, Farhan. I would like to welcome you to this call. By the way of introduction, I joined Micron 2 years ago from Qualcomm that I worked for 9 years and was part of the leadership team that brought Snapdragon to market. Prior to that, I was at Texas Instruments for 16 years and also part of the mobile leadership team there. Having worked for over 25 years, focused on mobile segment of the semiconductor industry, I can tell you that the need for memory and storage in mobile has never been greater, and this presents a tremendous opportunity for Micron. It's a big reason why I joined Micron. Before diving into the presentation, I want to briefly cover the near-term business environment. As we mentioned on the last earnings call, COVID is adversely impacting the mobile demand. Since then, as China has reopened, it has been widely reported that the mobile demand has rebounded. We are hopeful that as Europe and U.S. start to recover, we will similarly see a recovery in the mobile demand in the coming quarters. While the near-term environment is clearly challenging, I will primarily focus on longer-term trends in our memory business unit and how we are positioned to win in the 5G era. Now turning to our presentation today. Today, I'm going to discuss the 5G opportunity with a focus on Micron's Mobile Business Unit, our mobile strategy and the financial results we are achieving. I'll start off with an overview of 5G. 5G presents a huge opportunity because it will mean a 10x-plus change in network performance, which will drive the proliferation of connected devices. As for mobile specifically, it will be a catalyst for unit growth and content growth. The 4G to 5G transition is unlike any before because it combines a high magnitude of change across all 3 of the most important technical vectors that define a network performance: speed, latency and the density of devices that can be supported by the network. 5G download speeds of 20 gigabits per second make it possible to stream 360-degree video in high resolution, enabling AR and VR applications, and also sports streaming where users can control the vantage points and the camera viewing angles. Low latency means systems that are far apart can operate as if they're in the same place. This means massively multiplayer online games with imperceptible lag times. And in health care, the possibility of remote surgeries via connected surgical robots. And the 10x higher density enabled by the 5G network means you can have 1 million devices per square kilometer, enough to support the Internet of Things and machine-to-machine communication. Today, I'll focus on what 5G means to our mobile business unit. But it's important to understand that 5G will also drive secular demand across other Micron business units. 5G will drive a proliferation of connected devices at the edge and networking and data center infrastructure at the core for applications and services that support these devices. Together, the interaction of these devices with each other and data center will result in an acceleration of data creation, analytics and storage, along with it, the demand for Micron's products and solutions. 5G will lower the cost of data transmission and will usher a new era of applications and services, many of which are hard to imagine today, but that will be realized over time. These will be transmitted to our lives as creation of the Internet has been. Turning to mobile. One of the early applications we expect to see is video everywhere. The faster speed and lower latency of 5G will enable 8K live streaming, AR/VR-enabled content and multichannel broadcasting, offering users selected viewing angles. Video is data-heavy and will drive the need for both DRAM and NAND content. As consumers interact with more video, the amount they will store for off-line access will also increase. 5G phones will have greater levels of memory and storage content than 4G phones, and we can see this already in phones being introduced in calendar 2020. Even early generation of 5G phones need more content because they need to be able to support the future applications. In addition to video, which I mentioned earlier, these applications will include AR and VR, gaming, multitasking and more sophisticated always-on AI assistance. In flagship Android phone, defined as those selling for more than $600, the average DRAM content will increase from 6 gigabytes plus to 8 gigabytes plus, and the average NAND content will increase from 128 gigabytes and 256 gigabytes to 256 gigabytes and 512 gigabytes. For some premium flagship phones, we see phones with 6 -- 16 gigabytes of DRAM and 1 terabyte of NAND. In high-end, we see similar increase in DRAM and in NAND as in flagship, but using MCPs instead of discrete DRAM and NAND. MCPs are multi-chip packages, which combine DRAM and NAND in one package and simplifies smartphone design and production for our customers. We see the strongest content growth in the low to mid segment, which is also the biggest part of the market in terms of units. 5G phones priced near $300 have 6 gigabytes of DRAM and 64 and 128 gigabytes of NAND. As integrated application process and modems reduce the 5G phone BOM cost, we see the potential for sub $200 5G phones. In addition to the continued content growth, we expect 5G will reaccelerate the replacement rates and return the market to unit growth. Whereas the smartphone units slightly declined from 2016 to '19 and declined unexpectedly in 2020 due to COVID, we anticipate that there will be growth from 5G every year for the next few years. This return to growth will be a positive driver for our demand. 5G smartphone demand has remained more resilient during this period of demand weakness due to COVID. Data released is showing that 40% of the mobile phones sold in China in April were 5G phones. We remain confident in the growth trajectory of 5G phones and expect approximately 450 million 5G units in 2021. Combining the unique trends with the content growth, we see bit demand CAGR of 15% in DRAM and 30% in NAND from 2019 to 2022. Our demand forecast have been conservatively revised lower due to COVID. As always, we remain disciplined in growing our supply in line with the evolving market demand trends. Now I'll discuss our strategy in mobile. Our team execution and technology and product leadership has put us in an outstanding position to win in the 5G era. Our global network of manufacturing, technology and business centers of excellence play a key role in our strength. Our global footprint allows us to benefit from scale while streamlining processes and operations. It also brings together some of the world's brightest talent to work on our most advanced memory technologies. We also support our customers' innovation and success by being where our customers need us with local sales offices and customer labs. Through our diverse global operations, we deliver comprehensive collaboration, support and quality throughout the product life cycle. Micron's Mobile Business Unit has expertise in understanding the usage of memory and storage in various mobile applications. We take a holistic approach that considers the overall mobile system and examines how our products can affect the application performance and ultimately, the experience of the smartphone user. To do this, we attract the industry's best talent with extensive expertise that spans silicon to systems. Additionally, we have organized our team so that they're highly collaborative, both internally within Micron and externally across the ecosystem. Our teams work with alignment across products, operations and sales so that our road maps align to our customers and ecosystem partners' needs. The new Micron has gone through a dramatic transformation by rapidly improving our execution. Looking back at FY '18, we were late to introduce 1x DRAM, and our mobile products were coming even later as we prioritize other segments in our early ramps. The net result was that our mobile products were more than a year behind competitors. Fast forward to today, we've gone from a laggard to a leader. We have been the first to introduce 1z nanometer based LPDRAM, first to bring LP5 technology to market, the first to introduce uMCP5, combining LP5, the advanced UFS3.1. We were also the first in the industry to support the ultra-high speed 6.4 gigabits per second throughput on our LP5 products. In NAND, we've gone from lagging industry by more than a year in 2018, to catching up with industry this year as we are targeting the introduction of UFS3.1, in line with the market. In summary, whereas 2 years ago, we used to lag the industry, today, we are leading on process technologies, new interfaces and speed. As a result of this execution, Micron has emerged as a technology and product leader in mobile space with a complete portfolio of mobile products that are qualified with all the major chipsets. Qualification with mobile chipsets from companies such as Qualcomm and MediaTek accelerates the qualification of products with smartphone OEMs. Additionally, the breadth of our product portfolio offers our customers greater flexibility to design phones that better match their target market segment and that redefines the user -- end-user experience. Micron's LP5 is a great example of a product that redefines the end-user experience. Building on the success we have had with LP4, we'll lead in LP5 technology with a leadership in power efficiency and speed. Our LP5 product has approximately 20% lower power consumption for bandwidth-intensive use cases, including HD recording, compared to our competitors' LP5. Our LP5 is also the fastest low power DRAM in the market, delivering 6.4 gigabits per second, translating into better performance for AI applications, superior photo and video and more. Now I'd like to highlight a few key results that demonstrate how we have improved the performance of our Mobile Business Unit. Micron has been delivering solid results, gaining share in the most attractive parts of the mobile market and driving strong cross-cycle profitability. Although the smartphone annual unit sales declined from 2016 to 2019, Mobile Business Unit demonstrated a strong growth during this period. Our revenues doubled from 2016 to 2019, much faster than the growth of industry, which grew approximately 28% during the same period. This relative outperformance to the mobile DRAM and NAND industry was predominantly driven by growth in our mobile high-value solutions, which include MCP and managed discrete NAND products. Mobile high-value solutions combine a controller and firmware with our NAND component. Combining these parts together increases the product cost per bit, but it increases the value even more. The value in high-value solutions comes from reducing our customers' R&D expense and speeding up the development and qualification time. Mobile high-value solutions now represents the majority of the mobile market. We have made tremendous improvements in our product portfolio in this area, allowing us to nearly triple our high-value solutions revenue from 2016 to 2019 and to broaden our customer base to include all major smartphone OEMs. Micron also continues to maintain a strong position in discrete DRAM market with our industry-leading LPDRAM offering. Our LP4 and LP5 products have approximately 20% lower power consumption in bandwidth-intensive applications -- in bandwidth-intensive use cases compared to our competitors. This advantage also differentiates our MCPs from our competitors. The growth in our high-value solutions also positively impacts our profitability and reduces margin volatility. Combined with our faster-than-industry cost reductions and improving industry dynamics, this has helped to improve the MBU cross-cycle profitability. Through the last 4 years, we have demonstrated operating margins of a very strong 33% for the Mobile Business Unit. In summary, Micron's Mobile Business Unit is strong. We have a world-class team, technology and product leadership, and we are executing to win in the 5G era. I will now open the call for questions. And our CFO, David Zinsner, will join me.

[Operator Instructions] Our first question comes from John Pitzer with Crédit Suisse.

Raj, I have kind of 2 bigger picture questions that I think capture some of the concerns investors have. First, on the NAND content story for 5G. Just help me understand kind of the tension between storing more at the phone versus having ubiquitous connectivity and storing more at the center, especially as OEMs kind of struggle with BOM cost on 5G phones? And then secondly, just relative to the Huawei bands, I'm kind of curious to get your perspective on whether or not that slows the deployment of 5G infrastructure in China and what impact that may or may not have on handset unit demand.

Okay. It's glad to hear from you, John. Yes. So let's talk about the first question, which is NAND content in mobile phones because of 5G. Interestingly, we're actually seeing that 5G will actually drive more storage need in the phones than less. And the reason is I think the speed at which now you can connect with the network is much higher, so people really want to download a lot of content in a short term. That's one of the big reasons. Like -- and also the content itself, if you look at a lot of the content places where you get content from, like Netflix and Prime and so on, are actually moved to this model of a lot of 4K video now, and they'll let you download entire seasons of episodes, and people just want to get it all down in the phone. So they have it even when they're connected or not connected. That's one reason. The other thing we actually found is actually very interesting. Most people, when they buy a new phone now, we find that most of them will start by restoring the backup they've had with their old phone. And so almost everybody, when they upgrade to a new phone, starts to back up the old phone, that takes up quite a bit of the NAND that they had before, and now they're forced to buy a new one. And the third one we found is the app penetration. If you look at things like WhatsApp and Instagram and in different parts of the world, different chatting apps, there is so much video and image content in those apps that they're very quickly eating up the storage space that you have. So those are actually some of the trends that are actually making it go faster. And also the latency of -- at which you can actually have that performance is kind of very, very important. Going up to the network and coming back, really, you don't have that immediate experience that you had on high-resolution pictures. And the last one we found is as your latency gets low and people start playing multiplayer games, these game applications are actually very, very large apps and they take a lot of storage because they have a lot of scene content that actually goes into the NAND when the app launches. So we're actually seeing it move up as 5G launches. To your second question on 5G rollout, we actually are seeing the 5G rollout continue as planned in China and also in U.S. We haven't seen that slow down at all. Europe may be slowing down a little bit because of COVID and the money it takes to build into the infrastructure. But on the whole, we are fairly optimistic that, in the next year, it'll just continue to be -- move as fast as we expected because 5G does provide, as I explained in the presentation, a lot of clear end-user benefits, and that's what ultimately drives the penetration of these phones.

Raj, maybe another way to ask the NAND question, how do you see handset OEMs dealing with the increasing BOM cost pressure of 5G phones, just given the margin world that they live in?

Yes. So one thing to realize, I've been in the industry for a while, right? I've been there at 3G launch, 4G launch, now this is my 5 -- third generation -- 5G launch. What typically happens when new technologies launch is people like Qualcomm and MediaTek and who actually make these processors, initially launch with a separate application processor and a modem, and so they have a separate memory system on the application processor, separate memory system on the modem, that increases the total BOM cost quite a bit. And also, in some cases, the RF is actually duplicated because it was not all integrated to -- because of time to market. Now as we see -- so the total BOM cost is actually very high because of that reason. Now as we see single-chip solutions come out, which we've already seen them come out, very quickly, the BOM cost gets a lot more manageable. And we see that people are actually adding memory and NAND because that actually matters to the end-user experience.

Our next question comes from C.J. Muse with Evercore.

I guess first question, and perhaps this is a few years into 5G, but given how real estate is increasing in importance with the move to 5G, are you seeing any sort of changes in packaging requirements in this transition?

Yes. So I think a couple of things. That's why I believe that high-value solutions and MCPs are going to be really important, and that's why we launched our first uMCP5. When you have discrete solutions, you basically need 2 different packages, one for NAND and one for DRAM, and also many times, DRAM is actually put on top of the apps processor, so it increases the thickness, thereby taking away some space from things like battery. So we actually see that as 5G rolls out more and more, what we call MCPs, or multichip packages, which actually integrate DRAM and NAND, and what we call high-value solutions are going to be more and more prevalent, and that's actually something we are quite excited about.

Great. And as a quick follow-up, on the last earnings call, you guys talked about migrating certain wafer starts to servers away from mobility. But here we are into some sort of seasonal ramp for smartphones into the back half of the year, you are integrating into MCP package. How in this uncertain environment do you kind of manage production starts? Would love to kind of get your thoughts on that.

Yes. It's a good question. So as we mentioned last time, we saw -- because of COVID, we saw an effect in mobile phone demand and also some production in China. And then we reacted by moving some of the bids to other areas where we saw stronger demand because of work-from-home economy and all that, which actually was a good move that we did. What we're seeing now is we actually see China recovered now. And of course, COVID now is in the rest of the parts of the world, and we are watching that closely. We talk to our customers very regularly. Now actually, the cadence of our communication to customers have actually increased quite a bit. I'm on calls every month with all our key customers. And as we see things recover, we are quickly making the adjustments to make sure that we have the right products for them.

Our next question comes from Blayne Curtis with Barclays.

Maybe to follow-up on John's trade question. I'm just kind of curious, obviously, it's been a moving target to try to quantify all the effects. But the fact that Huawei can't get chips and share moves to Android, I'm kind of just curious, your perspective as to your positioning in non-Huawei, Android and whether it will benefit you. And then I'm kind of also curious, just on DRAM content for phone. I think the incremental unit is going to be kind of more mid-range phones. I think 5G has been more of a -- you've seen this total handset number go down and -- but the mix go up dramatically. I'm kind of just curious of the content uplift in more of a mid-range phone versus high-end for DRAM.

Okay. Sure. Yes, I mean, I think one thing that, as you saw from my presentation, that we have done over the last couple of years is to really focus a lot on our product portfolio. And if you look at our product portfolio now, it's actually pretty wide and pretty broad, both in discrete DRAM and MCPs, which actually helped us have a much better customer penetration across all the customers, pretty much most -- all of them use our products now in different segments of the market. So we believe that, that brings the business a lot more stability to any one customer fluctuations up and down. And we work very closely with them. We work pretty closely with Huawei. It's really hard to predict the effect of the ban, but we do feel like our products are designed in across a large number of customers in Android. So we feel like we have a good penetration to weather any changes in the customer market share and so on, both in the short term and long term. As far as your comment on mid- to high-tier phones content, I mean we've done some work recently to really see -- like we actually took one of the higher-end phones of the flagship phones and we try to see what exactly is the memory footprint and where is it getting used. While this is early data, we have a lot of expertise in this kind of stuff because I came from Qualcomm and I understand apps process pretty well, and I brought in a team of people from Qualcomm to work here. And we actually saw that the minimum footprint needed today in a 5G phone between the Android OS and the system is 3.5 gigabytes just to start. And then by the time we load 6, 7, 8 apps, it's actually over 4 already. So that's why there really are no 5G phones in the market today that use less than 6 gigabytes of DRAM. Now that will change in time. We believe people will try to optimize that. Android will try to optimize, maybe camera resolutions will adjust. But I still believe that 4 gig is probably where even the mid-tier phones will need to be to really provide that user experience in 5G because what happens, typically, I've seen in the phone market is when people buy their first 5G phone and they get this great experience of apps and everything that they get used to it, they try to want the similar kind of experience as you go lower and lower into the tiers. So it's one of those things that kind of drives content because you don't want to compromise the user experience that you are used to.

Our next question comes from Harlan Sur with JPMorgan.

Very informative. On the higher value-added mix within the team's NAND portfolio, it was 50% of the mix in fiscal '19. It hit 70% of the mix last quarter. You guys have a target to get to 80% next fiscal year. I believe your MCP business has been driving record revenues, and I think your managed NAND business has also been doing very well. So can you just help us understand, is the biggest part of the higher value-added mix your mobile products? Or is it more the SSD products? And within -- with the big uplift in content for 5G, would you expect the mobile mix to continue to go faster relative to the other higher value-added solutions?

No, I mean for us, as you pointed out, we've done -- because of the things I mentioned in terms of understanding the system, building the right product portfolio of managed NAND and also MCPs to address the different customer mix, we've been able to grow both managed NAND and MCPs at a record clip. But our SSD business is doing really well, and we believe that both SSD and mobile are important to us, and both of them are growing nicely and both of them will grow. So we look at them both very important in growing our total high-value solutions in our share.

Again, in the near term, Harlan, the -- clearly, like just more recently, because we have been not as broad in our portfolio of SSD NVMes, MCPs clearly has done better more recently. But SSDs as we grew that in SATA, certainly was growing at a great rate. Now that we do have a more robust portfolio of NVMe, SSD will be back on track and growing nicely. So as Raj alluded, we think both of them will have good growth rates.

Okay. No. I appreciate the added insights there. Maybe more near term, I appreciate the update on the demand side, but I believe many of your MCP and/or managed NAND solutions are assembled in your Malaysia factories. I know the team was having some challenges given the COVID-19-related movement controls in Malaysia. Can you guys just give us an update on your operations there?

David? I can take a shot here, David.

Yes. No. Go ahead. Go ahead. Raj. You take it.

Yes. So our MCPs are actually done at multiple locations, not just Malaysia. I mean, we make them in Malaysia, in Singapore and Taiwan. So we kind of have a broad footprint and where we make different things. So I think we have been fairly fortunate that we've been able to keep them going.

The other thing is, at this point now, Harlan, we're actually back up and running at essentially full capacity in Malaysia now. So at this point, we're in pretty good shape across the supply chain. Now of course, anything can happen. But at least as we stand today, we're in good shape.

Our next question comes from Timothy Arcuri with UBS.

Actually, I had 2 first. I don't know whether, Raj, you want to take this, or Dave. But I'm sure you saw overnight the talk about the fact that the U.S. government, they might extend the use of U.S. equipment to produce memory to then sell to Huawei. So I think the comment was specifically more about the Koreans, but obviously, it would impact you as well. So a, I just wanted to get your feeling on that. I mean, I know that, over time, you'd be able to fill that hole, but I just wanted you to address that. And I guess, secondly, Raj, if you could talk about is there any incremental content difference as we roll out millimeter wave versus sub-6? I know it's not necessarily due to the connectivity standard. But are there other things about millimeter wave that will drive more memory content versus the sub-6?

Yes. So I mean, I haven't seen the U.S. government thing that you talked about, so I can't really comment on it. But about -- but the second one on millimeter wave versus sub-6, I think the millimeter wave clearly provides much higher bandwidth than sub-6. And sub-6, of course, helps you more with mobility. So we believe that the content on millimeter wave will be higher for -- will need higher content for things like video and gaming and so on. And so I think that's probably going to be one of the drivers also.

Dave, can you address the equipment issue, the China issue with Huawei? Can you address?

Nothing that came out last night I can comment on. I mean, up until -- or anything that I've seen, at least in the -- directly wouldn't impact us. Of course, challenges around businesses of our customers is always something we have to manage around. But as we kind of see it, we've been able to do a fair job with that.

Our next question comes from Ambrish Srivastava with BMO.

Raj, I was wondering if you could comment on the difference in bit growth that you saw in 4G versus the numbers that you gave us. And then also, if you could please comment on what does it do for demand on the infrastructure side? And then I had a follow-up for Dave as well, please.

Difference in bit growth we saw at 4G versus 5G. Is that your comment -- your question?

Yes, yes, yes.

I mean, I think, 5G, clearly, we believe, will drive more content growth because of the -- as I mentioned, the things like 4K video was possible in 4G, but in 5G, now people are going to 8K video, which by itself makes a pretty big difference. And also the number of cameras and the camera resolutions during the 4K time has been 15, 20, 30 and 24 megapixel. But if you look at now, there's 100-megapixel phones, and I've even seen customers talk about 150-megapixel phones coming out now. And again, the reason for the higher megapixel, a lot of people ask me this, is because you can now do real-time zoom, right? That's actually a pretty big deal because phones always struggled with how to make zoom and not lose quality. And the other thing people are finding in these cameras is multiple cameras, some wide, some zoom, some in the front, some in the back. So these are innovations that have timed with the 5G launch, and that, coupled with the fact that you have a very high bandwidth connection to actually get the content down, both those are driving much higher 5G -- much higher content growth in 5G era than the 4G era. Now I mean -- but in the premium tier side, we've already seen quite a bit of growth, right? I mean, more people are launching 8 gig, 12 gig, even 16 gig phones. So there's probably going to be a little bit of saturation at the very high-end just in terms of how much more you can put in there. But clearly, on the mid and high tier, where you kind of see a lot more units in 5G, I believe that the content will be higher in 5G than in 4G era.

Yes. I was just trying to compare the bit growth you gave for DRAM and for flash, 15% and 30%. And I was just wondering, what was the bit growth when you had the ramp in -- going from 3 to 4. That was the question.

Yes. I don't remember that. Yes. I don't have it handy. But I mean, I think I also will say this, so we're actually a little bit conservative, I would say, in what we are estimating moving forward. And it's actually just prudent to be conservative because we use those numbers to manage our investments and CapEx and supply, and we always want to make sure that that's done carefully so we don't [ out-index ] in CapEx. So in that sense, it's probably a little conservative number. But we'll probably have to get back to you on what it was in 4G. I don't remember that.

Yes, Ambrish, it's fair to say that the content growth has been -- has come down. It used to be higher because the innovation in the smartphones were a lot more early on, and they were starting from a very, very low base in all the tiers and so you saw a pretty strong growth. And as we look ahead as Raj said that on the very premium end where you're getting like 16 gigabytes of content already, it's hard to see that continuing to grow at the rate at which historically it has grown.

Okay. And Raj, I was wondering if you had any insights on the infrastructure side for 5G. What are the deltas and the differences that enable you to grow content on that side?

I don't know that part very well. So maybe we had to get someone from our infrastructure group to talk about that. Maybe, Mike Lemus knows them.

Yes, we'll have to get back to you on that one. We are not addressing the broader...

Okay. And then my -- and maybe I'll squeeze in one for Dave. Dave, last -- through several quarters, you have addressed this that there has been some sort of overbuying going on in China. And I think a couple of quarters ago, you said that you didn't expect it to come to a stop. And I was wondering, has that abated but -- or given all the additional ratcheting of trade tariffs, has that continued on?

Yes. I mean, what we saw in the second half of calendar '19 was some inventory build in China based on concerns around trade and I think most OEMs trying to derisk their supply chain. As we came into this calendar year, we had a bit lower expectation around DRAM growth rate, and part of that was an assumption that they would make some adjustment to their inventory levels over the course of this calendar year, and that would affect the growth rates a bit. Of course, early on in the calendar year then, the COVID situation arose, and I think, for the most part, most of those OEMs in China are probably just as concerned around supply chain, also because of trade because that is still swirling around. But in addition, concerns around the coronavirus and shutdowns disrupting their supply chain. So our view is that they kind -- they have certainly not adjusted their inventories down. And inventories in China remain at somewhat of an elevated level. The magnitude of which is different across markets and across specific OEMs.

Our next question comes from Karl Ackerman with Cowen.

Maybe I guess for Raj or Dave, you have traditionally been underexposed in mobile because you didn't have the right cost structure. Now it seems you are in the lead, particularly in DRAM with 1z. But in mobile, when should we see replacement gate being introduced? And when do you expect to reach revenue crossover of uMCP and UFS devices versus eMCP and eMMC?

Yes. So I'll take one shot at the initial part of it, maybe Dave can add to it. But I mean, I think that the reason -- we have been underexposed in mobile, and it is not just the cost structure, I think we didn't have the right product portfolio also. Because, really, to be successful in mobile, you have to have a fairly broad mix of MCPs because mobile as a whole market is very different between premium tier and mid-tier and high tier and low tier. And that was one of the changes that I was able to bring when I came from having spent quite about 15 years in mobile is to bring that market knowledge and the relationship with the customers to see what products we really needed to make to be able to address the largest parts of the market and also growing the MCP space because that's clearly where the higher value is and then get the right DRAM into that space. And as you mentioned, getting 1z really helped because it brings lower power and we were able to get cost structure also much better. So those are -- and the other one I'll mention is that we spend a lot of time now in the last couple of years, as I said in my presentation, spending time with the customers to understand what's the problem they're trying to solve and what kind of SKUs of products help them and how the end-user phone experience changes based on the products we put in, and that also helped us grow the share quite a bit. As for replacement gate versus floating gate, we are actually just sampling our replacement gate products, and we believe that launch will start next year, and it'll just continue through that. Again, I can't comment exactly when the revenue mix will change. But from next year, we'll be shipping our replacement gate. Your question on eMCP versus uMCP, I don't remember exactly when the revenue crossover will happen, but we do see the market moving more and more aggressively to uMCP, and we can actually see now that the U.S. uMCP market coming into even the -- usually, they're predominantly used to be only in the premium tier and high tier. Now you can actually see it in the mid-tier phones also. So that's one of the reasons we're also growing share because we have a pretty strong UFS portfolio now.

Our next question comes from Srini Pajjuri with SMBC Nikko.

Raj, just to follow up to the previous question, I'm trying to better understand how the capacity allocation works between mobile and some of the other segments, and where in the process do you have to decide whether a particular wafer is going to be allocated to mobile, et cetera? And also along the same lines, when you design these products for mobile market, for example, an eMCP, are they designed for a particular customer? I mean, are they standard products? Or are they more of custom products? Are they designed for particular models? So if you can just go through that, that will be helpful.

Yes, absolutely. Well, let me talk about the second part of your question, how do we define products. So the way we define products is actually very much market- and customer-based. So what I mean by that is we study trends in the market, like for example, we would say, okay, with 5G launching, when we'll -- what will be in the 5G flagship phone versus what will be in the high-end phone, what will be the mid-tier phone. Then we talk to people like Qualcomm, my old team, and figure out what are the products that they can support, and to MediaTek and HiSilicon and Samsung and whoever else sells processors. And then we define products that actually will interface well with those guys. And then we qualify those products with them first so that they're ready. Then what we do is we also make, like for example, in the uMCP case, we'll make 128 plus 32, which is basically 8 gig plus and so on, our 6 gig. And we'll make multiple products with different SKUs. And then we work with each of the customers to see which part of the market are they addressing, like for example, one of our customers may want to go more into India and they may want 64 plus 32. Another customer may want to go more into China, and they may want 128 plus 48. So we actually have different SKUs based on the different customers, but these are not custom, by the way. They can easily be more -- if one customer find that they're not doing so well here, they want another part, we have that SKU also. So that's kind of the point I was making earlier. The reason for our success has been that flexibility and understanding of the market and defining products that match the requirements of the market and match the requirements of the customers based on which part of the market they want to go into at any point of time. That's kind of how we define them. And again, understanding the system requirements is very key. So we really need to know what use cases will drive what products, which has been the new thing that we brought on in the last couple of years. For your last comment on how do we do the allocation, we have a small team of people, Sumit, me, and there's a couple of other GMs, and we all get together and we have a plan for the year, but we review that very, very closely and try to make changes as we see the market going on. And...

I would just...

Yes. Go ahead.

Sorry, Raj, I was just going to jump in on this one a little bit, too. So one way to think about it is Raj and -- meets pretty regularly with his customers, probably more frequently now than he's ever met with them, and uses that, coupled with his own sense of the business. All the other GMs do the same. That builds up what their expectations are going to be around their particular needs over the course of the next, say, 12 months. We meet frequently with operations then, and we've stepped that up probably twice a month now to take that information and start to build a plan around how we're going to manage the mix. And we have to do that, obviously, because we need to start wafers now for Raj to get product 4 months from now. So we have to constantly try to anticipate how things are going. And in addition, we then aggregate all that up and look at what the overall bits would suggest we need to build and plan our CapEx around that as well. So we're now, every other week, kind of looking at capital spend in relationship to what we think is the bit demand over the next several years and adjust our CapEx assumptions based on that. And of course, suffice it to say, if we went back to, call it, February or something, I had a pretty good line of sight that our CapEx next year would be growing at a reasonable rate relative to where we think we will land for fiscal '20. And now it's pretty clear that's less likely. And so CapEx, clearly, now the assumption is to spend less than what we were originally thinking for fiscal '21. And now it's just a matter of exactly what the magnitude of is -- magnitude of that will be. And so over the next couple of months, we'll use all the data that we get from Raj and his compatriots, the other general managers, to make some plans around next year's capital spend.

Got it. Just a follow-up to that. The CapEx spend for next year, the change in view, is that primarily driven by the weakness in mobile? Or is there something else, any efficiency gains or any other markets where you're seeing weakness or any competitive factors playing a role there?

Yes. It's a combination of an aggregate sense of where we think -- clearly, there will be less bit demand this year than we came into the year thinking. And so that adjusts the inventory levels that we had originally planned out for this year and next year or so. And as you kind of saw, our inventories were up in the kind of 100 -- mid to 130s, and potentially we'll be at that or somewhere around that for this quarter. So we want to bring that down. So if we want to bring that down and we have a different assumption around bit demand for this year, we have to adjust the CapEx accordingly.

And our final question comes from Tristan Gerra with Baird.

You talked a little bit about the next-generation DRAM architecture that 5G mobile is going to need. Could you talk a little bit more long term about how 5G is going to reshape memory architecture and potentially also the positioning for 3D XPoint longer-term in phones.

Yes. So as I mentioned, I think, in 5G, one thing that's happening is we clearly see the need for much higher bandwidth from the DRAM and also need for much lower power point, power consumption by the DRAM itself. So we continue to work on both those fronts. For example, after LP5, there'll be something, a new standard called LP5x. So we just continue to drive that standards to make sure that we can get that. And the other thing I think that's going to happen in 5G is that as 5G moves from the premium tier to high tier and mid-tier, the single chips will only have 2 channels of -- to connect to the memory. Typically, in the premium tier, they have 4 channels. And as we come down to 2 channels, but still the use cases still need as much bandwidth, so what happens is you'll just need to run the DRAM faster and drive more throughput over just 2 channels where you had the luxury of doing at 4 in the premium tier, and that needs innovation and that needs higher clock speed. That's why we're excited about LP5 because we're actually the first ones to run it full 6.4 gigabits per second. So we actually expect that trend to continue in 5G, higher speed, more bandwidth over lesser number of channels and lower power. Your comment on 3D XPoint, our initial focus on 3D XPoint is in the data center, and we're going to get that to production and get it to market. And once we do that, we will see if it makes sense to have some advantage in mobile apps. So we're constantly evaluating that. Right now, our focus is on data center.

Ladies and gentlemen, this concludes today's conference call. Thank you for participating. You may now disconnect.