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

Welcome to the second day of JPMorgan's 53rd Annual Technology Media and Communications Conference. My name is Harlan L. Sur, I am the Semiconductor and Semiconductor Capital Equipment Analyst for the firm. Very pleased to have the team from Micron Technology here with us. We've got Manish Bhatia, Executive Vice President of Global Operations and Samir Patodia, Senior Director of Investor Relations here with us as well. Samir will kick it off with safe harbor, and then we'll go ahead and kick off the Q&A. Gentlemen, thank you for joining us this morning.

Yes. Thanks, Harlan. So we'll start with the safe harbor. We'll be making some forward-looking statements today. Those statements have risks and uncertainties associated with them. So we refer you to the risk factors disclosed in our public filings.

Great. Again, thank you for joining us this morning. I'll kick it off. You're nearing the end -- the team is nearing the end of our May quarter. Coming into the quarter, you had anticipated quarter-on-quarter bit shipment growth in both DRAM and NAND as excess inventories in the more consumer-oriented smartphone, PC markets had been cleared, calling for record revenues. On the gross margin front, the team guided gross margins to decline slightly, 140 basis points to about 36.5% on a higher mix of consumer-oriented volumes. Obviously, NAND underutilizations and lower calendar Q1 pricing trends, which I think have recently turned, obviously offset by strong execution on the HBM front. Since then, DRAM pricing has been strong, both DDR4, DDR5, contract pricing and spot pricing have both been strong. NAND pricing declines have started to stabilize. And so good trends overall over your May quarter. And so much so that at the end of March, the Micron team did issue a letter to customers indicating your intention is to increase pricing as the team was seeing an increase in unforecasted demand, right, across various of your business segments. So relative to your guidance in March, I mean, it certainly seems like pricing has come in better. But wanted to get your views on how the May quarter has played out so far.

All right. Do I need to do? Great. Thanks, Harlan, and thank you for hosting us here at the conference and for all those kind of market trend highlights since our last call. I'd say that you're right on the trends. I think we've seen healthy demand for both DRAM and NAND. And a lot of -- the dynamics have played out much as we had discussed them not just in the last quarter but even going back further to some of the trends in terms of strength in data center as well as what we had said before would be a springtime resumption in strengthening the consumer markets due to inventories having to -- having come down over the sort of calendar Q1. We're not updating guidance at this time. But I will say that the quarter is progressing well and we are driving price inflection as we had said we would. And so that's playing out -- that's evolving well. And as we see the trends in the quarter continuing, we are going to be continuing to make our plans for the second half of the year.

On the pricing inflection trends, is that more a statement around DRAM because we have seen a strong upward trajectory of DRAM but does that include NAND as well?

Yes. So we had talked about actions that we had taken back during the December call, actions we were taking from that point on, and we've continued with some of those supply actions and others in the industry have done similar. And so that has helped with the trajectory of NAND pricing.

Perfect. And then I think on the earnings call, you had given us a little bit of a sneak peek into fiscal Q4, right? And off of the -- and I think the view there was that in fiscal Q4, your August quarter, you would see slight improvements in gross margins. With the better pricing trends here in the May quarter, pricing inflection certainly seems like both DRAM and NAND pricing remain on sort of this positive sort of trajectory, at least as we look -- at least as we look kind of near the midterm on a forward basis. Again, you had previously anticipated fiscal Q4 gross margins up slightly. But it would seem that the gross margin trajectory in Q4 could be -- could improve at maybe a slightly better trajectory than what you had laid out in March, I'm not sure if that's a fair assessment or not?

We're not providing forward-looking guidance in terms of margin outlook. But we'll say that part of our comments around F Q3 as well as F Q4 around the mix of the business and that pickup in consumer-oriented business that we expected for F Q3 and that increase in mix does tend to be lower on the margin side versus, say, some of our higher-end data center products. And so that increase in mix in consumer was kind of the key driver for F Q3, and we'll kind of provide more commentary relative to forward-looking margin structure overall as well as the mix that we expect as we go through the next fiscal quarter and into the second half.

Perfect. We'll have a lot more discussions around some of your end market segments but obviously, a lot of your here -- a lot of us here in the room are extremely focused on the dynamic of accelerated compute and AI. The Micron team we know has been designed into NVIDIA's next-generation GB 300, which will be starting to ramp in second half of the year. Concomitantly with that, right, the Micron team did start production of your HBM3E 12 high just recently. Maybe you could just give us an update on the ramp, the yield? Is it tracking towards expectations?

Yes. No, we're very pleased with where we are with HBM. We know our 8-high has been in production for over a year now. It's formed a really solid foundation for us, and we started production on 12-high earlier this year. And the progress there has been good. In fact, we are seeing the -- we had said before that we expect that to -- the 12-high to be the majority of the demand in the second half. I think we can now say that we see the crossover between 12-high and 8-high in terms of volume shipments happening in our fiscal Q4. So that's -- that trajectory is ramping well. And in terms of yield progress, the learning rate on yield for 12 high is really faster than we had on 8-high. So along with that ramp that -- the ramp up in 12-high as it crosses over Q4, I'd say that we're expecting that we'll see the 12-high yields to be able to kind of reach that knee in the maturity by the end of Q3. So kind of going hand-in-hand with the ramp. So we're very pleased with where we are in terms of the progress on the 12-high production ramp, in terms of the process stability as well as the yield improvement.

I'm going to go back to kind of the near to midterm environment because in addition to the strong pricing inflection that we've seen so far here up through your May quarter, much has changed, especially on the tariff and trade and geopolitical front, right? And as you mentioned, AI and accelerated compute demand trends remain strong, strong cloud, hyperscaler, CapEx spending trends, even in the non-AI markets, right, smartphone, PC, general purpose server memory demand trends have been relatively good. And as you guys had predicted 2 quarters ago, right, demand combined with the discipline of the industry has driven excess inventories down, and that's part of the strength that you guys are seeing. But I think with respect to tariffs and trade, the big question is how much of the strength in near-term demand and therefore, strong DRAM pricing increases and stabilization in NAND ASPs are due to tariff pull-ins, right? I know one of your large memory competitors recently called out preemptive tariff-related demand pulled into the June quarter. How was the Micron team interpreting the current positive demand and pricing environment, cyclical improvements that you called out back at earnings versus tariff-related pull-ins?

Yes. Yes. So I think the key point there with regard to the demand trends, I mentioned before that we're seeing healthy demand, right? And it's playing out much as we talked about with the data center continuing to be strong. And with consumer's -- consumer end markets, having worked through inventory and we're seeing that pick up in this May quarter. And so there could be, for certain customers, some pull-ins, but I would say that the bigger trend is that are the ones that we had laid out and forecasted back in our December call, playing out largely as we expected.

The other side of the potential tariff-related dynamics, again, a view by a few of your competitors during the earnings call that -- was that these pull-ins are potentially pulling from demand in the second half of this calendar year and therefore, could potentially drive a weaker bid shipment profile. Additionally, end market trends in second half, PC, smartphone, general purpose server forecasts have been -- are being trimmed, right, by the big Wall Street firms including our firm. We recently cut our smartphone forecast, our PC forecast and our server forecast, a little bit less seasonal, right, in the second half of this year because of tariff and trade-related uncertainty. If I recall in the last tariff and trade war back in 2018, 2019, the semi industry and the memory industry did enter into a downturn, coincidentally or not, right after the final large round of tariffs that were put in place of that year. Are you seeing any indications by your customers that they're starting to plan for a sub seasonal, maybe slightly weaker environment in the second half of this year?

I think we've -- you've talked about some of the macro and tariff uncertainty and trends. And obviously, everyone is having to react to that. But I think the commentary from many of our customers, even public commentary, certainly the cloud service providers commentary over the recent rise is for them to continue with their capital spending plans to be able to build out in the data center and continue to invest in their AI solutions businesses, right? And I'd say that's one area that's reassuring for us with regard to the future data center because so much of our portfolio and so much of our focus is on that area and driving more strength there. On the consumer side, there are just -- there are a number of different product launches that are still happening. I think that we're seeing customers who are really -- have a new wave of offerings to be able to have, I think, smartphones and PCs that have native AI capabilities designed in versus maybe over the last year, they were there were platforms that weren't designed with AI agents in mind, and they were trying to be able to add some capabilities but frankly, the reset in terms of the hardware platforms, the amount of memory, the capabilities and the processors, all of that is still, I think, ahead. So I think that there are still strong demand trends ahead. I think in terms of tariffs, obviously, there's been a lot of different signals, and it's been difficult for every everyone, all of you, all of us, everyone trying to parse waking up every day reading the news. But I think encouraged by the direction of travel over the last couple of weeks in terms of de-escalation on the tariff front. As far as it impacts the macroeconomy broadly, of course, as you -- well, I should say, of course, the semiconductor-specific products, including DRAM products and NAND products were exempted from the reciprocal and baseline tariffs before, and those are part of a Section 232 review. So we'll have to see how that plays out.

Post- tariff announcements, there was some news around the Micron team reaching out to their U.S. customers on the potential for passing through tariff-related price increases. But on April 12, the President did exempt SSDs, memory modules, chips from tariffs. So I assume that the team has not had to increase prices to your U.S.-based customers. Number one, is that an accurate statement? And then the second question is, what percentage of your overall revenues is for shipments of SSDs and module form factors across the U.S. border?

Yes. So we have commented a little bit on the overall pricing trends. You're right that -- and I mentioned just a minute ago that the products have been exempted. So there haven't been tariffs levied at least on our DRAM or our NAND products that are coming into the U.S. So we're just keenly watching this 232 review to understand what the implications will be. In terms of percent of products, that really varies quarter-to-quarter depending on where our customers' supply chains are located and how much we provide to them in their -- in which region where they're going to be either building out their data centers or where they're going to be assembling their final products. And we're focused on understanding the 232 review and also trying -- we actually provided some comments to the Commerce Department for that review. Those will be -- eventually be public. And we're trying to make sure that the policy is constructive relative to the administration's priority to be able to build the U.S. semiconductor manufacturing. And so we've been -- our comments will provide kind of a road map of what we think are key ways that this can be enabled or accelerated and not somehow slowing down -- that semiconductor-specific tariffs don't somehow slowdown that initiative.

Yes. And maybe as a follow-up to that, the U.S. administration's goal has been to fast track a return of manufacturing back into the U.S., right? And the good news is that the Micron team is well into the $15 billion expansion of your Boise, Idaho fab, right? And can you give us an update here? The original plan was to bring Boise online in 2027. Where you -- where are you in this build-out? Are you potentially accelerating maybe your plans here? And have you seen a step-up in potential customer commitments, right, to Micron longer term because of your U.S. bit supply capability?

Yes. So we are in construction in Idaho. We have 3 projects really that we've announced in the U.S. One is in Idaho, co-located with our research and development facility there. So we'll get the benefits of time to market of our newest technologies as well as some synergies between R&D equipment and manufacturing equipment and R&D engineers and manufacturing engineers. In terms of -- and then we have a project -- large-scale project in New York that's currently in permitting. And then we have another project that we are still discussing with the Commerce Department around modernizing our facility in Virginia, which has been for supporting automotive and aerospace and kind of long life cycle opportunities there. We have had, frankly, really good response from multiple different customers who are interested in understanding how they can leverage our -- or be part of our U.S. manufacturing supply. And there are multiple reasons for that. I think one of them is the fact that the energy in the U.S. has the potential to be able to have more green energy here. So that's been one of the reasons that people are interested. And then there's others, I think, just because of the overall push to be able to have more U.S.-based supply chain even for themselves. Many of our customers, as you know, have announced larger U.S. supply chain footprints in the future. And so I think they want to kind of see how they can -- how we can play into those plans.

Before I'm going to start talking about products and markets and overall end market dynamics. But before that, I wanted to open it up to the audience to see if there are any questions out there. If you do have a question, please raise your hand, and we'll try to get a mic over to you as quickly as possible. Any questions from the audience? Okay. So let's talk about AI and accelerated compute. Obviously, a big driver of the business, the mix on the accelerated compute and data center side for Micron has richen significantly over the past few years. In addition to tariff and trade-related dynamics, there's been recent concerns around the potential for AI compute demand normalization, maybe some compute digestion. Although we have seen -- we haven't seen any pullbacks from cloud and hyperscalers on their cloud CapEx spending intentions. And so have you seen any signs, indications of a pullback in accelerated compute demand? In other words, has your HBM TAM outlook, right, greater than $35 billion for this year changed at all?

Yes. So we provided that HBM TAM update during our last earnings call, and we typically provide those on those calls. I'd say there's a lot of moving parts there with regard to changes even as late as yesterday on the AI diffusion and how that AI diffusion pullback is going to impact demand. There's a lot of different moving parts. So we are not going to provide an update at this time. But I will just come back to the comment you made -- I made earlier that we are seeing robust demand for our HBM product, and we're seeing robust demand for our high-end or high-capacity DRAM modules as well as expanding demand for our low-power solutions for data center, and we're encouraged by the capital spending plans of virtually every cloud service provider as well as new entrants into cloud service providers that are trying to build out, I think that's also good news for creating multiple vectors of demand and even multiple different architectures, right? I think the fact that there are going to be different workloads and different applications focus areas for different cloud service providers as they're coming up, that just gives us an opportunity to go in and leverage the strong capabilities we have across our portfolio. We believe we have the best -- customers consistently tell us we have the best HBM product, 30% lower power than our competitors for actually even better performance than them. And we have -- we're the only ones really shipping low-power DRAM into data center applications right now as well. So across the board, I think that heterogeneous set of workloads for data center across expanding number of cloud customers, that just plays to our strength.

If I rewind back 1 year ago, you gave us the update on your HBM3E 12-high ramp. If I rewind back a year ago during your June earnings call, you did say at that time that you were sold out on your HBM supply through calendar '25 and that the majority of that committed supply was locked in from a pricing perspective. One year later, where are you on negotiations for calendar '26 HBM supply and pricing discussions? I mean is the team's supply outlook for next year maybe almost fully committed?

Yes. So we're in discussions right now on 2026, which, again, is earlier than we are for other products and traditionally. I mean HBM certainly commands longer lead time in terms of discussion given the criticality it plays into our customers' road maps. And I think the discussions with our customers center around the mix of HBM3E 12-high, when their platforms are going to be ready for HBM4, how our HBM4 will ramp. I mean, we think we're going to have, again, a leadership product in HBM4, we expect to and be able to ramp at the very leading edge of our customers' platform ramps. And so those discussions are ongoing but we feel very good about where they are as well as really the feedback from our customers has been terrific in terms of how we've been able to ramp our HBM3E. Really, this time last year was essentially the very beginning. That's why we started production last March, having not had a product in HBM3 had a very little -- coming from very little learning base, little scale. Where we are today, our customers have been very, very pleased. And that give them the confidence to be able to count on us for their future ramps, and we feel very good about our future...

And it's been a great trajectory, right? You guys did over $1 billion in HBM revenues last quarter and on track to exit this year based on our estimates at double that rate, right, at roughly about a $2 billion per quarter annualized run rate. So very, very strong ramp and execution by the team. Let's...

We actually said that both HBM as well as those other categories of high-capacity DIMMs and LP. When you think about each of these, they'll each be multibillion-dollar businesses for us in fiscal '26.

Yes. Let's switch over to data center SSD. We're seeing, as you mentioned, as you mentioned, numerous AI GPU, but also a lot of XPU programs coming into place, right? All these ASIC programs that are going to start to ramp into what we think is strong second half of the year, right, which would be good for DRAM memory product but should be good for your storage solutions as well. And I always tend to remind people that your data center SSD share has gone from 5% to 7% in 2022, 10% to 12% in 2023, and you exited last year at 15%, right? I think either #2 or #3 global market share in enterprise SSD. So strong share gain performance over the past 3 years. You've got your next-generation 9950 PCIe Gen 5 SSDs out there. What are we going to see your next-gen G9-based data center SSDs come to the market? And more importantly, does the team still expect continued share gains in enterprise SSD going forward?

Yes. So you're right. Thank you for that, Harlan. And you always call out our data center SSD execution. I remember even here last year, and we've been very focused on that. We think that the -- over the last couple of years, our vertically integrated platforms, having our own self-designed ASICs, so that to go along with the leadership we have in memory and the capabilities that we've developed in terms of integrating that together and manufacturing has really played well. And I think our customers also really appreciate the single point to be able to build out their data centers, right? So for example, being one of the first to be qualified SSDs on the Grace, Blackwell platform, right? So I think as we see the overall data centers getting built out, the story has been about DRAM but you're right to point out that NAND definitely has a role to play and we will be expanding. And we are looking to continue to increase from that record level of demand and share that we saw at the end of calendar year '24.

On the last earnings call, you had -- the team had anticipated data center SSD bit starting to reaccelerate over the coming months. With a bit more visibility versus the March earnings call, would you expect maybe that data center SSD re-acceleration to start maybe in the August quarter, your fiscal Q4?

I think we'll probably give some more color on that later. But I think certainly, the actions that we and the rest of the industry took in NAND late last year and into the first half of this year, that should help with overall balance, and that should help with inventory clearing across the channel, and that will help with the healthy -- provide a healthier environment for all NAND products, including data center SSDs.

Maybe you can just give us an update. The team is doing a lot of technology upgrades. You gave us the update on the HBM3E 12-high ramp. Your team is starting to ramp your next-gen 1-gamma DRAM EUV-based technology. As I just mentioned, you're starting to ramp your Gen 9 NAND technology. Can you just give us an update on these new technology ramps, yield, cost downs and just more importantly, just overall ramp execution?

Yes. I think both are going very, very well. I think our Gen 9, we announced earlier and have progressed well on that one. We're managing the ramp of that Gen 9 in line with demand, right? So we're trying to make sure that we're bringing capacity online responsibly. But at the same time, feeling really good about where it positions us relative to competition and the capabilities in terms of the NAND device performance, we think, are best in the industry for that Gen 9. So that will help us, obviously, in high-performance applications like data center SSDs. 1-gamma, we're very proud of. We just announced that qualification and the initiation of production ramp on our last call, and that's going very well. It is our first EUV node, and that's gone exceptionally well. We actually had, I think, a very good strategy between our technology development team and our manufacturing teams. We did a lot of prework to be able to, first of all, get a more mature EUV tool than maybe some of our competitors started with. That tool has performed well. And then we use that tool on some of our older nodes, including 1 alpha and 1 beta to be able to ensure that all the different elements of the exposure process and improving the manufacturability of that process would be well rung out before we got to 1-gamma. So now as we're ramping 1-gamma, we're focused on normal yield learning, and we feel pretty good about where we are with that. And we think it's going to be -- I think we've talked about that being 30% bit density gain, which we think gives us multiple generations, 1 alpha, actually 1Z, 1 alpha, 1 beta, 1 gamma. We think we've been first to market now for different generations and feeling really good about where that positions us for the future.

You talk about the 30% power reduction relative to your competitors, 30% better power performance on HBM. And I think it has been a big differentiator for the Micron team. As you move to gamma, can you give us a sense, are you anticipating a similar, not only power consumption savings, but better performance dynamics 1-gamma? And maybe where does that intersection point come? One -- you've got your HBM road map here. You've got 1-gamma. Where do you think 1-gamma lines up on your HBM road map?

Yes. No, we haven't provided details on which nodes are going to be for our future HBM yet, we'll do that a little bit later. But definitely feel very good about -- I think a year ago, I mentioned to you that we felt really good about our 1-beta as the foundation for HBM, right? And our implementation of high-k metal gate, our implementation of various other things in process, combined with some of the decisions we made in design and in packaging together to be able to deliver really the best HBM product on the market. A year later, if anything, I think we're even more pleased as we see competitive offerings now trying to work out at different platforms and trying to get qualified at different customers, we feel great about where that is. And I think that just bodes well for our -- all that learning and all that of integrating our high-performance product went into the 1-gamma node and the fine-tuning that we'll have there across our portfolio.

China domestic -- let's talk a little bit about the competition. China domestic memory suppliers, we estimate account for about 6%, 7% of total industry DRAM bit supply, mid- to high single-digits percentage of total industry NAND bit supply. In DRAM, I mean, your China competitors have been more focused on legacy DDR4 applications primarily to serve the domestic markets. But they are reportedly making some inroads on DDR5. I mean, do you see them at all in the market?

D5 is really a significantly more complex product. We're on our fourth generation there, and we've seen the performance requirements continue to step up. And we're also using LP5 as well now into those high-performance applications. So we're very focused on that -- on being able to tune our road map for that. We do see that the domestic Chinese competitors are becoming players in the market, and we do take their presence into account as we make our plans, and we do have that their supply model as part of our long-term supply-demand balance in the industry as we think about it. But we are -- really believe that we -- as we focus on the very highest end applications, the ones that are most demanding in terms of performance, in terms of reliability, in terms of quality and interoperability, the ones that are, frankly, most difficult for qualifications that provide the maximum value to our customers. Those are the ones that we're focused on.

Lots of questions that we get from investors as it relates to the largest memory supplier out there, Samsung and what happens to HBM supply and demand dynamics when they finally do get qualified on HBM3E 12-high right. What's your sense on how the pricing power dynamics may or may not change when they become the third large and qualified supplier of mainstream HBM?

Well, I think what's important to note about HBM is that the qualification process is longer with just about everyone than it is for standard DRAM products. That's really not just a function of how complex the product is, but it's also a function of the cycle time, right? Because you're not just talking about in a standard DRAM product, putting it on a board and then running a call, you're talking about having to go through a long chip on wafer on substrate or KOAS process and then often integrated into a system and then you were able to do the reliability testing, right?

And maybe even in some cases, a full-blown rack scale.

Absolutely. Full-blown rack system. So that cycle time to do a qualification is substantially longer than standard products. Now couple that with the fact that virtually every one of our customers has announced wanting to accelerate the cadence between ramps of new platforms. So instead of the 18 to 24 months on new server platforms or PC or smartphone platforms, we're talking about every 12 months, they want to come out with a new -- they're challenging us, right? Last year was HBM3E, 8 high, this year, 12 high, next year, HBM4, right? So when you have a shortened cadence and a longer time, it's going to be just difficult for many providers to be qualified on same platform. So there could be different platforms that different competitors are qualified on. But we think that, that shortening of the cadence sort of benefits those who are leading.

Got it. I want to talk about something back in mid-April, the team announced that it was reorganizing its business units to a more market segment focused based structure, right, our business units, more transparency on your data center focus. Walk us through the rationale and what you will be disclosing on a quarterly basis, starting with your fiscal Q4 results.

Sure. Thanks. I think this is really an evolution of the way that our end markets are organizing. And so we want to organize the same way. And the high-end computing markets were becoming our huge opportunity. We've talked about the long-term HBM by itself by 2030 being a $100 billion market opportunity, right? So that's as big as all DRAM last year. So we felt like it was a good timing to be able to organize around the fact that, that high-end computing market HBM, high-end, high-capacity DRAM modules as well as low power into these -- some of these high-end applications or supercomputing, AI supercomputing applications. We wanted to have really good focus on that. We then have another segment that's focused on we call core data center, which is more enterprise server class. And then we have -- we combine the client applications into one smartphones and PCs, and then we have our automotive and embedded business. So we think that we're organizing to be able to have better market focus, provide maybe more transparency into each of these areas. We will be providing by each of those segments, revenue, gross margin and operating margin across -- for those 4 segments going forward and I think that should give us -- give everyone better insight into how we're managing the business, what the opportunities are and how we're allocating capital between them to be able to really maximize our opportunities.

Yes. Manish, Samir, thanks for the great participation today. Always appreciate the support. Yes. Thank you.

Thanks, Harlan. Appreciate it.