Thank you for standing by. Welcome to Micron's Post- Earnings Analyst Call. At this time, all participants are in listen-only mode. After the speaker's presentation, there will be a question-and-answer session. As a reminder, to ask a question during the session, you only need to press star 11 on your telephone. As a reminder, today's program is being recorded. N ow I'd like to introduce your host for today's program, Satya Kumar, Investor Relations. Please go ahead, sir.
Thank you. Welcome to Micron Technology's fiscal first quarter 2024 Post Earnings Analyst Call. On the call with me today are Sumit Sadana, Micron's Chief Business Officer, Manish Bhatia, EVP of our global operations, and Mark Murphy, our CFO. As a reminder, the matters we're discussing today include forward-looking statements regarding market demand and supply, market trends and drivers, our expected results, and guidance in other matters. These forward-looking statements are subject to risks and uncertainties that may cause actual results to differ materially from statements made today. We refer you to documents we filed with the SEC, including our most recent Form 10-K and upcoming 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 to conform these statements to actual results. We can now open up for Q&A.
Certainly. One moment for our first question. Our first question comes from the line of Ambrish Srivastava from BMO. Your question, please.
Hi. Excuse me. Thank you. I had a couple of clarifications. Just wanted to make sure I was on the same page as with prior comments. Just talking about the data center inventory cleaning out by first half 2024, that seems to be extended versus, I think, prior comments were in the beginning of the year. And kind of related to that, is my understanding correct that now it's pushed out to first half 2024? And then is that a comment because of the strategic buys that you guys have seen? I t is a Micron-specific comment, or is that an industry-wide comment that you made? And then I had a quick follow-up.
Yeah. Hi, Ambrish. Yes. O n the data center side, there are some customers whose inventory will be normalized, let's say, in the Q1 time frame. There may be some who will get into Q2 before their inventory is normalized. That's why we said sometime in the first half because certain customers have experienced some lower end-market growth in the last many months. And consequently, their consumption of the inventory has been slower. And that is what may bleed into Q2 from Q1. And I'm talking about calendar Q1, Q2 for 2024. And it's nothing much to do with any of the strategic buys so much. And it's more to do with the total extent of the DRAM and NAND inventory at large data center customers. So we believe it'll be mostly normalized in the first half and then more robust buying patterns coming in the second half.
Then as the overall data center general-purpose compute growth accelerates, which you have seen, these general-purpose servers actually are likely to fall for calendar 2023 at a double-digit rate, server unit volume for general-purpose needs. We expect that to stabilize in 2024, be sort of flattish general-purpose server unit volumes, and then grow from that point on. Second half of 2024, calendar year, and then calendar year 2025, more robust growth along all of those fronts.
Got it. Got it. Thank you. So I was correct. A bit of a push-out, but not a Micron-specific comment. That's more an industry comment. For my follow-up, maybe for Manish, as we think about cost down, I think you guys have been talking about high single digits for, I think, a couple of years from when it used to be mid-teens DRAM cost down. Just trying to understand with HBM, higher packaging cost, stacking, does that change? You did give guide for 2024, which is high single digit. But beyond that, does DRAM now start to come lower than that in terms of cost down?
Yeah. Thanks, Ambrish. For front-end, you're right. We did say mid to high single digits was our long-term expectation for front-end or memory-level cost reduction. And that factors in a few different things. One, of course, is technology and how technology is getting more challenging, cadence of technology that's capable for the industry and for us. And yes, there is an element of mix of technology, certainly. We're transitioning from DDR4 to DDR5. And DDR5 has some structurally larger die sizes, and so some cost increases as we move there. And yes, as HBM penetration grows in the industry, we've said for the industry in HBM3E, the die size is double. E ven within the front-end cost, which is what we were guiding to, we do see that the industry's cost reduction capability will be coming down somewhat from what we had seen previously.
Got it. Thank you.
Thank you. One moment for our next question. Our next question comes from the line of Thomas O'Malley from Barclays. Your question, please.
Good afternoon, guys. Thanks for taking my question. At the risk of splitting hairs here, if I look at your deck from last quarter, you talked about DRAM calendar year 2024 bit demand exceeding the long-term CAGR. Now you're saying closer to near the long-term demand. Is that the server side that's causing that downtick? And if you put yourselves back in the position of last quarter, has pricing improved enough where you would expect that that would offset that downtick in demand when you're looking at revenue and free cash flow kind of for the full year? In other words, can that more than offset what you're seeing as a slight downtick in demand? Thank you.
Our overall expectation of demand for 2024 at the aggregate bit level hasn't really changed that much. But you may have noticed that we have taken up our view of 2023 demand growth in DRAM from mid-single digits to high- single digits. And so it's really a movement of some bits across the year-end boundary. And that's what increased some of the bits in 2023. H ence, with that higher base, the year-on-year growth in 2024 is, on a percentage basis, not as high as it used to be earlier. With that said, we have definitely a more confident and optimistic view about pricing than we did three months ago and w e do expect the pricing to and you've seen that in our guidance in terms of revenue as well as gross margin for FQ2.
We do expect pricing to continue to improve for 2024 throughout calendar 2024 and our financial results to improve alongside, mainly driven by the fact that the leading-edge nodes for both DRAM and NAND are oversubscribed for the full year. And so I don't see any kind of an impact on the change that you were mentioning on a revenue basis. The revenue change should be a positive.
Helpful. And then just a quick follow-up. You obviously started the year on the DRAM side very strong from a bit perspective. You're guiding slightly down, I would say, for the next two quarters for February and May. Could you talk about just how severe that downtick may be? And when you look at what you're talking about for mid-teens demand kind of near the long-term CAGR, I mean, it would require a pretty big downtick, at least from your supply perspective, to get there. So just kind of help frame out those next two quarters. Is it stepping down materially from February and then May off of that, or just the linearity of those bit declines would be really helpful? Thank you.
Yeah. Yeah. I can provide a little bit of color on that. I f you just go back to our 2023 fiscal year, our bits bottomed in FQ1 of 2023. Our revenue bottomed in FQ2, an d so throughout the fiscal 2023, our bits have been growing through the course of the fiscal year. W e enter 2024 at a pretty good run rate of bits compared to where 2023 started on a fiscal year basis, and t hen, as you noted, in Q1, our bit growth was pretty strong in DRAM. W e actually don't need a huge amount of sequential growth in 2024 to have pretty robust year-on-year bit growth fiscal 2024 over fiscal 2023 because of those base effects.
With that said, if you just look at the environment right now, our leading-edge supply is obviously pretty constrained due to all of the history of CapEx cuts that we have gone through, the industry has gone through. O ur leading-edge bits are fully subscribed. We do have oversubscribed. We do have some inventory that we will obviously use to supplement our sales, supplement our production in 2024. W ith that said, we have also some seasonality that typically occurs in Q1 for the consumer portions of our business, an d we have not enough supply of leading-edge bits in this time frame. T hose are all of the things that are coming together to cause our bit shipments in DRAM and NAND both to be down a little bit in FQ2 sequentially. On the DRAM side, not growing in FQ3 either, will likely be down in FQ3.
NAND, we'll see where things stand in FQ3. But we do expect both DRAM and NAND to have some modest sequential growth in FQ4, a nd we do expect pricing to be the prime driver throughout each of those quarters for revenue, margins, financial performance throughout that time horizon.
Extremely helpful. Thank you, guys.
Thank you. One moment for our next question. Our next question comes from the line of Joe Moore from Morgan Stanley. Your question, please.
Great. Thank you. I wonder if you could give us some sense of the relative profitability of NAND versus DRAM. At this point, I know you had mentioned in the call that the DRAM mix was still helpful to gross margins. I would have thought with the increase in price that you saw in NAND that you're starting to get close. Can you just give us some sense of how close the two quantities are?
Yeah. I mean, the profitability of DRAM at a gross margin level is definitely higher than that of NAND. This is continued trend that has been in place for quite some time where DRAM profitability throughout the cycle tends to be higher than NAND profitability. With that said, NAND did decline at an industry gross margin level more than even DRAM in this downturn. W e do expect that because it declined more from an ASP perspective and from a gross margin perspective, the bounce will obviously have to be steeper in NAND in order to restore some level of acceptable ROI, s o the pace of price increases in NAND to dig out of that hole will have to be just mathematically bigger. W e are seeing tightness in both DRAM and NAND, especially on the leading-edge nodes. They are in shortage in both DRAM and NAND.
Both of those should continue to improve. But just because NAND has a deeper hole to dig out of, its percentage increases will just be higher.
Great. That's helpful. Then for an unrelated follow-up, I wonder if you look at the high capital spending in memory in China, obviously, severe limitations on what they're allowed to spend money on. But you might have geopolitical incentive to use those bits in devices where you wouldn't ordinarily use trailing-edge bits. Can you just give us some sense of how you guys perceive that supply coming out of China to be relevant to the global supply demand?
Sure, sure, Joe. So I think you correctly highlight that technologies that are in production in China are behind the leading-edge nodes in the industry, a nd so we do have modeled the supply for both DRAM and NAND as part of the global demand that we see or global supply against the global demand models that we have. And we do model them being able to continue. However, their impact on the market, given the fact that they're trailing, is commensurate with that position, both based on the ability for those technologies to generate supply as well as their ability to compete in higher-end applications when they have nodes that are less competitive in terms of features or performance. F or example, as industry transitions to more complicated products like HBM or others.
Great. Thank you very much.
Thank you. One moment for our next question. Our next question comes from the line of Quinn Bolton from Needham & Company. Your question, please.
Hi guys. So thanks for taking my question. I guess maybe just wanted to ask a question on the front-end cost reductions in fiscal 2024 for mid- to high-single digits in DRAM and low double digits in NAND. I thought previously, you guys thought that the cost reductions in fiscal 2024 might be affected by underutilization charges and those costs flowing through inventory and through COGS. And so I guess I'm wondering, can you clarify, is that inclusive of underutilization charges, or is that just front-end wafer and excludes any utilization or underutilization effect?
Yeah. It does include the underutilization effect. And Quinn, it's a good question. And in the case of DRAM, it's all in. But operationally, we do expect FY 2024 to be a year of improved efficiencies across the factories. As we've talked about, we've got structurally lower capacity as we migrate idle tools from lagging to leading nodes. Now, there's still some underutilization in those lagging nodes. But we will see the idle charges decrease through the year and starting in second quarter. And between that and then just the tech node migrations, more of the capacity going to leading nodes, we're going to see that positively affect the front-end cost reductions for both DRAM and NAND. So you're seeing that happen here, incrementally positive to the view we had before. We've covered it quite a bit over the past year. So I'm not going to go through all that.
But in summary, some of the underloading effect, as you know, hits COGS as period costs while some of it's absorbed, a nd the part that's absorbed is in the higher-cost inventory. In fiscal Q2, we'll have much lower idle-related period costs. A gain, it's mostly related to this increased utilization from and the idle costs remaining will be mostly legacy node-related. T hen you'll have this portion of underload that's in inventories as higher costs. And as we move through the year, this will clear and become less of an effect. So in total, we end up with what we communicated today, which is front-end costs and, in the case of DRAM, all-in costs, mid to high single digits down for 2024. Now, in NAND, we said low teens in front-end costs down, which we will see.
But because of more of a mix shift on components to modules, we'll see below 10% more of mid to high single-digit cost downs for NAND all-in just for your modeling.
Got it. That's helpful. And then I guess this follow-up question, just wanted to come back to the question that was asked on the call about HBM3E. If you guys have 10% better performance, 30% better power, do you expect to be sole-sourced on the SKUs you win, or to the extent that there are multiple sources, will the customer have to derate your HBM to have similar performance across the two SKUs?
Yeah. I mean, it's a good question. Our customers have been amazed at the capabilities that we have created with HBM3E versus what others have been able to do with this product, a nd this has been part of our strategy to de-risk HBM3 and go to HBM3E, be first to market with it, create a world-class product better than anyone else, a nd this is panning out. Our customers have different approaches that they are assessing in how to take advantage of this capability on the performance and power side in terms of what type of products they use our HBM in, what kind of SKUs do they create with that, what kind of multi-sourcing strategy do they have to support their whole portfolio.
While we are working with them on our designs and integration of our designs with their GPUs, we are not at liberty to discuss those strategies. We'll leave it at that.
Understood. Thank you.
Thank you. One moment for our next question. Our next question comes from the line of Vijay Rakesh from Mizuho Securities. Your question, please.
Yeah. Hi, and a q uestion on your when you look at your DRAM and NAND CapEx and the comments that you made on the bit supply, can you talk to what your wafer capacity would be exiting calendar 2023 and how you see that exiting calendar 2024?
I think what we've given the Vijay, I think we've given guidance mainly on sort of fiscal year to fiscal year. If you allow me to do that, we've said that our wafer capacity for both DRAM and NAND wafer output capacity was approaching 30% below our prior peak levels as we began this fiscal year. The majority of that was underutilization of equipment. As we transition through the year and as Mark noted earlier, we'll be utilizing some of that underutilized equipment towards transitioning to the leading-edge nodes, 1-beta and 232-layer NAND.
As we do that throughout the year, by the time we exit fiscal year or beginning of the next fiscal year, fiscal 2025, the majority of that reduction will be actually the derated capacity or reduction in capacity due to the longer process cycle times and the equipment being utilized for more complex nodes. We haven't really and the production capacity will be down meaningfully from our peak levels. We haven't really given specifics on how much. It'll just be meaningfully down even as a result of this. We still will have some underutilization, primarily on legacy nodes, going through into fiscal 2025 as well.
Got it. And the NAND said very similar, or?
I'm sorry. Those were all about DRAM and NAND comments. Yeah.
Okay. Got it.
Yeah. By the beginning of 2025, yeah.
Yeah. Sorry. Go ahead.
No, I was just going to say, by the beginning of 2025, we would expect that the majority of the reduction in capacity came from converting equipment to the newer nodes versus a decision we've taken on underutilizing and idling equipment.
Got it. T hen on the HBM side, any similar commentary on what your capacity would be on HBM within DRAM, I guess, exiting fiscal 2023 versus fiscal 2024 or fiscal 2025?
We haven't provided much detail other than to say that we expect our share of HBM to intercept our share of the overall DRAM market sometime in calendar year 2025. T hat gives you some sense of sort of where we have to go. As you know, we have not had as large a share. So we're starting from a trailing position. We had some capacity and some production on HBM2E to get good learning. But the HBM3E will be a steep ramp. We have good alignment agreement with the equipment suppliers, a nd we have a plan to be able to meet $several hundred million of revenue within fiscal year 2024, which I think we talked about in the script.
Then, of course, that will position us well to be able to grow continuing into fiscal year 2025 towards that goal of being able to intercept our overall industry market share sometime in calendar year 2025.
Vijay, maybe I'll add just maybe a CapEx-related comment on profile. We gave an estimate, $7.5 billion-$8 billion per CapEx based on the guidance we gave for the first- half CapEx. The second- half CapEx will be a bit higher than the first- half. That's a function of HBM spend and also greenfield capacity spend. Certainly, we've made some grant assumptions on size and timing. Of course, that's reflected in our guidance. If that were to change materially, then it may affect our numbers, both CapEx and cash flow.
Right. Just a slight clarification amongst us, greenfield capacity, but it's really greenfield facilities to support future capacity.
Yes. End -of- decade related demand.
All right. Great. Thank you. Thank you.
Thank you. One moment for our next question. Our next question comes from the line of Sidney Ho from Deutsche Bank. Your question, please.
Great. Thanks for the question. So just to follow up to a couple of questions from the last two questions, what was the idle charges in fiscal Q1? And what are you baking in for fiscal Q2 in your gross margin 13% guidance? W hy would idle charges come down when bit shipment is expected to decline somewhat in fiscal Q2 and again in Q3? Is it just the use of underutilized equipment for no transition that you guys just talked about?
Yeah. Sidney, the idle charges in Q1 were $165 million, that's down from over $220 million in the fourth quarter. They'll go down significantly first quarter to second quarter. A gain, this is a function of utilization is going up in the network, a nd underutilization is going to largely reside in legacy nodes. A gain, as we talked about last call and this call, that's happening in part because we've redeployed some legacy node equipment to leading nodes, a nd the wafer output's down. So the denominator in the utilization's larger or smaller. A gain, the idle charges are related to when utilization is under a certain level, you cannot capitalize those costs in inventory. S ince utilization is increasing, we pass that level for leading node and have idle charges remaining to lagging nodes.
Maybe Sidney, just a little follow-up on the second part of your question on bit shipments and timing of bit shipments. Keep in mind the process cycle time in fab as we transition to these newer nodes is longer. T he cycle time in the fab is longer, right, as we make these transitions, so you're converting to a node more complex, better technology. But it is longer cycle time in the fab. And then we're also, as Mark mentioned, particularly for NAND, transitioning and targeting our portfolio towards higher-value solutions. Those higher-value solutions also have longer cycle time in assembly and test.
So sort of the cycle time and that's why when we say we're constrained on leading-edge nodes, it's not immediately when you're talking about leading-edge technology into leading-edge high-value products, the cycle time for us to be able to deliver more bit shipments is longer than maybe it was last year when we had more of a component focus.
No, that's super helpful. I mean, my follow-up question is, if I look at DRAM ASP, it was up slightly in fiscal Q1. That's kind of expected. I think previously, you said it was limited by customers. Strategic buys occurred in prior quarters. Is there any way you can help us think what would have been gross margin would have been without the headwinds from these strategic buys? If you look at fiscal 2Q, part of the quarter is still within that, we'll probably see the same headwind, right, because most of these pre-buys end at the end of the calendar year. Is it fair that there will be another step up in fiscal Q3 that is above and beyond market trends that we should be thinking about? Thanks.
Yeah, it is true that fiscal Q1 did get impacted by some strategic buys in terms of pricing realized. But keep in mind that those strategic buys were implemented or agreed upon in CQ3. F urther, FQ1 has the effect of September in there, which is part of CQ3 pricing, which was still under pressure. T hen two months of CQ4 where we did better. The other aspect was we came into FQ1 with robust bookings even beyond the strategic buys. And these robust bookings included pricing that had been negotiated earlier before the inflection happened. So we knew based on those robust bookings, we were able to then use those robust bookings to push pricing higher than we had originally expected. And so CQ4 pricing ended up becoming much better than we had originally planned because of that.
Of course, you're seeing that our FQ2 guidance on revenue is going up substantially from $4.7 billion in FQ1 to $5.3 billion ±$200 million without the benefit of bit growth, right, because bit growth is not a driver in those numbers. Yes, December is one month of CQ4 in our FQ2. Then we'll have two months of CQ1 in there in our FQ2, January and February. But all of the expectation that we have right now is for continued increases in pricing. For a lot of the customers for whom we do pricing negotiations on a calendar quarter basis, every fiscal quarter, we'll have this effect of one month previous quarter, two months next quarter. That's how it'll flow into our P&L.
Okay. Great. Thanks.
Thank you. One moment for our next question. Our next question comes from the line of Vivek Arya from Bank of America. Your question, please.
Thanks for taking my questions. I had two as well. The first one on just the concept of the pricing improvement. Historically, the memory industry has had a challenging time forecasting pricing for more than 1+ quarter. But now you're suggesting pricing can continue to improve not just in the current quarter, but I think I heard for 2 more quarters. And the main reason I heard was tightness in leading edge, which I assume you mean your supply tightness in leading edge. So if you could just give us more color, what's giving you the visibility that pricing can continue to improve for several quarters? Because when I think of leading edge, I'm reminded of the data center. But that's the segment you are saying is most oversupplied, at least in the near term.
So just help us understand what is helping drive that confidence and visibility that pricing can continue to improve for several quarters?
Yeah. Yeah. I think the pricing outlook, we can think of from two angles. One is the demand side. One is the supply side. I think on the supply side, you have seen that the industry has had a challenging time with margins and so many announced CapEx cuts, underutilization, etc. So the supply has been impacted quite a bit. We have also spoken in our last earnings call as well as in this current earnings call about how underutilization is going to morph into lower capacity. So the overall capacity is getting impacted by all of these capital-efficient approaches to moving tools from the underutilized older nodes to newer nodes. So we have constrained capacity overall. And then, of course, HBM ramp that is taking place in the industry. We have mentioned to you the more than 2-to-1 trade ratio.
If we manufacture 1 bit of HBM based on the end-to-end yields, we end up with more than 2 bits lost of DDR5 in the process. So all of those things are pressuring supply. And when you get to lower amount of overall capacity and then you decide at some point that the ROI justifies increasing CapEx, you obviously have to first reach that point where you can justify increasing CapEx. From that point, you have 2 lead times. And then you have all of the cycle time of the front end and back end, which typically is 4.5-5 months depending on the complexity of the product. So tool cycle time, let's say scanners and so on, 9 months, 5 months for fab front end and back end. You are at 14 months before output can be impacted once capacity needs to be increased.
So it's a long cycle time to increase capacity and get those bits in the hands of customers in terms of products that they need. So there is that supply timeline and supply impacts that are happening in the industry. On the demand side, we have mentioned multiple drivers in our earnings call and other public comments about what's driving the demand in the short term. We have normalized inventory in smartphones and PCs. Auto and industrial continue to grow robustly from a bit demand perspective. So these things are helping drive and of course, the average capacity increase is happening across the board. So these things are driving bit demand. And yes, you're right. The data center is certainly not yet normalized in terms of inventory. But keep in mind that data center is not the only user of leading-edge technology products.
For example, when we think about smartphones and when we think about PCs, they all use leading-edge technology. We have spoken about DDR5 transition from DDR4. Most of our DDR5 is in leading-edge nodes. And so that's one example. All of our client and data center SSDs, we have mentioned how we have reached a second consecutive record in data center SSD. That's all leading-edge nodes. And so it's not just the data center that is using leading-edge nodes. It's also multiple other parts of the market. And then you're right. We don't usually forecast pricing. This time around, we did talk about our expectation of pricing increasing throughout calendar 2024. I mentioned the supply side of it. The other aspect is we are seeing this tightness in leading edge even before data center has become a material purchaser of DRAM and NAND. They are still in inventory consumption mode.
So their purchases of DRAM and NAND are still below their consumption, right? So they are consuming DRAM and NAND from inventory more. And so once their inventory normalizes, they will become a more active purchaser of DRAM and NAND in the second half so starting in sometime first half and certainly second half of calendar 2024. So that'll become an additional driver. And all of our 2024 numbers don't require very dramatic growth in PC units or smartphone units or even general-purpose server units. Our growth expectations are all very modest in these areas. So any kind of uptick, especially later in calendar 2024, will just add to demand in an environment where supply is not easy to come about in this time frame given this timeline I mentioned on when supply can be added.
So the last point I will mention is this has been a very unusual decline for the whole industry, certainly most challenging in 15 years, but unusual in the aftermath of the COVID upswing that happened, which needed to be worked off, etc. So there has been an impact to industry balance sheets. There has been an impact to the overall ROI. And certainly, from a Micron perspective, we are going to be very focused on ensuring supply discipline, ensuring we invest based on improving ROI, and ensuring we do think very carefully about supply-demand balance in how we make our own plans. So when I put all of these factors together, that's what gives us confidence about a little bit more extended timeline where we provided our guidance and view of our expectation of pricing strengthening throughout 2024.
Very helpful. And then maybe as a follow-up, actually, two clarifications for Mark. So one is on the CHIPS grant, what has been contemplated. For many other companies, it's about 10%-20% of CapEx. So that suggests gross CapEx could be higher next year, right? And I don't know whether that means if WFE will also be higher on a gross basis even though on a net basis maybe. So I don't know if there's any issue there. And the other clarification, Mark, I don't know if you have already spoken to it, but the drivers of gross margin expansion in Q3 and Q4, is it just pricing, or is there anything else? Maybe you mentioned where perhaps I didn't catch it before. Thank you.
Yeah. Vivek on your first question. As we disclosed, we did submit applications for CHIPS funding for two fab expansion areas in the U.S. Those negotiations are underway. We're not able to disclose any more at this time. We've made assumptions, as I mentioned earlier, about grant timing and size based on our understanding of potential there, but nothing else to disclose at this time. On drivers of gross margin expansion through the year, it's going to be we didn't provide any color other than it will expand, we believe, through the year, second to third and third to fourth. That will be almost exclusively priced in the second to third. There will be some net cost benefit. Though that'll be a bit muted because we will lose that lower-cost inventory benefit, which that inventory last clears in the second quarter.
And then we'll see price appreciation again. And that will be the largest factor again in the fourth quarter gross margin expansion.
Thank you.
Thank you. This does conclude the question-and-answer session as well as today's program. Thank you, everyone, for your participation. You may now disconnect.