Thank you for standing by, and welcome to Micron’s post-earnings analyst call. I’d now like to introduce your host for today’s program, Satya Kumar. Please go ahead, sir.

Yes. Thank you, Jonathan, and welcome to Micron Technologies’ Fiscal Third Quarter 2025 Post-Earnings Analyst Call. On the call today with me today are Sumit Sadana, Micron’s Chief Business Officer; Manish Bhatia, EVP of 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, and our expected results and guidance, and 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 have filed with the SEC, including our most recent Form 10-Q and upcoming 10-Q, for a discussion of risks that may affect our results. Although we believe that the expectations reflected in the forward-looking statements are reasonable, we cannot guarantee future results, levels of activity, performance and 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 the call up for Q&A. 1

And our first question for today comes from the line of Karl Ackerman from BNP Paribas.

I have 2, please. This quarter, you raised your DRAM bit demand outlook for calendar ’25 to high teens from mid-teens. Is that only coming from your outlook on HBM? Or do you believe, excluding tariff pull-ins from certain customers that demand outlook for non-HBM DRAM has also improved? And I have a follow-up, please.

Yes. Karl, this is Sumit here. This is a small change to our DRAM outlook for calendar ’25. We upgraded our view to the level you mentioned. Now there is no impact of pull-ins in this assessment because whatever modest amount of pull-in there might have been, for example, in CQ2 from CQ3 or FQ3 from FQ4, et cetera, that kind of pull-in activity is net neutral for the full calendar year view. So the improvement is not impacted by any kind of pull-in activity. The strength in our view of the AI business, i.e., the data center growth, continues to be pretty robust. And our view of the AI demand is certainly supportive of the forecast that we have provided you for bit growth in calendar ’25. And the other positive thing that has happened over the last 3 months is that we are seeing, after many, many quarters of a very challenging environment in the broad distribution markets, industrial markets and some parts of the automotive market, we are now starting to see really good improvements in demand in these broad distribution and industrial markets. And that is also providing a tailwind to the overall calendar ’25 bit growth estimates.

2 Mark, given your expectations for another quarter of healthy free cash flow in the August quarter, I was hoping you could discuss how you might prioritize improving net leverage versus share buybacks, just given you elected to not repurchase stock in the quarter?

Yes, Karl, as I mentioned in the earnings call, we’re pleased with the deleveraging that’s happened. We’re down now to a net debt of $3 billion, substantially down from the prior quarter. And I said on the call that we expect to generate free cash flow in the fourth quarter. I would add that we stand at record levels of cash and record levels of liquidity for the company. We’re at $15.7 billion of liquidity, including our untapped facility. So we’re in a great position to first continue to invest in the priorities of the business, maintaining our technology leadership, investing in needed capacity in DRAM to serve HBM and other high-value, high-return markets and then do return capital to shareholders through our routine dividend, which we hope to grow over time, and then oppor- tunistic repurchase. So I think the balance sheet is in a great spot now, and we have a lot of flexibility. And on a covenant basis, I would add that we’re actually basically at no leverage on a covenant basis.

And our next question comes from the line of Aaron Rakers from Wells Fargo.

Two, if I can as well. I think, Mark, first of all, I appreciate the gross margin guidance. I think last quarter, we talked a little bit about start-up costs and the potential for underutilization flowing through inventory on the NAND flash business. Obviously, demand dynamics have improved. I’m just curious, how should we think about those inputs to gross margin over the next couple of quarters? 3 And then I’ll just ask my second question right out of the gate. The LPDDR business in the data center side, we’ve talked for several quarters about you being predominantly the sole provider of those solutions into some of the AI silicon. I guess how do we think about that business as it progresses to $1 billion plus of quarterly revenue? I think you’ve alluded to in the past as far as diversity of customer base there. Has that expanded? Do you see a broadening market for that solution in data center?

Yes, I’ll start, Aaron. So as it relates to under-load utilization-related period costs, we brought those up, I think it was on the December call. And over the past 6 months, we sort of worked – the last call, the March call, we talked about how the structural capacity had come down and that we weren’t incurring those period costs, they would end up being absorbed in inventories and pass-through. And that’s what’s happening. We have structurally lowered capacity in NAND, and those costs all now just passed through on inventories. The NAND volumes were very strong. So we’re benefiting from some more loading there. But still, we’re very careful about capacity in that market because that part of the business, the market environment is challenging. So again, we talk about low CapEx levels, careful about node transition, where our bits are going and premium products in that market and watching inventories carefully. Inventories improved in NAND actually in the quarter, but they remain less healthy than DRAM, which DRAM, we actually expect to be below target by the end of the year. As it relates to the start-up costs that I mentioned last quarter, we do see those beginning to increase, and very modestly now, but through ’26, we’ll see some increase related to – as you know, we’re building Idaho 1, and we’ll begin to see those beginning to increase in ’26. We also see some FX-related costs. I mean, everybody knows that the dollar has been 4 weakening, and so we’re seeing a bit more of that. But since we made those comments, the revenue outlook that we have is much higher. So those effects are 1 basis point less than what we talked about on the last call. Now I think that will have me reinforce, what’s important is are we growing and we are positioning our bits in the right place. And we serve a market that’s growing, and we’re getting leverage on those costs. And then our price performance is critical. So we’ve done a great job of pricing in this market that’s turning and becoming more constructive and which is why we have a strong guide in the fourth quarter of 42%. And then we feel positive about the trajectory of the business beyond that.

And in terms of your question about LPDDR in the data center, we continue to be very excited about our sole source position in that market. It’s become quite a large market and, of course, pioneered by one large company in the data center. Of course, as you know, the power consumption in the data center continues to be an important driver of decisions and designs and architectural choices by our customers. And overall availability of power is an important driver of many decisions, including location of these data centers. So certainly, the interest in more broadly deploying LPDRAM in the data center amongst customers of scale continues to be high. And we do expect, based on our interaction with customers as the pioneer of LPDRAM in the data center in the world, we have multiple interactions going on with customers of different types and scale. And we do expect that LP will grow in its penetration in the data center over time. And we are very excited by our position and we’ll be able to leverage that position as that growth unfolds.

And our next question comes from the line of Brian Chin from Stifel. 5

I know you indicated on the call, in the slides, that DDR4 will be a low single-digit percent revenue exposure in fiscal second half of ’25. But is there a way to quantify how much of the 300 basis point sequential higher gross margin guide for fiscal 4Q maybe is attributed to DDR4 pricing? Because I think that is a question that we’re getting sort of post-call here.

Yes. I’ll take that. The D4 pricing, that market has gotten tight as was talked about in our prepared remarks. Customers are beginning to see increasing shortages for D4. As you mentioned, it’s a smaller part of our business and, on a volume basis, getting smaller. The D4 pricing is a positive Q3 results and Q4 guide. But it’s one of the many factors contributing to the positive results, it’s not a driver of that margin expansion. So I would leave it at that.

Got it. And that’s sort of – I know previously, you said 10% revenue could be sort of DDR4 and now it’s lower, maybe part of that is because of strengthening in other, DDR5 and other areas, a mix. But I guess is this pretty consistent with where your production is going as well as your finished good inventory in terms of DDR4 being at sort of like a sub-10% kind of exposure?

Yes. Just to clarify, right, I mean our earlier comment about the 10% of revenue was a combination of DDR4 and LPDDR4. And the comments that Mark made and what we provided in the prepared remarks as well relate to DDR4 only, which is a low single-digit percent of our revenue. So LPDDR4 and DDR4 is together around 10% of revenue. DDR4 alone is low single-digit percent of revenue. And a lot of the things that you see on the spot market in terms of pricing relate to DDR4 only, and that part is a very small part of 6 our overall business.

Got it. Maybe just one quick follow-up. It sounds like mix, less consumer orientation in terms of the bit shipments in fiscal 4Q. Is there any reason not to – I know you don’t want to guide fiscal 1Q, November gross margins per se, but it would seem seasonal and kind of reasonableness that, that mix should only continue to get less consumer-oriented into fiscal 1Q. So any reason not to think that wouldn’t be a bit of a tailwind on top of sort of shipping more 12-high HBM3E into the fiscal first quarter in terms of gross margins?

Yes, Brian, we didn’t guide – we’re not going to guide first quarter. But to your point, in the fourth quarter, we do see 2 forms of favorable mix at a high level. One is just more DRAM growth relative to NAND, and so mixing more DRAM, and then more data center relative to consumer. So you’ve got those positive mix effects. And then to your point on what’s the trajectory into the November quarter, the market, we’ve talked about our inventories are very lean. We’re a bit constrained as we look out to that quarter. So we’re focused on the best pricing decisions. And then we’re continuing to work for a favorable mix. So that’s why on this call earlier, we said that we can expand gross margins in November quarter, we believe.

And our next question comes from the line of Krish Sankar from TD Cowen.

This is Eddy for Krish. Congrats on the strong results. If we go back June of last year, you guys did give us some color about 2025 demand for HBM. I think you mentioned like a multiple billion dollar opportunity, and you gave some color about the pricing, that pricing was being negotiated. But this time, you guys chose not to. I wonder like are there 7 any specific reasons? Like, are lead times different for HBM? Or is customer behavior different or the competitive environment? Like, any color would be great. And on that note, I wonder like when you guys are expecting to hear from your customer about 2026 demand. That’s it for me.

Yes. So in terms of 2026 HBM, we do expect that HBM bit growth in ’26 over 2025 calendar year will be significantly faster than the overall DRAM bit growth. So HBM will continue to outrun the DRAM bit growth. Now when it comes to the actual discussions with customers, compared to this time last year, of course, we provided you a lot of details about 2025 expectations, and a lot of updates on color every quarter on what kind of milestones we were reaching on HBM, and what kind of capacity ramp and yield improvements and percentage growth rates on quarters, and $1 billion quarter, 50% growth after that in FQ3 from that $1 billion quarter in FQ2, et cetera, $1 billion-plus quarter in FQ2. So we have provided you a lot of these data points over the last several quarters because we were still in the process of ramping our capacity and ramping our revenue and ramping our scale in HBM. Of course, HBM is now over a $6 billion run rate business for us based on the last quarter’s reported numbers. And so we are now at a place where we can not only offer customers with our HBM products, world’s best technology based on our very mature capability in 1-beta node, but also the best power consumption capabilities. So we have not only that, which we also had last year, but we also now have the capacity in place and the scale in place to be able to meet their growing demand. So we have earned the trust of our customers. We have terrific engagements with all of the major HBM consumers in the world, and we are deeply embedded in their road maps, not just for HBM4, which we had just sampled, but also for HBM4E, where we have multiyear R&D engagements to do co-design work that relate to the custom-based die that HBM4E will have. So we are 8 very excited about the growth opportunity. We are very confident about our place in the competitive landscape. We are very happy and glad to see the contributions of our team on the execution front from ramping the high-quality output to getting the financial benefits of growing this business in our portfolio. In terms of the discussions with the customers, our customers are going through a lot of rapid transitions. We’re going from 8-high to 12-high that is in progress right now on HBM3. HBM4 has been sampled to them by us, and they are considering the timing of their platform transitions that will leverage HBM4. And they’re still – most of our customers are still in the process of finalizing their plans for 2026 in terms of the transition time lines of their key platforms, which will then determine the mix of products they have to purchase. And HBM products, as you know, come with very different type of lead times than the rest of the business. We like to ensure that when we start these wafers, we have very good visibility to the demand and good agreement and understanding with our customers on what volumes they will take in what time frame. And consequently, our customers have a higher bar in terms of providing those forecasts to us, a higher bar of fidelity of quality of those forecasts. And so they have to go through the assessment of what their platform transition timings will be, what their volumes by product will be. And that business, as it has scaled, has become much more complex due to its scale. And we are just going through those discussions with those customers. And as and when these agreements – as and when we have the full visibility to all of these numbers and the agreements will fall in place, then we have confidence that we will end up with a terrific trajectory of our HBM business.

And our next question comes from the line of C.J. Muse from Cantor Fitzgerald.

9 I was hoping you could hit on, I guess, where you think you are seeing tariff-related pullins. And if you can kind of comment between both DRAM and NAND and kind of con- sumer and nonconsumer. And you did take up your DRAM bit outlook for the year. Can you also discuss what’s driving the uptick there?

Yes. I think I had mentioned earlier in the call, I’ll take your second part of the question first, I had mentioned in the earlier part of the call that the improvement in the DRAM outlook for the calendar year 2025 has nothing to do with any kind of tariff-related movement that may have occurred for some customers. It has a lot to do with the fact that we continue to see robust demand driven by AI in the data center. This is not just HBM demand, but all data center-related DRAM demand. We are seeing improvements from the CQ1 levels of data center SSD type of demand there as well. But really, the other area of improvement beyond the data center in DRAM has come from our improved view of what’s happening in the industrial and broad distribution markets. These markets have not grown much for many quarters, and you have seen that view across the semiconductor industry. But now these markets seem to be improving their growth trajectory, firming up their growth forecast. So that’s helping 2025 view of DRAM growth. Now as it relates to the tariffs and what some customers may have done, first, I just wanted to reiterate that whatever impact tariffs may have had on customer order patterns, we think that overall impact is fairly modest in our FQ3 numbers. And we actually are not too concerned about what the tariff-related impact has been because the customer – our customers’ aggregate demand signals for the remainder of calendar 2025 continue to be healthy. And so we feel like we are in a constructive demand environment for the next – for the remaining part of calendar ’25. And of course, there is uncertainty related to tariffs and whatever may hence happen to the macro environment, and we are mindful of that uncertainty, but we also know that we are one of the most agile businesses 10 out there in responding to any changes that may occur. So with that said, the tariff-related movement from certain customers may have mainly been to stage inventory of their own finished goods in different parts in different geographies and that may have created some need for memory products in terms of a pull-in effect. But again, the overall impact fairly modest.

— Operator Instructions — And our next question comes from the line of Vijay Rakesh from Mizuho.

Just a quick follow-up. When you look at your HBM4, I think you guys mentioned the CMOS logic die. Can you talk to how that compares to – I think some of your peers have talked about FinFET as well and how that could drive higher ASPs on your HBM4, I guess, versus HBM3E?

Yes. The HBM4 cost and price is expected to be higher than HBM3. We do expect that – the product itself has remarkably better specs, as you have seen from our press release as well on HBM4. So in terms of value that it creates at the system level is also very robust value for our customers. But we do expect that the pricing on HBM4 on a per bit basis will be higher than that of HBM3E. And so we definitely will look forward to the tailwind from that growth as well.

Vijay, I think Sanjay made this point – this is Manish. I think Sanjay made this point on the call, and maybe this is – everyone knows this, but the JEDEC standard die size for HBM4 is larger than it is for HBM3E for all players. So that’s why the trade ratio, the cost with HBM4 11 will be higher. And just to reiterate, we’re building our HBM4 on our now 2-years-old 1beta technology, very, very proven, very strong performing product and where, similarly, our logic die is going to be built internally. So the integration between the memory and the logic is also proven from our HBM3E processes. In the advanced packaging process, we’re leveraging all the learnings. So we feel very good about the road map decisions that we’ve made relative to HBM4 in order to intersect the market as soon as our customers are ready with a terrific performing product as well.

Got it. And just one quick last one. On the pull-in, I’m sorry to beat a dead horse, I know you said still seeing second half strong data center, but is there any pull-in from Q1 into 4Q? Or is that – should we be concerned about that? And can you frame that a little bit?

Yes. We think that any kind of pull-in activity, in aggregate, whether it is Q3, FQ3, FQ4, is modest amount of impact. And we continue to feel that the overall demand environment beyond the FQ4 time horizon, which is the rest of the calendar year beyond FQ4, remains in a healthy place in terms of our customer signals. This is part of the reason why we also feel good about driving the pricing of our products. Our inventories are in a good place. We expect the pricing to improve in FQ4. And again, we don’t see that as just a very much tariff-related or pull-in related effect. It’s more of an overall market environment impact. And that’s why we have also provided you a view of DRAM bit growth in calendar ’25, which is now improved versus what we mentioned to you just a quarter ago.

And our next question comes from the line of Timothy Arcuri from UBS.

I’m trying to figure out just the pricing modeling on HBM. So nearly half of the dollar rev12 enue growth in DRAM came from HBM, where we know that the price is many multiples of the non-HBM. I don’t know, most would say it’s probably 4.5 to like 5x. So I get that the HBM ASP – or sorry, that the non-HBM ASP would have come down because of the mix to consumer. But why would the HBM ASP have come down? Because it had to come down a lot for overall DRAM pricing to be down. If half of the dollar growth came from HBM, pricing had to be down in HBM. So why would that have been the case? Is this like product transition related? And I guess I’m just asking, like, is the pricing noisy on a sequential basis? Yes.

Yes. I mean the HBM pricing is not noisy, and it’s not going up and down in the way you’re describing. I think if you look at our results, and particularly, I’ll point you to the growth in our mobile business, which has been very significant, the price points in mobile tend to be quite low compared to data center. And so definitely, when you look at the overall mix – and the mobile business is a pretty sizable business. It’s not a trivial business. It’s a large part of the TAM as well. And so when you look at the overall effect, on a like-for-like basis, we have improved pricing, we have growth in HBM happening, but we also have fairly significant increase in consumer-oriented mix in FQ3. We had highlighted – when we provided guidance for FQ3 last quarter, we highlighted to you that we would be having a significant increase in our consumer mix. And that has been driven by the fact that it took several quarters for the inventories in client and smartphones to get to healthy levels after being heavy in inventory in the second half of calendar 2024. And we were shipping into these markets on the consumer side at a very low rate compared to the ship out rate from these markets because of that inventory level. So when that inventory level became normalized, our growth rate in FQ3 has suddenly jumped in a lot of these consumer end markets, which carry lower price points. So even though we have had very steady pricing on the HBM front and like-for-like pric13 ing increases, the mix change has been so sharp that it has caused the impacts on the high-level pricing that we were observing in the numbers.

Right. So the message is that HBM price was still up. It’s that the non-HBM price was down such that the blended ASP was down low single digits. That’s the message, correct?

Yes. I mean the message is that the HBM pricing is steady, and the pricing we have on the DRAM side like-for-like has improved sequentially. And it’s really the mix change on the consumer side that has done this. Now we spoke about this on the DRAM side. But of course, on the NAND side, we have some different dynamics, including the fact that the market environment is not as healthy as it is on the DRAM side, it’s a more challenging environment on the NAND side. And so the NAND dynamics are different. But I just wanted to make sure that I clarify that the DRAM dynamics are very healthy.

Totally. Totally great. And then just one last quick one, just on the cost. So costs had to be down a lot in both DRAM and NAND, had to be down like mid-single digits in both, definitely better than what I would have thought. Why was that? I know, Mark, you talked about these NAND headwinds, but it seems like they weren’t anywhere near as bad as what you thought they’d be. So how do the costs work so that they were down so much?

Mark? And then I’ll add some color.

Yes. I would just say, Tim, that I mean, mix affects costs, as you know. And so there’s some of those effects. I think just on costs overall, front-end bit cost reductions, HBM 14 and DRAM and for NAND, they’re generally on track with the front-end cost reduction ranges that we’ve talked about before. So I think if you’re looking to model things going forward, I think that’s something to keep in mind. Our business is changing to this mix issue. Our business is changing with our focus on the higher performance part of the market. Sometimes that has some higher costs associated with it, but it’s higher margin. And so we get higher value out of that. And again, the mix is really affecting the cost. So you’ll see us move away from – traditionally, what we would do is drill into the cost detail every call because it’s becoming a harder metric to follow with the improving mix that we are driving in the business.

And Tim, I’ll just add a little. I mean, I mentioned on the call that our HBM 12-high yields are doing better than our initial HBM 8-high ramp was a year ago at this time and better than our expectations. So overall HBM came in a little better than our expectations last quarter. And there were certain other areas that Mark and Sanjay both commented on a little bit that our cost performance was operationally better. But as Mark mentioned, a lot of it is due to mix.

And our final question today comes from the line of Chris Danely from Citi.

This is James filling in for Chris. I just had one question. I think it was probably about a conference a couple of months ago, but you said the August quarter gross margins could be flattish to slightly up compared to this quarter. Was there any major change in that other than some pricing and you talked about the data center mix? But just wanted to dig on that. 15

Yes. I think I covered this on the main call that when we provide that guidance, it was in a period of transition. We did know that inventory outlook was improving. And we indicated on that call that we intended to work to inflect price. The conditions of the market are just better than we expected when we gave that guidance. And our performance, particularly on price relative to what we thought the market would bear was better. And then that’s carrying into the fourth quarter. So you reset the baseline, we did 39% versus our guide of 36.5%. And so from that baseline of 39%, we’re seeing still a constructive market environment, particularly in DRAM or especially in DRAM. And then we’re – so we’re still focused on price. But the 39% to 42% guide, the majority of that is driven by favorable mix effects. So more DRAM growth relative NAND and, if we cut by markets, more data center than consumer-oriented mix.

This does conclude the question-and-answer session as well as today’s program. Thank you, ladies and gentlemen, for your participation. You may now disconnect. Good day. Copyright © 2025, S&P Global Market Intelligence. All rights reserved 16