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

Thank you for standing by. Welcome to Micron's Post Earnings Analyst Call. [Operator Instructions] And now I'd like to introduce your host for today's program, Farhan Ahmad, Vice President, Investor Relations. Please go ahead, sir.

Thank you, and welcome to Micron Technologies Fiscal Fourth Quarter 2022 Analyst Callback. On the call with me today are Sumit Sadana, our Chief Business Officer; Manish Bhatia, our EVP of Global Operations; and Mark Murphy, our CFO. As a reminder, the matters we will be discussing today include forward-looking statements regarding market demand and supply, our expected results and other matters. 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, including our most recent Form 10-K and 10-Q for a discussion of the 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 results to -- sorry, to confirm these statements to actual results. We can now open the call.

[Operator Instructions] Our first question comes from the line of Ambrish Srivastava from BMO.

This is [ Jameson ] calling for Ambrish. So I just had 2 quick questions. The first one is on -- given the fact that you had guided free cash flow to negative $1.5 billion in 1Q. And what seems to be a similar loss in 2Q given the front-end weighted CapEx, similar revenue and profitability profile as well as inventory continue to rise. And then leading to your commentary of free cash flow generation in the back half of the year, can you comment on your expectations for fiscal '23 free cash flow?

Yes. We -- just to maybe reference the script, I believe we stated over $1.5 billion negative free cash flow in the first quarter. We're not guiding the second, though, I did say it would be challenged because we do have still low levels of revenue and income. And then the inventory -- elevated inventory levels. And then as you pointed out, we said that we would have CapEx weighted heavily in the first half. So all those are going to weigh on second quarter free cash flows as well, but we did not provide a number. And then we do expect to return to free cash flow generation in the second half, but I -- we're not providing a full year estimate at this time.

Okay. And then one other question on long-term bit growth, especially for DRAM. I think you guys have lowered your bit growth several times over the last few years, 20% to mid-high teens and now the mid-teens. What would it take for long-term bit growth to be 10%? Is that something that you guys can see over the next several years or next decade happening? Or is that something that is very unlikely, and we should maintain that teens?

Sumit, do you want to take that one?

Yes. I'm sorry, I couldn't catch the question.

I can. The question was about the long-term -- so the question is, we used to say 20% for DRAM, 4, 5 years ago, it came down to mid- to high teens. And today, we're saying mid-teens. So this question is, what would it take for your assumptions to have the long-term CAGR come down to 10%?

Yes, but that's not our assumption.

Yes. I mean I think right now our best projection we have given the strong trends in artificial intelligence and 5G and eventual -- the content increases that will be driven in autos by autonomous. Right now, our outlook is the mid-teens, and that's kind of where we see it now.

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

Yes, sorry about that. I missed my name there. It broke up. I appreciate you taking the question. I have 2 quick questions, if I can, as well. On the first question, Mark, I just want to understand kind of how you're framing the expectations of bit growth. I think in the call you had alluded to an expectation of returning to year-over-year bit growth. It sounded like for both DRAM and NAND into the back half of the second of the fiscal year. I guess the first question to that is, is that the expectation you're currently operating to? And I guess if you believe that you've got to assume a fairly steep north of 20% sequential bit growth expectation in your mind looking into the back half of the fiscal year. Is that how you're thinking about? Is that a fair assumption?

Yes. Aaron, I did not break it out on the call, but you're right in the -- conceptually that in the second half we are expecting a robust recovery and volumes.

And that was DRAM and NAND?

Yes.

And Aaron, I'll just add that if you look -- think about it, we are right now shipping well below end demand. And so the demand on our bit shipments are artificially in some sense low because of the inventory adjustment going in with the customers. Once that is done, it will bounce back. And you can think of it like from where we were in the Q3 level. From that, we have volumes come down almost 1/3, maybe even more than that into the Q1. So at some point, there should be a bounce back from that as the inventory is adjusted.

Yes. I appreciate that. I guess it kind of segues to my second question, which is, I guess your best assessment of the cloud, the server demand profile that you're seeing right now, any views of what inventories are looking like? I think on the call, you alluded to some map set issues, supply chain constraint issues still going on there. But how would you assess what you're seeing from a cloud inventory perspective at this point?

Sumit, are you on?

Yes. I'm here. Can you hear me?

Yes.

Okay. So yes, I can take that question. So yes, I mean, first of all, the demand in the cloud overall is healthy and there is demand out there for servers that just cannot be built because certain components for servers like network interface cards or certain types are still in pretty significant shortage. And so definitely, because the servers cannot be built, the other portions of the supply chain, which have more than adequate inventory are being cut back in terms of demand. So that is one aspect that is impacting the demand even though the end demand in cloud is relatively healthy. Now within cloud, if I were to just look at different segments based on discussions with customers, the portion -- there are portions of the cloud demand that are somewhat impacted by the macroeconomic environment and then there are portions of the cloud demand that are very robust from a demand perspective. So when you net all of that out, the overall demand is relatively healthy. But I mentioned the server issue. The other aspect is memory and storage inventory at cloud customers is generally high and does need to be worked down. So that is part of what's impacting the demand in the near term. Now we do think that the cloud demand is going to remain better than other segments of the market even as the economic environment proceeds along the trajectory that most people expect because there is going to be a desire from companies to find a way to cut costs, to convert capital expenditures into operating expenditures through the use of cloud to defer IT, CapEx and use cloud more and things like that. And so there is definitely going to be an outperformance in terms of cloud demand versus other parts of the market. And then obviously, the whole trend towards digitization of the economy is going to also continue. So we think that this is going to continue to be a good environment. It may have some ups and downs based on the macroeconomic demand, but we are optimistic about how this will play out over time.

[Operator Instructions] Our next question comes from the line of Thomas O'Malley from Barclays.

Mark, just one for you. You talked about some decisive action on the OpEx side of things. Obviously, you're not seeing that flow through right away. It takes some time, but could you just try to lay out the cadence of what kind of -- well, the cadence in the vector of how extreme that could be throughout the year, just as that's a pretty good offset as revenue comes down, just the shape of that OpEx decrease and how much you could see.

Sure, Tom. As with CapEx, we did respond quickly in this unprecedented downturn. We've got clear line of sight on our spend and it's in control, and we're projecting it to come down. I think in the fourth quarter, we actually -- it may not have been noticed, but we are actually below the low end of the guidance range we provided on OpEx. So I think some of that was incentive comp. But I think the actions of the company have already started. We expect to decline sequentially off this $1 billion level that we guided for the first quarter, then we would expect the decline through the year as the actions taken hold and more actions kick in. We would plan to end the year down closer to the levels that we did in FY '22. So I would say $3.8 billion or above. Yes, the actions are wide ranging, headcount free, use attrition, other actions focusing on highest value and highest probability projects for development. So those that are not -- we're taking a harder look at looking at office space, outside services, discretionary spend and so forth. A lot of actions in flight, more coming. And of course, we would adjust our spend as the market conditions warrant. We are clearly trying to balance the short-term challenges in the business with retaining the long-term capabilities of the company and just working through that very carefully.

Really helpful. And then just a follow-up is on the tax rate. You made a comment that it should be $300 million. Were you referring to the total for calendar year '23 being above $300 million? And if that's the case, it's like a 25% tax rate just kind of held consistent through November of '22 through November of '23 kind of the right area? Or does it need to move around a little bit? Just any color there would be helpful.

Yes. So I think what we said, Tom, was the minimum number of $300 million, could be above that, and that's a dollar basis for the full year. And so you can assume that as a floor, it could be a bit more. And I won't give you a percent because that will -- you'll be able to back into full year income. But the -- but it will be a strong double-digit rate. And -- but we expect that to be -- the rate that we're experiencing or really -- expect to experience in FY '23 is really driven by 2 things. One, it's the introduction of this capitalization and amortization of R&D expenditures, and that was a headwind on tax. And then the second thing is just the low levels of profitability. And the way our tax planning works is optimized for higher levels of profitability. And so when you get these lower levels of profitability, you have basically sort of some fixed costs or fixed taxes at these levels. Now as you look out into future years and we return to more normal levels of profitability, we would expect to be in the low to mid-teens. And that increase from FY '22's rate, which was 8%. And there were some, let's say, a normalized rate there is 8% to 10%. That several points of increase as we look forward on a normalized basis is principally driven by that R&D capitalization and amortization effect. And then also just as income grows, the GILTI parts of the U.S. tax reform put some pressure on the rate as more income gets exposed to GILTI.

Super helpful.

And our next question comes from the line of Rajvindra Gill from Needham.

Just a follow-up question on the assumption of returning to year-over-year bit growth in the second half of fiscal '23. Sanjay mentioned some color around smartphone market and kind of the PC market starting to normalize next year. That kind of assumes that the inventory adjustments that the customers will take place. I'm also wondering, are there other assumptions that you have kind of built in with respect to those particular end markets as well as the other end markets to kind of give you confidence that we are going to see kind of this demand inflection point exiting February quarter. Are the inventory levels at the end market level at a sufficiently low level that gives you confidence that they'll start to rebuild? So any color there in terms of that inflection point would be helpful.

Yes. I can take that question. So I think there are a couple of -- there are actually several moving parts in the demand environment. Of course, there is different segments of the market growing at different rates. There is different levels of inventory in the various segments of the market. And the macroeconomic environment is also expected to impact going forward these segments in different ways. And then the other important thing to just keep in mind is, #1, the shipments that we are making currently to customers are below the consumption rate of DRAM and NAND because they are -- obviously customers are trying to reduce their own inventories. Now keep in mind that the PC market and the smartphone market entered the down cycle before the other parts of the market and well before the economic environment broadly started to weaken. And so we also expect that since these are consumer devices that are pretty essential when the volumes of these devices get down to really low levels, we do think they will stabilize. And there is a good chance of improvement in these volumes as countries like China start to open up and forego some of their zero-COVID -- lockdowns related to zero-COVID. We are assuming that those lockdowns will improve as we go through our fiscal year. So sometime in early next calendar year, we are assuming that the Chinese economy will start to improve, and there will be some reinvigoration of consumer demand there. So the last point I will make is the shipments that we are talking about growing in the second half of fiscal 2023 versus the first half of fiscal 2023 have a heavy component of the [Technical Difficulty] that's catching up to what the end demand is because the inventory correction has largely -- we expect would have largely run its course by the end of our first fiscal half of fiscal '23. So by the end of the February quarter, we are expecting that the customer inventories would have been materially improved because they would have been in inventory improvement mode for many months up to that point. And if you think about the PC and smartphone segment, they would have been in that inventory improvement mode for the better half of the year. So I think those are some of the factors. And then the final thing I will mention in terms of the overall health of the industry, although not directly related to demand itself, is that just like right now, the supply is significantly higher than demand in calendar year '22. We expect that situation to reverse early in calendar '23, where we expect that due to the CapEx cuts that we are making and what we expect the industry supply growth to be next year, which is only in the mid-single-digit sort of percentage range for DRAM in terms of supply growth, demand growth being in our model in that mid-teens range. We get a situation next year where the inventory starts to come down sharply because the supply growth has fallen well below the demand growth. And so the situation that you have today, you'll see sort of a mirror image of that sometime next year. And it's obviously based on certain assumptions of the macroeconomic environment. If the environment is better, things will get shifted to the left in terms of improvement of shipments. If the environment gets worse than what we are modeling, then obviously, things will shift to the right. So that's sort of how we're looking at it. I hope that helps.

Yes. That's super helpful to understand those -- those puts and takes. And just for my follow-up, with respect to the cloud server inventory, you made a distinction between some portions of cloud being healthy, regardless of the macro, some portions of the cloud being impacted by the macro. You also mentioned that the cloud inventory remains high, it needs to be worked down. They clearly are kind of are waiting for memory pricing to drop further before they start to kind of rebuild. But I wanted to get a little more insight on the portions of the cloud business that are being impacted because of the macro. Can you elaborate further? Are you seeing kind of an expectation that the hyperscaler customers will reduce infrastructure spending for data center? Or is it just they already have too much inventory and they're going to wait to rebuild until prices come down further?

Yes. I mean I think it's not -- I don't view this as customers are waiting to have prices come down in order to buy. I don't feel like that is the mindset they have. I think what has happened is they had a certain amount of demand that they were buying components for and then they -- then they couldn't put together all the servers they needed to put together, and that was mainly due to things like networking interface cards and so on that was in shortage. And so they found out that, okay, when they can't pull together all the servers they need, they do have built up a high level of inventory of things like DRAM and NAND that obviously they need to slow down the purchase of because they are only able to build a certain number of servers. So that's one aspect of it. I think the other aspect of it is, in terms of the areas that are somewhat impacted, I mean if you think about consumer companies who do leverage the cloud infrastructure for their own purposes of extending into some kind of a hybrid environment or a cloud-only environment, these consumer-facing companies are getting more impacted by the inflationary environment and changing customer -- consumer behavior in terms of spending patterns are generally getting more cautious about the macroeconomic environment and the trajectory of their own businesses. And those are the ones that are a little bit more susceptible in the short term in terms of what's happening to the macroeconomic environment. And the companies that are more B2B type of companies have a little bit less susceptibility. And of course, these are very generalized statements within those sectors, there are some companies that are very strong and some companies that are weak. And so at an aggregate level, I would say that, when you put all of that together, the end demand for cloud is still healthy. But of course, if all of the components have been available to put together servers, it would have been even stronger than that, but that's the situation we are in right now. I do think that it's not so much the putting off of purchases for lower prices. I think once the inventory normalizes at cloud companies, their purchasing will improve of DRAM and NAND somewhat regardless of the end demand because when our shipments fall below their consumption levels, that kind of pattern typically corrects itself after X number of months, but you can [ LIBOR ] that expense just based on how much inventory each customer has.

And our next question comes from the line of Sidney Ho from Deutsche Bank.

So more on inventory, about your own inventory days, it sounds like that is going to go up in the November quarter. Should we be thinking that is the peak level in terms of either dollars or days? And at what point should we be worried about you have to write down that the value of the inventory? Is it when gross margin turns negative? And do you expect that to happen in the next few quarters, especially given -- especially on the NAND side?

Yes. Sidney, so we do expect inventory levels to go up substantially in the first quarter and days go well over 150. We expect it to stay sort of at those levels into the second quarter and then begin to come down as volumes pick up in the back half of the year and -- but remain above our target level for sure through '23. In fact, stay probably elevated at that 150-plus through the year. Yes, we -- though prices have come down, and we have price assumptions in our outlook, we have a robust process internally to monitor the potential exposure on inventories. We review sales volumes, customer orders, contract prices, supply-demand signals, seasonal factors, other trends to evaluate the risk. But based on our assessment, as we see in our outlook, we project margins at such levels that we do not see currently any write-downs. However, we'll obviously continue to monitor things...

Due to lower...

Yes, due to lower profit. We always, of course, have inventory adjustments for various reasons that occur every quarter. But the lower cost of market that you're talking about, we do -- so I think that's where we are.

Okay. Maybe a quick one on the capital spending side. I know you've taken it down to $8 billion. You normally talk about how much of the CapEx is dedicated DRAM versus NAND versus back end. I wonder if you can give us a breakdown of that. And within the WFE cut, you talk about 50%, are you cutting more on the DRAM side or NAND side or is about the same?

Sure. So I can take that one, Sidney. I'll take the second question first. We're cutting on both, the reductions on both -- and what I'll just emphasize is that all of the WFE that we're spending is really going to next-generation technologies, meaning in DRAM, one beta and beyond, including even some early spending on 1-gamma. And then on the NAND side, 232 and beyond, even though we're slowing down those ramps, still all -- whatever spending we do have in WFE is really going towards the future. And whenever you get in a situation like this, you always have to make a choice. And typically, what you do is sort of stop spending on the existing technologies because that's capacity growth and you pivot your spending towards the future generations so that you're prepared when things reverse, that we're going to have leading-edge technology qualified. It's going to -- we're going to have multiple different products, qualified good yields. And that's sort of what we're investing for now as well as generation even beyond those 2. In terms of just giving you some rough breakdown of the overall $8 billion, I think we can say that the WFE is the largest part. It's more than half. We then have construction spending, which is actually, as we said in the script, above what we had -- what the levels were in '22. Construction spending, as you know, tends to be lumpy. It's sort of you build cleanroom space. We have been very proactive over the last several years before last year in building cleanroom space, if you go back to our years in '19 and '20, in '21 -- so '22 was a lower year. This year we have more spending. And inclusive of that spending is the spending for the U.S. project that we announced in Boise earlier this month. And then we have spending for technology and development as well as for assembly and tests that are sort of the next categories.

Okay. So maybe, Sidney, just to add to something that Manish was covering there. So I think it's noteworthy as he said that there's a sizable construction part of this spend. And as we cut levels, that becomes a higher percent. That actually is a higher percent in the back half of the year. And I just think it's worth calling out because these are investments for the long term, and we're going to begin spending monies around Idaho, for example, and that is for end of the decade later in the decade demand. And I just think it's important to call that out because we're certainly exploring ways to finance those expenditures. And there's certainly government grants that we are working through that chipset, for example, and the ITC benefits of that legislation. We would use, of course, as the business improves our own operating cash flow, our balance sheet and then other means to finance those expenditures.

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

This is Eddy for Krish. Thanks for squeezing me in. At your Analyst Day, you gave us a 50% target mix of DDR5 in servers by -- to Q4 2023. Obviously, this will be lower, but can you share your updated thoughts on that? And I have another question, please.

Yes. I mean definitely, the DDR5 crossover point has been pushed out largely due to the pushout of the schedule of certain new architecture introductions by the CPU vendors. And so as we look at the DDR5 ramp, it is continuing on the DC side. But on the server side, the numbers are far lower in 2022 as we look towards the end of '22, far lower than we had originally expected or the industry had originally expected. And we expect that by the end of 2023, the DDR5 penetration in the server space will be meaningfully below 50%, we'll get to 50% levels sometime in 2024.

Great. And on cost declines, how should we think about the consequence for fiscal '23 across DRAM and NAND. That's it for me.

Sure. I can take that. I mean, both of them are going to be -- are going to be below the long-term CAGR this year, in fiscal year -- actually to this year. In fiscal year '22, last year, we performed very well. Both NAND and DRAM at the memory level were above -- the cost reductions were better than the long-term CAGR that we have for both -- for DRAM, which we said is high single digits, and for NAND, low teens. So we did better than that in '22. In '23, we expect both DRAM and NAND to be below that. Obviously, the reduced capital spending and slower technology transitions have a big part in that. The underutilization has a part in that, Mark gave some color on the call before about that part. We have some inflationary headwinds, particularly in Singapore. On the NAND side, we have an inflationary headwind, it's unique there given what's going on with the electricity markets there, and that's a challenge for us. DRAM does have some benefit on FX, given we manufacture in Japan and Taiwan and both the yen and the Taiwan dollar have weakened recently. So we have some things there. But when you put all that together, we still expect DRAM and NAND both to be below the long-term CAGR for memory level cost reduction in fiscal year '23 and NAND to be more challenged than DRAM.

And will it be lower than fiscal 2022?

Yes. Both of them will be below the long-term CAGR, so therefore, below -- let me make sure -- I should -- below -- they will not reach the long-term CAGR. They will be worse than fiscal year '22, and there will be less cost reduction or shallower cost reduction than the long-term CAGR with NAND more challenged to DRAM.

And our next question comes from the line of Toshiya Hari from Goldman Sachs.

I had a couple of questions on the demand side. As your expectations have come down over the past couple of months, and this obviously isn't just you guys, but the overall environment has come in. Which end market or application, I guess, has surprised you the most? It sounds like everything is worse, but if you can differentiate between the big core end markets, that would be helpful. And then related to that, you talked to elevated customer inventory in cloud, but is there a significant difference in customer inventory levels across the end markets vis-a-vis what you consider to be normal?

Yes, yes, that's a good question. I think in terms of what surprised us the most -- I mean, the smartphone and PC segment first one to weaken and they weakened sharply early in the cycle. And we've started seeing inventory-related corrections then spread to other segments of the market. And then some segments of the market started to see some end demand erosion alongside inventory correction and that sort of compounded the demand issue. So in terms of the pullback of actual purchases, it has been pretty broad-based across all of the segments due to a combination of inventory and end-demand concerns. In some segments, the end demand concerns like cloud, I mentioned, end demand still healthy, but inventory and server issues. But generally, the pullback has been pretty significant across multiple segments. I would say that the end demand in automotive still remains in very good shape as well. The pullback in demand in areas like automotive, despite the inventory correction is less pronounced compared to the overall pullback in demand in other areas. But pretty much all segments have been impacted quite a bit. And I think part of this is also coming from the concern, a broadening level of concern among customers about the trajectory of the global economy and the U.S. economy and just driving that cautionary behavior of needing to manage or wanting to manage free cash flow and inventory levels in a tighter way than they would have then they were in a mode that semiconductor parts are generally in shortage for the last couple of years, and the mindset has been to just accumulate parts and not [ gauge ] revenue, and now the mindset is more to protect free cash flow and keep healthier levels -- lower levels of inventory. I think in terms of -- yes, I was just going to the other part of your question regarding cloud. I mean I think in terms of cloud, the only differentiation I wanted to make is the enterprise versus the cloud part of the data center -- overall data center market. The enterprise being more of the traditional on-prem type of OEM-driven purchasing. That has been weaker than the cloud trends, generally speaking, in terms of differentiating between enterprise and cloud. And I think that at this point, I would not say that inventories are at normal levels in any major segment. I would say that there is some level of excess inventory in most segments, a small amount in some, but -- and more in others, but there are definitely customers who have normal levels of inventory within each segment. There may be customers who have more normal levels of inventory. But if I look at it from an entire segment level, not easy to identify one that there is -- entire segment is healthy from an inventory perspective at this point. But I do think that -- you see that in our shipment numbers in F Q4, our bit shipment declined in F Q1, our bit shipment declined. Small bit shipment declines do happen from time to time. It's very normal on a sequential basis. But bit shipments, having 2 quarters of back-to-back declines in a pretty substantial way, typically only happens when inventory is being liquidated. So you can see that inventory is going down at customers between Q4 and Q1 of our fiscal year. And we expect those inventories to keep going down through the rest of this calendar year and into the early part of calendar '23, at which point we think they would have reached levels where the ship-in volume should start to rebound to meet the ship-out volume levels. And of course, all of this is based on assumptions of certain macroeconomic environment that obviously is ahead of us.

Great. As a follow-up on the supply side, so you're cutting WFE by 50% in fiscal '23. If -- for whatever reason the environment ends up being a lot better or less bad than expected, say, in the second half of your fiscal year or your expectations for the early part of fiscal '24 improve, how flexible can you be with that CapEx with your equipment suppliers? Meaning, can you go back to them and say, we actually don't want to cut this much, should we need tools or just given where lead times are, would that be difficult? I guess if you can compare and contrast the flexibility you have today versus past cyclical inflections, that would be helpful.

Sure, Toshiya. And that's a nice thing to think about -- to try to think about it [ how did you ] like today. So thank you for giving [indiscernible]. So look, we have a great relationship with the equipment vendors, some of it over multiple decades of all the vendors really whether in the U.S. or Japan or Europe. And what I'll tell you is that when they were having supply chain challenges over the last couple of years, we really partnered with them closely to help them get their -- figure out how to keep their shipments up to meet our schedule, doing things like qualifying alternate sources for certain parts so they could ship equipment or even doing the installations in place. So we've worked with them really well, and I think they appreciate that partnership, and they're working with us now as we have this difficult period. And as I mentioned, we are maintaining the manufacturing corridors for the ramp-up of these new technologies to get multiple products qualified across them to improve yield and make sure that when the demand turns around, we'll have the ability to ramp those leading-edge -- those leading-edge nodes. And in terms of the flexibility that we'll have, where we've exhibited flexibility with them to try and help increase or improve and keep our ramps on pace in the past, this is obviously a different period, but we can certainly go back to the playbook we had before if we need to expedite shipments again. And certainly, if we did, it would be for leading-edge technology that we would. But I don't think that it's something where we're going to do it in line with the demand trends and demand environment improving. And for now, we're kind of -- we're slowing the ramps down through '23 and looking at '24, where we'll get the majority of the impact from these 2 new technologies, 1 beta and 232-layer.

And our next question comes from the line of Hans Mosesmann from Rosenblatt Securities.

Can you hear me okay?

Yes, we can hear you. Go ahead.

Great. And just a clarification. There's a previous question on DDR5 and I think management indicated that the DDR5 transition and/or crossover had been delayed during the analyst meeting. But now I think that the answer has been that the crossover is going to be well below 50% by the end of 2023. Is that accurate? Or did I misunderstand?

Yes. That is what we said, yes.

And that's a change from what has been implied during the analyst meeting earlier this year.

I mean the DDR5 transition in the server market is heavily driven by the timing of the platform rollout from the CPU companies because the DDR5 meets and can attach to only certain server platforms. And it also then depends on the rate and pace at which those platforms will be deployed. If the deployment of these platforms is faster than we expect, and of course, we'll be able to take advantage of that because we have an industry-leading DDR portfolio, and we are -- we have a share position in DDR5 that's actually higher than our supply share. So we are in a really good position with DDR5. And regardless of what the timing of these ramps and DDR5 are, we are in a great position to take advantage of it. Our current expectation is that, yes, the 50% point would not have reached by the end of calendar '23 as we had earlier expected.

And what was the early expectation? Just to kind of confirm it, the earlier expectation was middle of 2023.

No, no, no, it was never middle. It was more like it was going to be towards the end, like fourth quarter of '23. Now it's pushed into '24.

Our final question comes from the line of Steven Fox from Fox Advisors.

I just had one question. You guys have given a lot of detail on your thoughts on inventories and demand by different segments. I guess the one overall arching question I had, especially since Sanjay called out China as being important. How much influence do you now looking back, I think that excess purchases out of China influence this whole cycle up and down. And how sensitive do you think you're going to be to what's going on in China on any sort of recovery?

I think in terms of China, we do have -- we do have YMTC selling NAND to customers for use in China, mostly into lower-end consumer applications. We do have a model of the industry TAM that incorporates certain levels of growth of supply from Chinese suppliers. So that's part of how we look at the world in terms of when you think about supply and demand and balance and so on. And there is no doubt that particularly in NAND, but to a lower extent in DRAM, but particularly NAND, there has been supply from YMTC that has added to the supply-demand imbalance that is currently in the market. But we do expect that the overall supply growth next year even in NAND will fall meaningfully below the demand growth and the environment to consequently improve once the inventories are in better shape, as we have described in the past.

This does conclude the question-and-answer session as well as today's program. Thank you for your participation. You may now disconnect. Good day.