Thank you for standing by, and welcome to Micron's Post Earnings Analyst Call. At this time, all participants are in listen-only mode. After the speaker's remarks, there will be a question and answer session. To ask a question during the session, you'll need to press star one- one on your telephone. To remove yourself from the queue, simply press star one- one again. And now I'd like to introduce your host for today's program, Samir Patodia, Investor Relations. Please go ahead, sir.
Thanks, and welcome to Micron Technology's Fiscal Fourth Quarter 2023 Sell-Side Analyst Call Back. On the call with me today are Sumit Sadana, Micron's Chief Business Officer, Manish Bhatia, our 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, 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 file 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. Jonathan, we can now open up for Q&A.
Certainly. One moment for our first question. Our first question comes from the line of Harlan Sur from JP Morgan. Your question, please.
Hi, good afternoon, everyone. Thanks for taking my question. So within the data center, the team is still anticipating excess customer inventories normalizing beginning of next calendar year. You've got enterprise, cloud, telco, data center markets. Product-wise, you've got DDR4, DDR5, HBM, LPDDR, and then you have your entire enterprise SSD portfolio. So either by end market or product type, like, where are you guys seeing the most weakness in areas of slower inventory drawdown? And then conversely, on the sequential data center growth here in Q1, what are the areas that are driving strength?
Yeah, Harlan, thanks for the question. So Sumit, who I announced will be joining the call, will join us just in a few minutes, so he hasn't arrived yet. We'll have him maybe follow up and answer, add more details of call to your question. But we'll start off with Manish, can maybe start off with an answer.
Mark, do you want to...
Yeah, no, I was just going to say, thanks for the question, Harlan, and thanks for, you know, calling out the fact that, you know, data center is this, you know, is not one amorphous area. There's a lot of different categories with their own, you know, inventory dynamics playing out. I think, you know, I think it's, it's, well known that, you know, the, the, the some of the higher value segments are the ones that are in shorter supply. You know, high bandwidth memory for sure, today, even, and as we, you know, step more into that over our fiscal year 2024 and become part of that, we would still expect that to be a relatively constrained area, of the market, of the data center market.
DDR5, you know, is, you know, inventories are tighter there and more constrained there, even as the industry, you know, continues to adopt more DDR5. It has the crossover points-
Mm-hmm.
For DDR5, you know, for the industry, should be middle of next year. For us, we think it'll be earlier in calendar 2024. And then, some of the LP solutions we've talked about, we've talked about are, you know, somewhat customized, and we have some of those solutions that are unique with some of our customers. In terms of the-
Yep.
And I'd say broadly that the DDR4 is where there's more inventory today, at least on the DRAM side. And then in the, you know, SSD side, you know, I think there are a couple of different segments that are there. There's the classic hyperscaler side of data center, and then there's some of the longer tail side. And you know, we're seeing you know, differing levels of inventory between those customers, but broadly, feeling good about where, you know, both DRAM and NAND are in terms of being able to see, you know, pricing improve in both the markets, including in the data center, as we go through 2024.
Okay, perfect. No, I appreciate the insights there. And then, Mark, can you just true us up on the gross margin impact due to underutilization charges and the associated higher cost inventory? So in other words, like, what's the current quarterly impact from underutilizations here in fiscal Q1, and how do we see that sort of trending going forward?
Yeah, so in Q1, we're going to see the underutilization charges, specifically idle charges, you know, we've been referring to as, you know, the period costs that are idle-
Mm-hmm.
will be about the same in Q1 as we saw in Q4. And it may be helpful, you know, at this point, since we're, you know, on the other side of all this complexity, maybe kind of summarize,
Yes
Y ou know, all the various puts and takes we've seen over the last few quarters. You know, we've had this underutilization, you know, that we've been sort of increasing at 5, then over 20, then, you know, approaching 30 as we sit here today. And so that's driving those underutilization charges that we saw in the fourth quarter and continue into first. And they'll continue through fiscal 2024, but they'll come down, as I've said before, they'll come down gradually and through the year, and in part because as we move to lower wafer output over time, the utilization will increase on that lower wafer output. Now...
You know, on the, on the write-downs and the associated sell-through, you know, we're on the back end of that as well. If you recall, we did the $1.4 billion write-off in second quarter. We did another $400 million write-down in the third quarter. We realized the benefits of the sell-through in the third quarter of under $300 million. So we came into the fourth quarter with about $1.5 billion balance. And in the fourth quarter, we recognized or had the benefit of about $560 million associated with the sell-through of that lower cost inventory. So we enter fiscal 2024 with about a $1 billion lower cost inventory benefit to realize.
We'll realize about $600 million of that in the first quarter, and then the balance, you know, nearly $400 million in the second quarter. And then that, you know, will be, you know, the GAAP and non-GAAP, you know, the results will be sorted out at that point. I think maybe one last thing back to underutilization. You know, I did say, you know, because we've talked about this, and I just want to make sure it's clear in your analysis. I did say that we would have about $1.1 billion dollar impact associated with underutilization in fiscal 2023. And, you know, we recognized about $400 million of that through period costs in fiscal 2023.
You know, we wrote down about, so that leaves $700 million of the $1.1 billion. You know, about $300 million of that $700 million ran through as inventories. So, you know, that's cleared. So that. And then there's another $200 million that actually got written off associated with the inventory write-downs. So we've got, you know, another $200 million left, or $300 million left that, yeah, that, that'll clear in early.
Okay. No, that's very helpful. Thank you.
Thank you. One moment for our next question. Our next question comes to the line of Ambrish Srivastava from BMO Capital Markets. Your question, please?
Hi. Thank you, Mark. It's just confusing. I think you had warned us last time. It's a bit complicated. So the $500 million you just called out, from the $700 million write-down, that's associated with inventory, and that's not part of the inventory write-down, right?
Which $500 million are you talking about?
Well, you just walked us through the $1.1 billion impact in fiscal 2023, which was from the underutilization-
Yes.
Of course, $400 million you took as a period cost.
Yes.
And then the remaining $700 million bridge, you said $300 million with inventory, and I thought you said $200 million also associated with inventory. So that's on-
Yeah.
Three-
Yeah, of the... And I'm talking the $1.1 billion is the-
Yeah
higher cost associated with underutilization.
Yeah.
$400 million of that was, you know, passed through-
Period costs
... as period costs.
Mm-hmm.
$300 million of it has already cleared in inventories, and $200 million of it was associated with the inventory write-down. Okay, so that leaves $300 million left that's going to clear with inventories in early 2024.
Off of the $1.1 billion, right?
Yeah.
So that's just-
Of the $1.1 billion. Exactly.
Okay.
Yeah, that's the $1.1 billion-dollar walk.
Got it. Got it. Got it. Okay, thank you. And then with the fact that we are, the factories are running 30% lower production versus when you started down the path, the underutilization overhang should continue until you come back. I'm just trying to understand the mechanics of what causes that gap to close between underutilization charges being taken and when the factories get back to normal loading.
So, Ambrish, maybe I'll start, and then Mark can cover the accounting mechanics. I'll cover the operational dynamics, and then maybe between the two, it'll give you and others a clear picture. So, you know, as we've said before, you know, with all this underutilized equipment and with the desire to be able to continue to migrate towards newer technology nodes, to be able to support the product portfolio and to, you know, provide better, you know, performance in the products to our customers, we have been utilizing more of the underutilized equipment towards converting to new nodes.
Mm-hmm.
As you do that, you are reducing the capacity baseline of the of each factory where you're, where you're implementing that, right? So the wafer start capability is coming down. But then as you're bringing those tools on and fully utilizing them in the longer process, you know, step counts of the newer nodes, they're not really being underutilized anymore once they become part of the new reset, lower capacity baseline. So as that happens, the amount that's sort of charged off the, towards underutilization or period costs is lowered gradually, and because you're utilizing the equipment towards the newer nodes.
Oh, I'm sorry.
By utilizing the equipment towards the newer node, we don't buy new equipment to be able to maintain a higher production baseline capacity. So it helps us offset CapEx. It helps us utilize the equipment that we have, so the period costs will gradually go down, as Mark has said. But we do end up with lower production capacity at the end of the day, which obviously, you know, isn't as cost-effective as if we would have just maintained the full production capacity while-
Mm-hmm.
We were doing the conversions, but it's more capital efficient to do it this way.
Combination-wise, you answer.
Maybe, maybe just to add on, Ambrish. So, you know, we were just talking about the $1.1 billion, and, you know, generally, the underutilization creates this higher cost per wafer.
Mm-hmm.
And that's the $1.1 billion, and it's when you enter these severe underutilization periods, that portion of that is going to be taken immediately to the P&L versus being hung up in inventory. What's going to happen over time, as the wafer capacity comes down, those, you know, those costs still occur. It's just that as the baseline of our capacity is reset lower, more of those costs then can be absorbed into inventories and will pass through as inventories, or pass through when inventory is clear, versus-
Okay, thank you.
Being recognized as varied costs.
Yep. Thank you. I'll cede the floor. Thank you.
Thank you. One moment for our next question. Our next question comes from the line of Chris Danely from Citi. Your question, please.
Hey, thanks, guys. So I guess, what level of revenue do you need to get to, to crank up utilization rates? And what revenue level do you need to get to, or would you be at, to get back to a full utilization?
Well, go ahead, Manish, why don't you?
I'll just start a little bit operationally, then you can talk about it from the business perspective. So just that, you know, I think of it more as we have to see our inventories come down, right? So, you know, and then we have to see the profitability and the business return to be able to, you know, change the CapEx profile that we've guided to, right? We've talked about WFE being down, and yet, the way we're able to do that while still migrating new technologies, as I was just explaining, is to be able to utilize more of the previously underutilized coming towards that transition to the newer nodes, so it's no longer underutilized. It's contributing positively towards us being able to make those, you know, transitions to the newer nodes.
So it's really more about how we see our inventories coming down that drives our need to be able to produce new supply, and then you know, the profitability of the business as we see pricing improve, then we would look into, you know, being able to invest more, whether in CapEx or other areas, to be able to manage supply or grow supply, more than what we're guiding to right now.
Yeah, I think... I mean, I just want to emphasize one thing that Manish said and connect it back to what Sanjay had said earlier as well, which is, as we have redeployed this equipment from other nodes to expand the map of our newer nodes, there is a structural reduction of our wafer capacity that is happening. This is happening across the industry because everyone's trying to be reducing CapEx in an environment where operational cash flow and profitability has been so dramatically impacted.
So it's an industry-wide thing that is occurring, and the reduction in capacity that is occurring as we do this, every month, every quarter, is going to transition our situation from more underutilization to just reduced amount of wafer capacity, which means there isn't the level of, you know, switching back on of underutilization that's even possible after we have done, we have gone through this transition, except maybe on some legacy nodes, where we will still be running underutilization after we are done with this whole transition. So, so just, you know, wanted to make sure that it's clear that ultimately, since the wafer capacity is coming down, there won't be the switching back on of underutilization to create a lot more bits, especially on the leading edge.
There'll be new CapEx investment required, which of course, will require increased pricing, increased margins, improved cash flow, improved financial performance for us to justify that CapEx investment to increase wafer capacity again.
Sure. And just a clarification on that, guys. Is there an optimal, inventory level in terms of dollars or days you're looking at or trying to get to? And then I'll cede the floor.
Yeah, we're targeting, you know, ultimately, we've said our target is, you know, 120 days, which, you know, if you look at the current, you know, if you just assume current COGS levels, let's say $6 billion or so. You know, we said we'd have, we have about $1 billion of what we call strategic stock on hand. So we've got, you know, $1.5 billion or so of inventories that we need to work through. We are working through those. You've heard a lot of talk today about supply discipline and really our spending and our transitions are driven by the pricing environment and the associated inventory levels in the industry. So we're watching those go down. We're actively managing our inventories down.
We do believe, you know, as we go into the back half of fiscal 2024, that we'll only have a few weeks of, let's call it, above target inventory on hand. And so we think we'll be in an even better place. But you're already seeing it. You know, what we said a long time ago is has transpired, and, you know, that is we said DIO would peak in the second quarter of 2023, which it did, and then the dollars would peak in the third quarter, which they did. And so we're just gonna continue to monitor it and manage it closely.
Yeah, that's helpful. Thanks, guys.
Thank you. One moment for our next question. Our next question comes from the line of Aaron Rakers from Wells Fargo. Your question, please.
Yeah, thanks for taking the questions and doing this follow-up call. I have two quick ones. And Mark, I apologize, I'm just trying to make sure I've got all the numbers correct here. Based on what you said, it looks like your in-period underutilization cost this last quarter is about $240 million to be at that $400 million for the full year. Is that correct? I just wanna clarify that.
You're talking period costs in Q4?
Exactly. Exactly. Yep.
$220 million.
$220 million, okay. And then, and then you've got, it sounds like, you know, $200 million-$300 million more to carry through the first half of the year. I guess my quick follow-up question would be, you know, trying to think about the pace of the recovery, obviously, the focus is on gross margin here. But, you know, I, I'm curious of how you think about the pace of operating expenses coming back into the model. You've done a good job managing operating expenses. As we think about gross margin turning positive, fundamentals improving, how do you think about bringing OpEx in? Or is there a structural thing to consider in OpEx just being structurally lower as we went through this last several quarters?
Well, we see OpEx moving up fourth quarter to first, and that's driven by resumption of certain compensation programs that are normal, that were reduced during the downturn or worst of the downturn. And then we have some timing on R&D-related expenditures. So you know, we've guided from $842 million up to $900 million. And then, you know, the rest of the year, we haven't guided by quarter, but we said that the year would be up slightly, low- single digits. So you do that and, you know, you end up with, you know, call it 1%-2% up.
Yep.
You know, 3.7 or below on the full year. So we're, you know, gonna manage it very tightly. We did have a lot of productivity gains through the year, and we're gonna work our best to, you know, hold those productivity gains.
And then, finally, just one real quick one. Is your impact from China still estimated to be kinda low, you know, low double digits? Is that... You didn't change that estimate.
We did not change that estimate. You know, we're obviously managing it, you know, actively, as you say. The impact has been below that number. It's below that number in the first quarter, but no update.
Yeah, I just want to clarify-
Thank you, guys.
you know, the low- double digit number you quoted is, was the total risk that we had said. So as Mark said, that the impact in China is less than that. And then we have had some success in some mitigation factors, which we are continuing to work on. And, so obviously, there is a net number, net of the mitigation that is, that is lower as well.
Perfect. Thanks, guys.
Thank you. One moment for our next question. Our next question comes from the line of Vivek Arya from Bank of America Securities. Your question, please.
Thanks for taking my questions. I had a few as well. But first, what kind of HBM market share is contemplated in your outlook for next year? And how would you kind of generally characterize the visibility in achieving that? Because I understand, you know, the technology strengths that you have, but when we look at your Korean competitors, they're also, you know, talking up their technology capability and, you know, their ability to hold on to a lot of this market share. So I'm just curious, what is your visibility and confidence, and what kind of, you know, rough market share assumptions do you have in HBM for next year?
Yeah, I mean, we have, we have skipped the HBM3 generation. We commercialized HBM2E, and we went straight to HBM3E in order to, get ahead. And, we are very excited about this technology because the samples that our customers have in their hands have industry-leading, head and shoulders above capability, versus competitors, both from a performance perspective as well as, amazingly good, power consumption capability, well below that of our competitors. And so we are very excited about this. Our customers are very excited, very enthusiastic about this. And so we are now focused on getting the product qualified.
As we have mentioned, you know, we have been embedded now in the lead, as a lead driver of HBM3E in our customers' roadmaps, and we have focused on qualification and the high quality, high reliability ramp, which will start, the production ramp of which will start in early calendar 2024, and then become meaningful in fiscal 2024, and then be pretty big in the back half of calendar 2024. And then from there on, build from there into 2025. And our goal is to ensure that we get our HBM share to be similar to our DRAM supply share, and to get there as soon as the, you know, hard ramp will allow. So that's what we are focused on, and we have, you know, confidence in our plan to execute that.
Right. And for my follow-up, I think, Mark, you mentioned gross margin. You said positive in Q3 or Q4. I just wanted to clarify what was said about gross margin. And how should we think about when EPS turns positive? Is that Q3, Q4, any... I imagine it's more Q4 than Q3, but just what was said so that our expectations are set the right way in terms of gross margins for Q3 and Q4?
Yeah, thanks. So on gross margin, and we—this is consistent with what we said several quarters, we do expect continuing improvement in gross margin quarter to quarter. You know, we guided this -4% down, improved off of 500 basis points off of -9% in the fourth quarter actuals. We do expect continued gross margin improvement in the second quarter relative to the first. We expect pricing to gain momentum in the second half and so expect further improvement on that and just the volume leverage in the business in the second half. And so through the second half, so third and fourth quarters, we expect there to be positive gross margin. And again, that pricing effect will be stronger in gross margin improvement.
As far as profitability, we did not update our profitability. We've said before that we expect to be profitable within the second half of fiscal 2024. And so no update to that, but I do think that it is notable that last time we said that gross margin would be positive in the fourth quarter. This time we're saying that gross margin is positive through the second half. So I think that's an incremental improvement as it relates to profitability.
Okay. Thank you.
Thank you. One moment for our next question. Thank you. Our next question comes from the line of Karl Ackerman from BNP Paribas. Your question, please.
Yes, thank you. Two questions, if I may. First, I suppose it's for Manish or Sumit, but just on the CAC restrictions, you know, you noted that your data center and networking business was impacted by the CAC in the August quarter, but should we assume that current restrictions are impacting your smartphone business in the November quarter?
Yeah. So the CAC impact was meant to be on CII operators, and that means that, you know, largely limited to data center and networking type of products that relate to customers who are deemed to be CII operators. So no, we don't see an impact to the smartphone business. And the impact that we spoke about is in the data center and networking.
Thank you for that clarification. You know, you indicated that test and packaging CapEx could double next year. But I guess, I'm just trying to find a baseline here. I guess, what is your packaging CapEx in fiscal 2023, perhaps both in absolute terms and maybe relative to the size of fiscal 2022? That would be very helpful just to, you know, set some baseline in terms of the investment that you're making relative to the total portion of WFE. Thank you.
I don't think we've broken that out before. You know, I think the color that we've provided, you know, as you think about 2023 going to 2024 on CapEx, is that both construction and assembly tests are going to be increasing meaningfully, while WFE will be down. And so you can imagine that the mix of both assembly and we said that the total CapEx number will be slightly above fiscal year 2023's number. So you can imagine that assembly test, as a percentage, is going up.
We can say generally that it's, you know, typically below 10%, and I believe next year it could be above 10%. I think it's a simple way to frame it, Karl.
Yeah. Good.
Okay.
You know, a significant portion of that increase is to support the HBM ramp, a strong HBM ramp that Sumit was talking about, that we're, you know, that our customers are excited about and that we're, you know, working to put in place.
Thank you.
Thank you. One moment for our next question. Our next question comes from the line of Brian Chin from Stifel. Your question, please.
Hi, thanks for letting us ask a few questions. Maybe kind of doubling back on sort of the output discussion. The die size penalty for HBM3E over DDR5, yeah, 2x, I think you said. It's a bit bigger than I would have expected. What else can you elaborate on that? And more importantly, as a comparison to your peers and the larger footprint. But then, you know, inclusive of that, inclusive of the, you know, the efficiencies and other line kind of conversions, et cetera, how much output do you think you and the peers, or just the industry, could take out next year? It could be something like 10% reduction in the denominator of capacity. What are your thoughts?
Sure. So, maybe I'll just clarify. Are you looking for a comment that's on just the HBM effect, or are you looking for a comment on sort of overall between the multiple effects we discussed? HBM will and DRAM will be a you know will absorb a significant amount of production capacity. And then you have the wafer production capacity that will be reduced based on the conversion to newer nodes.
Yeah
W hich we discussed earlier on this call. Are you looking for a number on both of those, or just on the?
Yeah
J ust on really an HBM question?
Yeah, maybe the all-in, since there's, you know, a couple, a couple different attributions, right? So maybe the all-in, but also maybe the, the amount of that, that's driven by HBM, because it's it should be a, a sizable portion of it as well, the HBM, it sounds like.
Yeah, I mean, I think that the HBM number, you know, just to give you some color, you know, this approximately doubles the die size per bit is kind of there for everyone in the industry. And, you know, it's derived from a couple of different factors. One is just the JEDEC standard for the size of the interface die. It's also in terms of what's needed to be able to get to the high performance of the spec, and it kind of gets more complex for each generation of HBM.
And, and then on top of that, so that's, that's 2x to begin with, and on top of that, you have the impact on, overall production capacity, because we have an interface die that needs to be produced, so a logic die that will take up production capacity in the fabs as well. And then you have the, the more complex back-end integration, actually, overall yields, for the product that are a little bit, lower than standard products, right? Because of the complexity of the package. So when you put all that together, you end up, you know, seeing, you know, more than 2x impact there.
And so you can imagine, you know, depending on, you know, the estimate for, you know, overall HBM percentage in bits for the industry, you can then estimate, you know, more than 2x of that coming out just because of HBM, right?
Mm-hmm.
And then, for the duration of capacity as we convert to newer nodes, you know, it'll vary based on, you know, each supplier. But, you know, what we've said is that we're approaching 30% of our wafer starts to be reduced from their peak 2022 levels. And today, most of that is underutilized capacity. As we look towards the end of fiscal 2024, for us, we're moving to where most of that will be on, you know, true structural wafer capacity reduction, because we have implemented—we've taken that equipment and utilized it in newer nodes.
That just gives you a sense of both of those two, and you know, you can project it for others in the industry and how they're going to take their underutilized equipment, repurpose it, along with whatever their CapEx plans are, to be able to end up with structurally lower wafer capacity. That's the statement both for DRAM and NAND. Obviously, the HBM is only for DRAM.
Oh, okay. And just to clarify that, you're 30% underutilized, and if you did not increase your wafer starts, you could actually approach full utilization rates based on the type of structural, you know, consolidation of the footprint?
Maybe I'll say, I'll say it again, just to make sure I'm clear. As we go through the year, we will be taking the equipment that's currently underutilized and converting it to be used in newer technology nodes. It will no longer be underutilized. It'll be part of a lower wafer start baseline, but with a richer mix of the newer node. Does that help make it clear?
Okay.
Does that make it clear?
Yeah. I think directionally, and maybe the last my follow-up really quickly, then. Just, is the right way to characterize this is kind of gross margins, so maybe Mark, but is the right way to characterize the fiscal one and directional fiscal Q2 gross margin guide, as that your expectations for improved memory prices is upward, but it's somewhat benign? Is that the right way to characterize?
Yeah, I think it's definitively improving from the fourth. And, you know, as we talked about, as the market is transitioning here, there you know, we need to meet the market where it is, and there are some deals that have been done, which, you know, will contribute to maybe a muted first quarter, and then maybe even into the second quarter price upward trajectory. We think it will be up, but how much? You know, prices have been going down for over a year. They're not going to go back up to the levels we want in one quarter. It's going to take a little bit of time. But the market is clearly recognizing that prices are going up.
We believe that the second half will see much stronger price improvement relative to the first. So we think that strength in the second half, and it will, you know, of course, have a favorable effect on the gross margin and the profitability of the business, and cash flows.
Okay. Thank you.
Thank you. One moment for our next question. Our next question comes in the line of Quinn Bolton from Needham & Company. Your question, please.
Hi, guys. Thanks for taking my questions. Wanted to follow up on the question or topic of the net reduction in wafer starts as you reallocate to HBM and more advanced nodes. What role does just improving demand play in that? I mean, are you assuming that, you know, most of that 30% that's underutilized today gets, you know, utilized in these more advanced nodes at constant demand, or are you sort of factoring in some improvement in demand as part of that equation?
So I mean, certainly, you know, we've said that we, we do expect there to be robust growth in DRAM and NAND next year in terms of bit growth, with DRAM bit growth being higher than the long-term CAGR that we've, that we've talked about before, and NAND being around this, that long-term CAGR. So that's certainly part of our plans. But I think what's important is that the demand for the higher value, newer solutions is also growing, maybe even bigger as a part of that growth. So you know, the industry is in the middle of multiple different transitions, whether it's DDR4 to DDR5, LPDDR4 to LPDDR5, HBM2 and HBM3 to HBM3E. All of those newer products are most efficiently produced or only produced on the leading-edge nodes.
So for us, in 1-b eta, for example, great node, strong leadership ahead of, well ahead of the rest of the industry, great power performance. So we've got it in DDR5, we've got it in LPDDR5, we've got it in HBM3E, and eventually, we'll have it in Graphics 7. So, so it's, you know, it's, it's definitely driven by, you know, assumption that demand will return as inventories have, at our customers, have normalized in many areas and will normalize, you know, as we go through the year in, in the other areas. But it's also about the demand mix being more towards high-value products that are, that are best produced on the leading-edge nodes.
And similar for NAND, in terms of the performance specs that we're seeing with our, you know, the 176, higher 176 performance that we're gonna, you know, achieve on our 232-layer. I think great examples on our, some with our UFS product that we've launched, as well as I think the fastest time, certainly in our history, if not the industry's history, from introduction of a node to a data center SSD shipping in volume to an OEM customer. I mean, it just shows the demand is there for higher performance, lower power, and higher quality technologies.
Maybe, Quinn, if I could just add, you know, the downturn here has been. It's lasted now over a year. And, you know, over time, you know, our products, you know, continue to advance. Customers want the best performance, and so we're finding the need for to serve those products, and HBM is one good example. But we are working to have that product capacity for the latest generation, and do it in the most capital efficient way, which is, as we've said, taking equipment from more legacy nodes. And so as a result, we end up with, you know, our wafer starts down over time. The rate and pace of what we do and the supply effects of it are going to be driven by the inventory levels we see in the industry and the level of profitability.
It's sort of a pay-as-you-go sort of approach here in the sense that, you know, unless we, you know, we're going to be constantly looking at the health of inventories and profitability as we make our capacity decisions.
Got it. And then, Mark, my follow-up, you gave us very detailed, your review of the underutilization charges in fiscal 2023 at $1.1 billion, $400 million period cost, $700 million flows through inventory. Do you have an estimate you can provide for fiscal 2024? Because it feels like, you know, you've got $300 million of inventory charges that hit you in early 2024, left over from fiscal 2023. But I assume that there's additional charges, underutilization charges, that are gonna be coming into the model for fiscal year 2024. I just wanna make sure I've got that, you know, thinking about that correctly.
Yeah. So I'm not gonna provide a number because, and I'll explain why briefly. Let's remember that we have period costs, which include the idle costs. So if you go back to the December call, I went through various period costs that we have. We have idle costs front-end, idle costs back-end, and then a whole bunch of other period costs, freight, royalties, and other things, inventory, scrap, and so forth. So we'll always have period costs. They've been elevated because of the underutilization that we had in 2023 and going into 2024. And that underutilization has been so severe that the front end has experienced pretty high what we call idle charges, which go to period costs.
That's part of what I gave you on the walk in 2023 and the $1.1 billion impact. We will have, in the first quarter, another $200 million of idle charges, related to the front end, related to underutilization. And we will have, charges in the third quarter as well, or in the second quarter as well. But over time, those period costs, or idle charges will go down over time, in part because as we mentioned, we're bringing our wafer capacity down, and when we do that, the utilization, just naturally increases.
So, you know, those costs will still exist, but they will go into inventories and then pass through the P&L through inventories rather than directly through period costs as you're seeing now.
Right. So-
And again, I wanna emphasize that the rate and pace of our capacity decisions are going to be driven by, you know, where we see inventory levels and the profitability in the industry. And, you know, we're seeing positive trends relative to last time.
Got it. Okay. Thank you.
Thank you. This does conclude the question and answer session, as well as today's program. Thank you, ladies and gentlemen, for your participation in today's conference. You may now disconnect. Good day.