Great. Good morning, thank you for attending JP Morgan's 21st Annual Semiconductor Technology and Automotive Forum here at the Consumer Electronics Show. My name is Harlan Sur, I'm the Semiconductor and Semiconductor Capital Equipment Analyst for the firm. I'm very pleased to have Mark Murphy, Chief Financial Officer, and Sumit Sadana, Executive Vice President and Chief Business Officer at Micron here with us today. I've asked Mark to briefly summarize the company's November quarter results and outlook, which they reported just two weeks ago, then we'll dive into the Q&A. Gentlemen, thank you for joining us today. Mark, let me go ahead and turn it over to you.
Okay. Thank you, Harlan. Thank you all for joining us today. I'll start with safe harbor. We will be making forward-looking statements. Those statements have risks and uncertainties associated with them. I refer you to our risk factors, which you can find in our public filings. As Harlan mentioned, we did just recently report on December 21st, our November quarter, which was our first fiscal quarter. The quarter was characterized or impacted, I should say, by lower pricing trends through the quarter than we had anticipated. As a result, we ended up printing results at the lower end of our range. It reflects, you know, the market conditions remain challenging. Customers continue to adjust inventory levels. Certain end markets are weak, especially consumer.
Just the, you know, the overall macro environment is weighing on the market conditions. We've taken decisive action in a number of ways. We've got additional enterprise-wide cost reductions that are underway, and you heard those details on the call if you had an opportunity to listen to the call. Furthermore, on the supply side, we have reduced utilization 20% from our peak levels in order to help achieve supply-demand balance in the industry. Also, we've reduced CapEx further. We now have our view on WFE for fiscal year is down over 50% versus last year.
As difficult as the market conditions are, we do see bits having bottomed in the November quarter, and we see volume strengthening through the year as customer inventories improve, and they begin to re-replenish their inventory levels. We would expect customer inventories to be in a better place by the second calendar quarter. The second half revenue, we do expect to be stronger than the first half fiscal revenue. you know, that would lead to, you know, just, you know, again, the strengthening volume, or driven by the strengthening volume. However, you know, given the market conditions and also, some of the cost effects of the underutilization, margins will be challenged in the second half relative to the first half.
you know, Micron's navigating these difficult, you know, this difficult environment with technology leadership and product leadership and world-class operations and a very strong balance sheet. you know, over the long term, you know, we're fortunate that we're serving markets that are supported by strong secular growth drivers, and expect that growth and a better supply-demand balance in the industry to yield a strong profitability and cash flows in the future.
Great. Perfect. I'll kick it off with the first few questions and turn it over to the audience. What is the team monitoring or seeing from customers that gives the team confidence on supply side that upstream inventories are trending the right way? On the demand side, you know, what are your assumptions for end market dynamics that are gonna help further drive the inventory normalization and subsequent revenue inflection, right. In other words, you know, what are your expectations for PC, smartphone, data center, and embedded market trends in a weaker global macro environment for this year?
I'll take that, Harlan. In terms of the overall dynamic, of course, all of us can see that the macroeconomic environment is challenging and quite uncertain. We have taken some prudent reductions in our estimates for forward-looking growth in calendar year 2023 due to the macro environment. I will state that the environment continues to be not easy to predict, as all of you have seen.
Yeah
... over the course of several quarters. Our approach is that, you know, we are very transparent with our assumptions with all of you and, you can, you know, assess and judge whether our own assumptions for end markets are reasonable compared to your own assessments. As we look at modeling of 2023 demand and talk to customers, talk to different players in the ecosystem and use many different data points to create our view. Our assessment is that the PC units will decline in 2023 calendar year from 2022 in that around mid-single digit levels, get it down closer to where the PC units worldwide were in 2019.
Mm-hmm.
Pre-COVID. All of that increase that happened in 2020 and 2021 would have been unbound through that 2022 and 2023 decline in PC units. Of course, the decline in 2022 has been much more significant compared to this assumption in 2023.
Right.
In terms of smartphones, there's obviously a lot of uncertainty there as well, particularly in terms of determining the trajectory of consumer demand. We have modeled flat to slightly up smartphone unit demand and a continuing increase of 5G penetration in terms of units as a percent of the total smartphone volume. Obviously, this requires second half demand in smartphones in the calendar year 2023 to be stronger and some benefit coming from the reopening of the economy that we have been expecting in China.
When we originally made this assessment way back internally in November and early December and then communicated to all of you, since then, obviously certain things have transpired in China that all of you have seen, a lot of the rollback of COVID restrictions and the country is going through a tough time right now because of all of that. Our expectation and hope is that the country will come out of it, China will come out of it, just like the rest of the world has over the next few months. Second half of the calendar year with government incentives and stimulus, et cetera, should do better, and that is the basis of our smartphone forecast.
Now on the data center side, yes, there is definitely continued changes happening under the covers with new server platforms being deployed in 2023, DDR5 being deployed in 2023. There is also considerable things on the macroeconomic front that our customers are dealing with there. There has been a reduction in the growth in data center for 2023 due to some of these end market challenges. I think whereas segments like PCs and smartphones have entered the inventory correction earlier in calendar 2022, data center has been relatively late in entering that inventory correction.
In terms of purchased, DRAM and NAND bits in 2023 calendar year, there is a pretty significant impact coming in the data center due to a fairly high level of inventory of, DRAM and NAND in data center customers entering 2023 calendar year. That's the part that's weighing on the demand and offsetting some of the underlying demand, that otherwise would be there. Having said that, I will point out that the underlying demand itself is weaker, than it has been in the past, and we have been, cautious about the trajectory of that demand for some time now, as you have heard from our public comments. Some of that concern is coming to fruition with data center demand weakening, combined with the higher inventory entering that year.
That's all part of the mix as to why we have felt that, yes, the inventory will improve.
Yeah.
We are seeing customer demand and purchases of DRAM and NAND fall further below their consumption levels and their ship out levels. We have a feeling that we have a reason to believe that the inventory at customers is on a trend to improve. It's not perhaps improving as fast as we would have liked, mainly because, you know, customers have had LTAs. They are under a lot of pressure to minimize the gap to LTAs or reduce some of the gap versus what they had originally planned to take. Of course, the pricing is pretty attractive, so they are also using the pricing to perhaps purchase more than they would have otherwise purchased given their inventory levels.
Right.
That's stretching out some of the timeline of the normalization of inventory at customers. We do think that by middle of calendar 2023, by that June-ish timeframe, we should have an environment where customer inventories are in much better shape than they are now and than they have been the past few months. That's our base view.
You know, the team is reacting exactly as we would have expected from a supply discipline, you know, cutting utilizations, cutting WFE spend, holding back bit shipments on current output, and slowing the pace of technology migration. Your industry assumption, you know, on supply and overall supply-demand normalization assumes that all of the major memory competitors act in a similar fashion, right? Talking with investors, you know, there are concerns that one of your large competitors may or may not be acting in a disciplined fashion. Obviously, your visibility as to what your competitors are doing, you know, is somewhat limited, but you are monitoring aggregate excess inventories at customers across various end markets, and the slope of the drawdown should be reflective of, you know, what you and your competitors are doing to a certain degree.
Based on the metrics that you monitor in real time in your end markets and with your large customers, are you seeing broad supply discipline by competitors and any differences between the DRAM and NAND segments of the market?
Yeah, I would say that, definitely the NAND segment has more challenges in terms of supply-demand imbalance compared to DRAM. You know, part of it is driven by, there are more, suppliers in NAND as well compared to DRAM. I think, you know, I can't speak to other competitors.
Right
... plans, but definitely, you know, there is no doubt that part of the reason that we have articulated in our earnings a couple of weeks ago that we expect a challenging 2023 calendar year is because, even though, as Mark said, you know, we do expect bit growth to happen, is because of the supply-demand imbalance that we think needs to be improved. Micron has taken extremely decisive steps, as you have heard from us over the last few months to reduce our investment in CapEx to curtail our bit growth. We think our DRAM bit growth is going to be negative on a year-over-year percent growth basis in calendar 2023 versus 2022. We think that DRAM bit growth needs to shrink year on year in order to improve the industry environment.
You know, we continue to watch for signs of, you know, what's happening with the industry supply-demand balance, continue to work with our customers. That's what we think is necessary to accelerate the improvement in the health of the industry, and we have definitely taken decisive action to do our part.
You know, the team has cut wafer starts by 20% versus the beginning of last year. You mentioned your intention to continue to keep utilizations low potentially into FY 2024. From a financial impact perspective, you know, underutilization charges forecasted to be $460 million in the second half of the fiscal year, with more of that weighting towards, you know, Q3, more weighting towards Q4. By our estimate, that's like a 4%, you know, gross margin impact in Q3, 6%, 7% impact in fiscal Q4. I believe you mentioned you still have another $400 million or so of these charges that will be recognized in FY 2024. If you do maintain utilizations below your capacity potential beyond FY 2023, how do we think about utilization charges extending into not just FY 2024, but calendar year 2024?
We've not provided any additional comments on the duration of the underutilization period. As we strive to, you know, deal with supply and demand imbalance, we'll continuously reassess that. As you said, you know, our current plans addressing this fiscal year, we will have higher cost inventories that more than half of that will pass through the income statement in the third and fourth quarter.
Yeah.
As mentioned, that will weigh on margins, as you said, in the back half. Less than half of the higher cost inventories related to FY 2023 underutilization will pass through, the first half of fiscal 2024.
Right.
If volumes are stronger than our current forecast, then more of those underutilization costs will be pulled into the back half of fiscal 2023. If volumes are less than we think, then, you know, more of those costs will be passed through to, you know, fiscal 2024. You know, if we do, you know, if the supply-demand balance does not improve and we have a more prolonged period of underutilization, then of course we'll have higher cost inventories, which then will affect FY 2024 and potentially, as you mentioned, depending on how long that inventory has to clear, and it could, you know, it'll just be a function of how long and then how long it takes that inventory to clear.
Longer term, the team talked about a slowing of technology migration cadence in both DRAM and NAND, given lower assumptions around DRAM and NAND bit demand growth. The first question is, if you slow tech migration cadence, doesn't this slow your cost per bit profile and your ability to drive through cycle profitability improvements? Secondly, you know, what are the demand assumptions that cause you to revise the long-term bit demand profiles lower? It doesn't really seem like assumptions pre-COVID-19 have changed all that much in terms of content growth and unit TAM growth in your various end markets.
Yeah. Addressing the second part first, yes, I mean, the, You know, pre-COVID, 2019 timeframe, now we are in 2023. A lot has changed in the world, and, part of what had transpired over the last few years was a view that, you know, the PC volumes are going to sustainably be higher than pre-COVID levels. Where those volumes ultimately settle remains to be seen. There'll be, at some point, some kind of an upgrade cycle for all the PCs that were bought, but typically, you know, those things are stretched out for several years in terms of upgrade cycles. We'll see when that happens.
Certainly as I mentioned earlier, the PC volumes have come down, and we have lowered our views of PC volumes as well as smartphone volumes in the longer-term view because the upgrade cycle in smartphones has continued to stretch every year. For the last few years, the upgrade cycle has continued to elongate. We have appropriately modeled that in our forward-looking views. The cloud is substantially bigger than what it was back in 2019 in the pre-COVID days. We have somewhat moderated longer-term cloud growth rates based on that as well. Those are some of the factors.
Yeah.
-driving the lower long-term growth rate assumptions for both DRAM and NAND bits. With regard to cost reductions, yes, absolutely, when tech transitions move out in terms of longer cadences-
Right.
cost reductions do become more shallow because of that. We believe, number one, it's an industry-wide phenomena. It's not a Micron-specific issue. There is a strong linkage between cost reductions, time of technology transitions, CapEx investments and pricing and margins-
Mm-hmm.
as you know very well.
Right.
Our view is that if cost reductions industry-wide become shallower, then pricing will consequently also become similarly impacted in the same way. If cost reductions are shallower, pricing declines will be shallower over time. That in itself should not cause any changes to the margin expectations of the industry. Really, it's a matter of, you know, the supply-demand balance and the supply discipline in the industry, at an industry level, and that ought to be the primary determinant of longer-term sustained margins.
You know, the team has had a very strong technology leadership position in both DRAM and NAND, right? First to market with 1-Beta DRAM, first to market with 200+ layer NAND. In response to the supply-demand challenges that we just talked about, you know, you are slowing the ramp of 1-Beta, you are slowing the ramp of 232-layer NAND, and it's, you know, pushing out some of the future roadmaps. You know, bears would argue that competitors will use this down cycle as an opportunity to close the technology, you know, gap with Micron. How does the team retain its technology leadership given the near-term slowing of tech migrations? Do you start to focus on more product differentiation, system-level differentiation, higher value add, or, you know, what are the things that you're gonna do to maintain this technology edge?
Yeah, great question. You know, I'll just point out that a couple of things. Number one, we have been very focused as a company on both process technology leadership as well as product portfolio leadership.
Mm-hmm. Mm-hmm.
You have seen fairly substantial transformation in the Micron product portfolio over the last five years, even as we have gone through the technology innovation curve and taken a leadership position on the technology side. For example, we lead the industry in QLC penetration on NAND. We are the first to the market with 176-layer data center SSDs, and are already have several qualifications completed, many more in process of being completed over the course of the next several months. Best number one share in the industry in DDR5, world's fastest graphics memory.
Yeah.
I mean, there's a lot of data points that we can go through that demonstrate our leadership on the product side while we have taken technology leadership in both DRAM and NAND in the industry and lead the industry by several quarters and ramping volume in both of the leading-edge nodes in DRAM and NAND. That has been a consistent strategy that will continue. In terms of this whole cadence, I again want to emphasize that, you know, a lot of that is driven by our expectations of demand bit growth, and every new node of technology creates a certain amount of increase in gigabits or gigabytes per wafer, as the case may be between DRAM and NAND.
If you use a model of a certain amount of demand bit growth in the industry, it almost dictates for a given amount of bit growth per wafer in going from one tech node to the other, that the cadence has to be aligned to the demand bit growth that you're trying to satisfy in order to keep demand and supply balanced. That's why we strongly believe this is an industry-wide phenomena and our relaxation or slowing down of the cadence of technology needs to really be something that we believe ultimately will happen in the industry just because the laws of physics and math work the same way in every company.
Before I jump into some of the product and end market segments, just wanted to see if anybody has any questions. Let's talk about the data center segment of your business. You know, the team's view on cloud data center dynamics, as you mentioned in your opening remarks, was probably one of the end markets where the memory demand dynamics changed pretty meaningfully relative to your prior views, right? Seems like the level of inventories of cloud and data center value chain is carrying was much higher versus your expectation. You know, obviously, cloud demand has moderated from the high growth rates of the past couple of years, but CapEx spend on by the cloud guys is still growing, you know, 6%-8% forecasted for this year. You've got new server processor RAM.
There's still strong spending on accelerated compute initiatives like AI and data analytics, and these clusters, like, suck up a ton of memory. Is more of the inventory issue your customers wanting to just maintain a leaner inventory profile, or is it really more a change in their compute and memory infrastructure sort of deployment plans?
I think there are a couple of factors, that I'll just very briefly touch upon. One is that, definitely from expectations of six months ago, the demand environment for data center customers has deteriorated for 2023, and that is partly reflected in our overall view of what'll happen in this segment in 2023 calendar year. The other aspect is, yes, indeed, I mean, the inventory ending calendar 2022, or as the customers, you know, started planning for calendar 2023, has been substantially higher in DRAM and NAND than what was our view, as well as what was broadly communicated by our customers to suppliers.
That view changed quite a bit over the course of the last several months, both as a combination of, you know, what they were buying in 2022 as well as their view of demand. There is no doubt that there are a lot of good demand drivers in the data center that long-term will be very positive for demand for memory, both DRAM and NAND. These include, you know, obviously, these DDR5 platforms that are, you know, certainly more expensive platforms to deploy. They'll take more dollars. Of course, there is increased content because these newer CPUs and GPUs have a lot more cores.
Mm-hmm.
Bandwidth per core is an important driver of system-level performance. A lot of the workloads are memory bound rather than compute bound alone. Increasing a lot of capability on the CPU or GPU side without increasing the amount of memory that's connected is not going to be terribly beneficial to system-level performance. The underlying demand trends are really solid long term. That's also the reason why CXL ultimately will connect to a lot more memory than what the DDR channels are able to connect to today, and it'll increase system-level average memory per server quite substantially over time. All of those things are great positive that'll play out over time.
CXL is in its super early stages, will take time to deploy. We are talking really about a, you know, 2023 calendar year where we still have to go through the inventory adjustment and all the other factors I mentioned. Once we get through that, you know, looking ahead, 2024, 2025, 2026, et cetera, things should continue to improve, from a demand trajectory perspective in the data center for sure.
You know, in terms of technology enablement, you know, I think there's a misconception in the market that EUV is new for Micron, right? I know the team has been working on EUV for many, many years now. In fact, you've had an EUV system in your Boise R&D fab for quite some time. You're getting ready to launch EUV into your next gen 1-Gamma technology that's targeted to RAM 2025. There's always a learning curve with new technology adoption. Has the EUV process, the module, the architecture been fixed? Is the team still optimizing it? If you are still optimizing, you know, what are the big challenges? Is it uptime, reliability, throughput? You know, would be great to get an update from the team here.
Sure. Yeah. In terms of EUV, like you said, Harlan, we have had EUV...
Mm-hmm.
In our R&D for many years, and we have really great improving performance of those tools compared to prior generation tools that existed. That tool is performing well. We continue to make great progress, and the capability of the EUV tools has substantially improved over the last couple of years compared to what it used to be. We feel really good about the timing of the introduction of EUV into our roadmap. We are super glad that we didn't rush to introduce this any sooner than we are, because that would not have been a very cost-effective way to do so.
Of course, you know, for something that's going to be introduced into volume manufacturing in 2025, we are just starting in 2023 now, so there is a considerable amount of development still to come. We continue to make progress with EUV as we have been needing to and wanting to. We have definitely areas focused on lithography, where we talk about, you know, look at engineering resources focused on pattern fidelity, leading edge features and the consistency of those leading edge features when we won't ramp a particular product into volume. Those are all kinds of things that are very natural for this stage of the development of a new technology node.
A lot of those issues have to continue to be worked on right up to the point we get to mature yields and mature levels of defectivity. That is an ongoing, you know, business as usual kind of thing we do. Yes, I mean, EUV continues to be on track to support that 2025 deployment.
We're just out of time, but I wanna ask one last question, shifting to the cross-cycle financial model that you guys laid out at your Analyst Day in May. You know, from a profitability perspective, op margins, 30%, EBITDA margins, low 50%, free cash flow margins of greater than 10%. With the current memory downturn in full force and a revision of the longer term sort of bit demand profile for certain of your end markets, what is the confidence level by the team at achieving these cross-cycle targets?
We think that the updated market view of volume growth does not change our revenue outlook. As mentioned earlier, the slower volume growth does in part result in less price erosion.
Right.
That helps offset, so no change there. Then we do expect when supply-demand balance improves that through cycle, we can deliver the strong profitability and free cash flow that we've outlined in our model.
Good. Look forward to monitoring the progress and execution of the team this year. Thanks, Sumit. Thanks, Mark. Appreciate your participation.
Thank you for having me, Harlan.
Yeah.
Thank you.