Good afternoon. This is the Chorus Call conference operator. Welcome, and thank you for joining the ASM International Third Quarter 2024 earnings call. As a reminder, all participants are in listen-only mode. After the presentation, there will be an opportunity to ask questions. Should anyone need assistance during the conference call, they may signal an operator by pressing star and zero on their telephone. At this time, I would like to turn the conference over to Mr. Victor Bareño, Head of Investor Relations. Please go ahead, sir.
Thank you, Operator. Good afternoon and welcome, everyone, to our Q3 earnings call. I'm joined here today by our CEO, Hichem M'Saad, and our CFO, Paul Verhagen. ASM issued its Third Quarter 2024 results yesterday at 6:00 P.M. Central European Time. For those of you who have not yet seen the press release, it is available on our website, along with our latest investor presentation. As always, we remind you that this conference call may contain information relating to ASM's future business and results, in addition to historical information. For more information on the risk factors related to such forward-looking statements, please refer to our company's press releases and financial reports, which are available on our website. Please further note that reference in this call to profitability numbers will be primarily on an adjusted basis.
Details of the acquisition-related PPA expenses can be found in the press release and in the investor presentation. And with that, I will now hand the call over to Hichem M'Saad, CEO of ASM.
Thank you, Victor, and thanks to everyone for attending our Third Quarter 2024 earnings call. The agenda for today's call is as usual. Paul will first review our Third Quarter financial results. I will continue with the discussion of the market trends and the outlook, followed by Q&A.
Thank you, Hichem, and thanks, everyone, for joining the call. In this Third Quarter of 2024, our revenue increased to a new quarterly high of EUR 779 million and 6% at constant growth compared to the Third Quarter of last year. Revenue in the quarter was at the upper end of our guidance of EUR 740-EUR 780 million. Spares and service sales grew by 45% year-on-year at constant currencies. This growth was above trend, including stronger-than-expected demand from China and with a relatively strong quarter for our outcome-based services. In Q4, we expect growth to return to a more normal level again. Equipment revenue in the Third Quarter increased 22% at constant currencies year-on-year and was led by our ALD product line, which accounted for the majority of our equipment sales. By customer segments, revenue in the Third Quarter was led by memory, followed by foundry, and then logic.
Combined logic foundry continued to account for the largest part of revenue and decreased slightly both year-on-year and compared to Q2. Gate-all-around had again a solid and increased. Memory sales continued to grow significantly, following the strong order intake in the previous quarter. Memory sales consisted mostly of high-bandwidth memory-related DRAM applications, which grew sharply. 3D NAND sales grew strongly versus the low level in the Third Quarter last year and were stable compared to Q2 and represent a smaller part of our memory sales. Sales in the Power/A nalog wafer segment were fairly stable compared to the Second Quarter. Year-to-date, sales in this segment are down by a significant double-digit percentage. This is compared to the very strong level last year and reflecting the soft demand and inventory corrections in the industrial and automotive end markets. Moving on to the gross margin.
In the Third Quarter, gross margin came at 49.4%, up from 48.9% in Q3 2023, explained by mixed effects, including increased sales year-on-year from China, which were somewhat higher than expected. For H2, we expect China sales to be below H1 sales, as stated before. SG&A expenses increased 1% year-on-year, reflecting our ongoing focus on cost control. For the full year, we expect SG&A to be slightly up compared to 2023, excluding the incidental charge of EUR 8.4 million in the Second Quarter. Net R&D expenses increased by 36% year-on-year and 16% compared to the Second Quarter, driven by headcount growth as well as higher amortization charges and lower capitalization. As indicated in previous quarters, several development projects are entering the commercial release phase this year, which means amortization of the related capitalized development expenses has started.
Also, in relation to the completion of these projects, capitalized development expenses decreased in Q3 because some resources previously allocated to those projects are now being used for a number of months to support the commercial release and must therefore be expensed instead of capitalized. For the full year, we expect net R&D expenses to increase between 15% and 20%. We also booked a one-off gain of EUR 7 million on the divestment of a building in Singapore, recognized as a separate line item, other income and not included in operating expense. Including this one-off gain, operating profit margin increased to 28.2% in Q3, up from 25.3% in the same period last year. The one-off gain of EUR 7 million had a positive impact of 0.9% on the operating margin. Below the operating line, results included a currency translation loss of EUR 48 million in Q3.
This compared to a gain of EUR 16 million in Q2 and a gain of EUR 3 million in Q3 of last year. These translation results mainly relate to our cash position, which we hold for the largest part in US dollars. Results from investments, which reflect our 25% share of the net earnings from ASMPT, decreased to EUR 0.7 million in the Third Quarter, down from EUR 4 million in Q2 and slightly up from EUR 0.4 million in Q3 of last year. ASMPT results were impacted by the continued downturn in the back-end equipment market. Let's now turn to ASM order intake. In the Third Quarter, our new orders increased to a strong level of EUR 815 million, up 30% year-on-year at constant currencies. In terms of customer segments, foundry was the largest segment in the Third Quarter, followed by memory and then logic.
Combined logic foundry accounted for more than half of equipment bookings and were up compared to the second quarter. Gate-all-around orders were again solid and increased sequentially, with most of the tool orders now for high-volume manufacturing. Mature logic foundry orders were also up sequentially but down compared to last year. Memory orders slightly decreased compared to the second quarter but were still at a very strong level. DRAM orders were roughly similar to the strong level in the second quarter and again driven by high demand for high-bandwidth memory applications. 3D NAND orders decreased slightly compared to last year and were still at the fairly modest level compared to our DRAM orders. Orders in the power analog wafer remained at low level in Q3.
Bookings from China dropped somewhat, both year-on-year and compared to Q2, but were still slightly higher than we anticipated at the start of the quarter. To the balance sheets, we ended the quarter with EUR 747 million in cash, up from EUR 637 million the previous quarter. This increase was largely the balance of a very strong free cash flow of EUR 242 million in the quarter, partly offset by cash spent on share buybacks. Next to the increased profitability, the free cash flow was primarily driven by lower working capital in the quarter. Days of working capital decreased to a relatively low level of 48 days, down from 64 days at the end of June. Working capital fluctuates from quarter to quarter, and in Q4, we expected to be back in the target range of 55- 75 days. We spent EUR 13 million on CapEx during the third quarter.
Furthermore, we spent EUR 93 million on share buybacks during the third quarter. On July 25th, we completed the EUR 150 million buyback program that we started last May. Combined with the dividend paid earlier this year, we returned a total of EUR 287 million in cash to our shareholders this year, up from EUR 223 million last year. With that, I hand the call back to Hichem.
As just discussed by Paul, we delivered a solid performance in the third quarter despite continued mixed market conditions. Growth in the major economies has been relatively soft this year, and the forecast for next year does not look much better. AI is currently the main driver for the semiconductor industry, in particular boosting data center growth. Other end markets, such as PCs and smartphones, appear to be bottoming, but expectations for a more sustained recovery have again been postponed. The industrial and automotive end markets are still in a significant downturn, with limited visibility for improvement. This picture in the end markets is also reflected in the spending patterns of our customers. Semiconductor devices that enable AI are the main areas of wafer fab equipment spending, in particular Gate-all-around in advanced logic foundry and high-bandwidth memory in DRAM.
If we first look at the leading-edge logic foundry segment, momentum in Gate-all-around remains strong. Orders for Gate-all-around-related applications were again solid in Q3, and we expect a further increase in Q4. The tools that we are shipping now are in support of the high-volume manufacturing ramp in 2025. Our customers have been talking about substantial demand for 2nm from their customers, including from AI chipmakers, with expectations for 2nm being a larger node than 3nm. News flow in this segment over the last couple of months has somewhat impacted the outlook for leading-edge logic foundry spending in 2025, but it's still within the range of outcomes that we assumed last year when we presented our 2025 revenue targets. We are still projecting a substantial increase in our Gate-all-around-related sales in 2025.
First, leading-edge logic foundry spending was relatively soft for most of 2023, as well as in the first part of 2024, and is now picking up with a majority of the spend on the Gate-all-around nodes. Second, the transition from FinFET to Gate-all-around has meaningfully increased our addressable market, with several new ALD and Epi applications. This is reflected in our previously communicated estimate of $400 million SAM increase per 100K wafer capacity. And third, with customers now moving to the phase of high-volume manufacturing, we can reconfirm that we have maintained our strong ALD market share in the transition to 2nm, while in Epi, we have substantially increased our share.
In addition, we have strong traction in R&D with all leading customers for the next transition that are expected to move into volume manufacturing in the 2026-2027 timeframe, be it the sub-nodes of 2nm or the next major node of 1.4nm. This transition will further extend our addressable market, including the introduction and wider adoption of applications that we discussed in our investor day, such as metal ALD for interconnects and contact vias, backside power distribution, selective ALD, and also increasing Epi intensity. In the mature logic foundry business, which for us is mostly about China, the demand trends are almost opposite to the leading edge. In 2023, this business was strong, and in the first part of 2024, even at an exceptional level, compensating for the soft conditions in the leading edge.
Now, demand in mature logic foundry has started to come down, as several customers in China are moving to a digestion phase following the substantial capacity investment in previous periods. Next, our memory business. Demand continues to be strong and mostly driven by high-bandwidth memory. Our key exposure here is our high-k Metal Gate ALD technology, in which we are a leading supplier, and that's used in the high-speed DRAM chips that go into the HBM stack. We expect our memory sales to grow at a very strong rate for the full year, with a percentage contribution in 2024 potentially approaching the previous peak level of 20% that we achieved a couple of years ago. In 3D NAND, demand has increased this year for our ALD gap fill solution, but it's from a very low level last year and relatively small compared to DRAM.
We are working with customers on new ALD applications for the next device generation, but in the foreseeable future, the 3D NAND business is expected to remain the smaller part of our memory sales and mostly driven by technology buys. Next, our silicon carbide business. As most of you are probably aware, conditions in this market segment have continued to soften. We are not immune, and while we have reiterated our forecast for a double-digit sales increase in this business line in 2024, the rate of increase is substantially softer than what we still expected at the start of the year. It's too early to predict if the market will start to recover in 2025, but the long-term outlook for wide-bandgap materials, such as silicon carbide, is bright due to the increasing power efficiency requirement in EVs and longer term also in other areas, such as data centers.
In addition, we are in an excellent position to further expand our market share. Since the acquisition of LPE, we have already won several new customers across the globe. Earlier this month, at the Silicon Carbide Conference, we announced the launch of our first 200mm single wafer cluster tool, the P208. We expect this tool to be a game changer. Similar to our other silicon carbide Epi tools, the P208 offers best-in-class film performance, excellent within wafer and wafer-to-wafer uniformity, and the lowest level of defectivity. These benefits will only become more important as our customers are planning to transition to the larger 200mm wafer size. New in the P208 is its dual-chamber platform compared to the single-chamber architecture of our existing PE108 tool. This substantially increases throughput and lowers cost of ownership for our customers. We have already received multiple orders for the P208 from several leading customers.
Finally, China. As Paul discussed, our China sales held up slightly better than we expected in the third quarter. We still project, however, that revenue will be down in the second half compared to the first half and that Q4 will be below Q3. The main delta from the first half to the second half is the mature logic foundry business in China, which was at an exceptional level in the first part of the year and, as just discussed, has now started to normalize. Let's now discuss the guidance that we issued with our press release yesterday evening. For Q4, we expect sales to be in the range of EUR 770-EUR 810 million. Taking the midpoint, this means that second-half sales are expected to be up by slightly more than 15% compared to the first half, and that full year 2024 sales will increase approximately 10% year-on-year.
We expect bookings to decrease in Q4 compared to Q3. This is mostly explained by some orders that were pulled in from Q4 and that drove the upside in Q3 bookings. Compared to Q3, we expect Gate all-around orders to further increase in Q4 and China orders to be down. Looking at 2025, we now expect our revenue to be in the range of EUR 3.2-EUR 3.6 billion. This compares to our previous target of revenue of EUR 3.0-EUR 3.6 billion in 2025. The most important driver for us in 2025, as just discussed, will be the increase in Gate all-around-related revenue. In addition, following a strong increase in 2024, we expect memory spending to be sustained at a healthy level in 2025, again mostly supported by high-bandwidth memories.
On the other hand, the power analog wafer market is not expected to recover before at least the second half of 2025. We expect China spending in the segments that are relevant for ASM to be down somewhat in 2025, but not falling off a cliff, and that includes our assumption of a limited impact from new export control regulations. Before we open up for Q&A, a few words about the longer-term outlook for ASM. After almost six months in my new role as CEO, I have only become more convinced about the great opportunities that lie ahead for our company, not only in the coming years but also in the longer term. AI is expected to enable new end market applications and to support productivity gains in many sectors and industries. At ASM, we are using AI, as just one example, in the process of screening future ALD materials.
This helps to increase the efficiency of the process, accelerate the time-to-market of new ALD applications, thereby strengthening our competitive differentiation. As AI functionality is expected to move more and more to the edge in the coming years in PCs, in smartphones, and also in industrial applications, broader segments of the semiconductor end market, in addition to data centers, are likely to see accelerated growth. To address the increasing performance and especially power efficiency requirement of AI applications, demand for the most advanced semiconductor devices is expected to rise. This plays to ASM's strength, as the next-generation devices will bring smaller geometries, more 3D structures, and new materials, which in turn will require more ALD and Epi steps. We continue to invest in our infrastructure and in R&D, and we have great people in our company to execute and deliver on these growth opportunities.
Also very important, we have the early and strategic engagement with leading customers, and I'd like to thank them for their continued trust that are key to enabling the next-generation semiconductor technologies. With that, we have finished our prepared remarks. Let's now move on to the Q&A. We'd like to ask you to please limit your questions to not more than two at a time so that everyone has a chance to ask a question. Okay, Operator, we are ready for the first question, please.
Thank you. This is the Chorus Call Conference Operator. We will now begin the question and answer session. Anyone who wishes to ask a question may press Star and One on their touch-tone telephone. To remove yourself from the question queue, please press Star and Two. Please pick up the receiver and ask questions.
Anyone who has a question may press star and one at this time. The first question is from Andrew Gardiner with Citi. Please go ahead.
Good afternoon, gentlemen. Thank you for taking the question. I just was interested in a bit more detail around the gate-all-around transition. Hichem, you mentioned that you're now shipping tools into the HBM production lines, right? So we've seen the shift from pilot to HBM begin. I'm just wondering about your visibility into next year, right? You've revised your guidance for next year, but for ALD in particular, given the timing of the ramps, would you expect a stronger first half than second half? Or any additional color around the timing would be helpful. Thank you.
Okay, thank you for your question. I think we'll limit our comments here to the full year 2025.
It's really too early for us to be specific quarter-over-quarter of the trajectory within 2025. Regarding backlogs and sales in next quarters, yes, there is, of course, a relationship, but it's not linear. Sometimes we work on the basis of forecasts and can ship the tool within the same quarter as the order.
Okay, thank you. And a second one, if I could, just, you also mentioned the strength and engagement that you're having below the 2 nanometer nodes. You've shared the EUR 400 million of additional addressable market for every 100,000 wafer starts per month for 2 nanometer. Do you have a similar guideline for us at the sub-2 nanometer node?
No, we have not at this moment in time, and at least not externally communicated.
I think what we typically communicate, Andrew, is that with every, let's say, major node change, there is a double-digit increase, a meaningful double-digit increase, but we have not, let's say, quantified precisely what that means in terms of incremental SAM per 100K wafer starts per month at this stage. We might consider to do that in the future, but that has not yet been done. But it will be, again, a step up.
Understood. Thank you.
The next question is from Stéphane Houri with Oddo BHF. Please go ahead.
Yes, hello. Good afternoon, actually. I've got two questions. The first one is on China. I would like to come back on what happened during the quarter. You talked about pulling orders in Q3. Could you help us to quantify? I understand there was also an impact on revenues, then on gross margin.
Also, quantification here could be helpful to understand what's going to happen in Q4 and what would be, for instance, the gross margin without this additional revenue from China? And I have a follow-up.
Thanks for the question, Stefan. Indeed, we saw somewhat stronger than expected sales and orders in China in Q3. It did not change our view that, as we just said, for the second half, that China will be below the first half, and we also said that Q4 will be below Q3. So that still stands. The orders that we saw were amongst orders related to spares and services. You've seen the relatively high growth in spares and services, which you should not get used to. We see this as more an incidental quarter, which was partially not only but partially contributed by China, but also because of good growth in our outcome-based services.
Then, in addition, we saw some higher sales in memory in China, which, as you know, is the smallest segment over a number of quarters. Of course, from quarter to quarter, it can be different a little bit, but if you take a number of quarters, the smallest play for us in China is memory. But this quarter, it was relatively high compared to other quarters, but it doesn't change the fact that it is still our smallest play in China. So that was another, let's say, dynamic that we saw in Q3. And I think also mature logic foundry was maybe slightly higher than we initially thought at the beginning of the quarter. Then, obviously, that has, as you all know, typically everything else equal a negative impact to margin.
So, knowing, based on our view today and expecting that Q4 China sales will be below Q3, it is also likely, again, everything else equal because there's other elements as well that play a role in the margin development. But everything else equal, that indeed margin in Q4 would come down compared to Q3. That's a very fair assumption to take into account. I hope with that. Quantification? Prefer not to guide specific on margin development, also because there's a lot of impact, as you might appreciate, in terms of mix. So it's pretty hard. We have several scenarios, of course, that we take into account. There's always pulls in and pulls out every quarter, so it's difficult to precisely say at this quarter, but it will be somewhat lower than we will not be falling off a cliff.
So the same remark that we say for revenue, we can also say for margin.
Okay, thank you. And yeah, and so the second question is about the guidance for 2025 now that you have reworked a little bit the assumptions. It's a bit of a general question, but what needs to happen for you to be at the top end and what needs not to happen to be at the bottom end? Some general ideas would be ve ry helpful. Thank you.
Yeah, the reason for actually, let's say, increasing the bottom end of the range is based, one, first and above all, that we get closer to next year. So there's a little bit better visibility, but there's still quite a few moving parts, as you know.
And two, based on everything we know today and also based on all kinds of external research, we believe that the WFE market will continue to grow next year compared to this year. And with that, given that, yeah, we have basic guidance for revenue this year north of EUR2.9 billion, we would expect to grow as well next year, and which would bring us then, of course, higher than the, let's say, the lower end of our range. So the first and I think most important one will be the development of the overall WFE market. The second one, of course, will be the mix development within that market. If leading edge goes faster than non-leading edge or mature, that's, of course, better for us, and would bring us more close to the higher end of the range compared to the lower end of the range.
Of course, if, let's say, somewhere in the year, as Hichem already said, some markets that today are still sluggish or in a cyclical downturn, even when they start to recover, that will also, of course, be a positive for next year. So yeah, I would say the logical things that you would expect, but the overall market is, I think, the most important data point, and the second one is the mix, of course, in that overall market.
Okay, thank you.
The next question is from Adithya Metuku, HSBC. Please go ahead.
Yeah, good afternoon, guys. Thank you for taking my questions. So firstly, I just wanted to get some clarity on OpEx in 2025. To the extent that you're able to comment, how should we think about R&D and SG&A growth in 2025? Any color there would be appreciated.
And then I've got a follow-up on which I can ask afterwards.
Yeah. In absolute terms, for sure, R&D will continue to grow for two reasons. One, our growth investments will further increase given, again, all the opportunities that we see ahead of us. And two, amortization, as you've already seen in this quarter, was up, I think, EUR6 million or so quarter on quarter, simply because we have started to amortize a number of projects that are now into commercial release, mainly related to gate-all-around type of applications. Capitalization, you also saw this quarter was higher. I would expect that maybe not yet in Q4, but beyond that, to get back to a higher level than we have seen in Q3.
That was actually even lower than what I would have expected, to be honest, at the start of this quarter, but that was mainly, yeah, for the reasons I mentioned. So anyhow, adding that all up, you will see an increase in R&D. As I've said before, it is even possible, but don't take this as a guidance, that as a percentage of revenue, we will be at the higher end of the guided range, as we said, low single digit to high single digit to low double digit for R&D. So we'll be at the higher end of that range, and in the initial years, maybe even slightly higher, given the number of opportunities. So that's the best that I can give you now for the reasons I just mentioned.
Yeah, if I add something here on OPEX for 2025, as Paul has mentioned in his remarks earlier, we showed that our SG&A expenses have decreased 1% year on year, which reflects our focus on cost control. We're really focusing on cost control. We're focusing on efficiency and improving our business processes. So from that point of view, we're watching our SG&A part of our business for 2025, and that's something we're taking very seriously. Yeah, so for SG&A, Aditya, again, if you look to our guidance based on what Hichem just said, there we said high single moving gradually towards high single digit. We're now still in the double-digit figure, so we should move, of course, to a certain extent depending on revenue, but also, of course, based on our actions to control cost and improve processes, we should move to the high single-digit area next year.
Understood.
And then just as a follow-up, now with the changing dynamics in the logic foundry and markets with three customers potentially coming down to one at the leading edge, I just wondered if you could give your thoughts on how you see pricing pressure for your products. There's been some concerns about what it may mean for the overall semi cap industry and maybe more specifically for you. But also, if you could also comment a bit on how your ALD intensity might change. For example, some customers have tended to be more ALD intensive in their recipes than others historically. So what do these changing dynamics at the leading edge between these three customers mean for your ALD and epitaxy intensity and your share gains that you've talked about before? Any color there would be very helpful. Thank you.
Okay, so as Paul has mentioned earlier, I think the move to a 2 nanometer technology node has really increased our ALD intensity tremendously. Yes, there are some differences between customers depending on their integration scheme, but that difference is really minuscule or small compared to the major increase going from 3 nanometer FinFET technology node to the 2 nanometer technology node. So that delta from one customer to a customer is really immaterial. We still see a significant increase in ALD intensity going to 2 nanometer. Right now, we're working with our customer, actually, on even the next generation 1.4 nanometer technology node, which will be in production in 2027. And we see even further increases in ALD intensity with the three customers.
So we don't really see a difference between whether it's customer A or customer C. Understood. Any thoughts on price pressure or? Any thoughts on what?
Price pressure. Any thoughts on pricing pressure that the industry might see given the changing dynamics, if you're able to comment?
Yeah, I can't comment on your answer. I think that pricing pressure really depends on value. So if you provide value to your customer, I think that you should be able to gain business and profitable business. I think what's important for us is really to provide the value. I think what we do usually is making sure that we innovate, making sure that we constantly improve productivity, improve reliability of our tools, of our machines. I think with that, you can still continue to be profitable. So pricing pressure, I think it's really immaterial as long as you continue innovating and providing value to your customer.
Understood. Thank you.
Thank you.
The next question is from François-Xavier Bouvignies, UBS. Please go ahead. Thank you very much.
My first question is maybe a clarification on the dynamic with Samsung and Intel. I mean, we saw in the press and we, of course, saw ASML guidance revision. You mentioned as well the outlook somewhat changed, but it seems that you confirm and you even upgrade 25. So I was just wondering, did you have any impact, I mean, from that Samsung, Intel? Because maybe you asked from confirmation of orders or maybe your orders are coming more from one customer's. Did you factor anything from what's going on at Samsung Foundry and Intel? I just wanted to make sure. Because when I read, it doesn't seem that you are impacted. You include the guidance. You say that you had confirmed orders from everybody, and that Gate-all-around is going to be very strong next year. So I just wanted to clarify maybe if you had any impact.
So I think to be sure, I mean, we're not going to be talking from one customer to the other. But what I can tell you is that, I mean, we have seen an impact in 2025, but it's not really a huge impact, and it's still within the range of expected outcomes that we talked about. It does not change our view that Gate-all-around related sales, we really still continue to see substantial increase in 2025.
And why is it not huge? I mean, it's two big players. They spend a lot of CapEx, and you are less exposed to memory, more logic. So I would think it would still be material. So why is it not big?
I think maybe the reason why it's not so big is because what we see today is that, yeah, this change, the Gate-all-around, is important for all, let's say, three customers, and it's important that they become successful, and they will continue to invest to become successful in this transition, yeah, which is a new architecture, which is strategically important for them, not only for next year, but for also, let's say, medium to longer term. And so the cuts, as we see today that they might talk about, yeah, publicly, are not so much, let's say, in the field of Gate-all-around, but more in other areas, which impacts us less.
Okay. Makes sense. Thank you very much, and maybe my second question is on moly. I mean, you talked a lot at Capital Markets Day.
Some of your competitors are talking about this from the migration from Tungsten, and I was wondering how you see, as the layers are being laid out, how your share is evolving. I mean, do you still feel strong about your share in moly to increase? If you could update on that front, it would be great.
Thank you. So I think if you talk about metallization, what the industry is experiencing recently is really a generational shift in metallization from moving from Tungsten to Molybdenum. This is an area metallization is a new area for us. This is something that we didn't play in. And this is the first time that we, ASM, have entered the metallization market with Molybdenum ALD. And this Molybdenum is really, like I mentioned, it's generation shift. It's going to be applied in 3D NAND.
It's going to be applied in DRAM, and also it's being applied in Logic. We are actually working with all customers on the three technology areas. We have made some wins in some markets. We're actually shipping Molybdenum reactors as we speak for our customer for 2 nanometer technology node production. So we feel good about our position. This is a newer area. We see introduction in Logic happening right now in 2 nanometer node, but I think as you go to the 1.4 nanometer technology node, you're going to see more and more Mo application. DRAM is slower from that point of view, but it's going to happen soon, and there's more and more layers happening in Molybdenum. So we are very excited about this market, and we look to be an area of growth for us in the future.
Great. Thank you, Hichem.
Thank you, François.
The next question is from Sandeep Deshpande, JP Morgan. Please go ahead.
Yeah. Hi. Thanks for letting me on. My question is, how many wafers of GAA do you expect? How many wafers do you expect to transition to GAA in 2025? And given what we had heard from some of the other players that they've seen some kind of slowdown, has the number of wafers that are transitioning to GAA in 2025 changed? Has it gone up or has it gone down since the changes have occurred in the industry? That is my first question. Thanks. And I have a quick follow-up after that. So I think, as we mentioned earlier, we see Gate all-around in 2 nanometer to be a strong node driven by AI. There is a huge demand for AI chips. There's a huge demand for Gate all-around chips.
So it's actually in o`ur remarks earlier. We said that 2 nanometer, which is GAA, is going to be a bigger node than 3 nanometer FinFET. So it's really strong. And I cannot give you the actual numbers, but what I can tell you, it's a significant node, significant investment in 2 nanometer happening right now. But is it bigger than what you had previously forecast, or is it lesser than what you had previously forecast? Actually, it's the same as we previously forecast. It's still very strong.
Understood. Thank you. My follow-up question is on China. I mean, you've had a stronger third quarter than expected on China. Many of the players exposed to the China market have not been able to forecast China that accurately.
So do you actually have a cl ear visibility into China into 2025, or is this your view into China and the decline into China, a view of ASM, or rather that what you see in terms of orders? Because, I mean, right through this year, China has surprised most companies on the upside rather than on the downside. Yes.
Thanks for the question, Sandeep. I wish we had good visibility. So similar to other companies, we also have limited visibility. And as you heard, we also have been, I think, positively surprised somewhat, at least in Q3, still for the two quarters together, not. We still believe that H2 will be below H1. We've said that for next year, we expect more normalization of sales compared to what we call the exceptional high sales, let's say, maybe in the last six quarters or so, six quarters. How much?
There's, of course, different assumptions. In addition, you have export restrictions that are not yet fully clear. Also there, we've made certain assumptions in, of course, the guidance that we've given now, but that's still to be seen how that will play out. So we have some level of visibility based, of course, on all our discussions with all our customers, all our intel that we have. But yeah, at the same time, also as we've seen in prior quarters, that might still change. But at least we have taken into account, yeah, a more normalization of the sales that we have been, I think, as one of the first ones been talking about already, that might happen in 2025 compared to 2024. And that's reflected in our current guidance. Yeah. Thank you.
If I add a couple of notes regarding China to what Paul has mentioned, for us, really, China is different from many other players. I think for the major, bigger player than us, China, they are very strong in the material technology node. We don't have, because we're an ALD company, we don't have much ALD in this material technology node. Our bigger play is really the company to the power analog wafer market in China. That's really the difference between us and our biggest competitor. We see the power analog wafer market has slowed down from the start of 2024. I think the softness for the global power analog wafer market continues through 2024, and we don't know when it's going to recover. Is it going to recover second half of 2025 or 2026? That's something really to wait for and to watch in the future.
Thank you, Sandeep.
Thank you.
T he next question is from Robert Sanders, Deutsche Bank. Please go ahead.
Yes. Thanks for taking my question. I just have a few questions, just about the last one, one question about 3D DRAM. There's been a bit more news flow about the companies outlining basic kind of structures and how that might work out. Do you think that that is still in the pathfinding stage or i.e., five years away plus, or do you think that there are some memory companies now moving into a more firm development stage?
So I'll answer this question. I think that what we have mentioned in our last previous conference call is that we know what's going to happen. We have better visibility on the DRAM roadmap. So we talked about the 6F-square cell architecture, which is happening right now.
And then the industry is moving to 4F-square, which is really shrinking of the cell. So that's actually going to happen, and it's going to happen in two, three generations in the future, which really delays the 3D DRAM to 2030. That's really what we have mentioned. Right now, we see actually 3D DRAM to be in production in 2031. That's the earliest we see it right now. So in DRAM, the next few years is going to be big in the 4F-square architecture, which, by the way, opens many new layers for ALD and epitaxy.
Got it. And on Molybdenum, you mentioned the transition from Tungsten, but is there not an opportunity from Copper as well? And isn't that as large an opportunity?
Yes. I think Moly replaces both Tungsten and Copper.
I think the beauty of Moly as a metal is that it does not need a barrier layer when you deposit it. Copper, everybody knows that copper diffuses a lot. It's a very high diffuser as a metal. So in order for us to stop the diffusion of copper, we have to put a barrier layer. This barrier layer that we deposit increases the resistance of copper, the final resistance of the copper. With Molybdenum, you don't need to use this barrier layer because Molybdenum does not diffuse. It's a big molecule, very big atom. So it does not diffuse this. And with that, actually, you achieve a much lower resistance. Absolutely. Molybdenum is replacing both Tungsten and also Copper in metallization.
Thanks a lot. Thanks, sir.
The next question is from Didier Scemama, Bank of America. Please go ahead.
Thank you so much. Good afternoon, everyone.
A couple of questions. So first, maybe a question for Paul. I think you said Spares obviously was a pull-in in Q3 from China, but that's going to normalize. You said it's a one-off. Just give us a sense of what Q4 is going to look like. What is a normalized level, or is that going to take several quarters to go back to a normalized level? That would be my first question, and I've got a follow-up. Thank you.
What we've seen this year in Spares is actually quite a few quarters in a row, a pretty strong double-digit growth, but not to the level that we've seen in Q4, obviously. Sorry, in Q3, obviously. I would expect to be a little careful here, but I would expect it to go back already in Q4 more to a normalized level that we've seen in the last few quarters.
At this moment in time, it would be a positive surprise if we would see something similar happening as in Q3, and I don't think that will happen. So I would not, at least so far, we have not gotten those indications that it's again to happen in Q4.
Okay. But a normalized level is what, 123 that you had in Q2? Is that what you're thinking about, or?
Specific number, of course, but yeah. It's definitely in Q3. But yeah, if you look, let's say, over the last few quarters, I think that's something that is likely going to happen.
Okay. Brilliant. And a follow-up to Hichem. So congratulations going into high-volume production for Gate-all-around equipment. My question is, when you look at your leading customer, I think the intention is to build about 40,000 of wafer starts per month capacity for 2 nanometer GAA.
c endar 25 over throughput, or do you think that actually this is going to carry through calendar 25 because 26, the capacity should again double? And as you mentioned, 2 nanometer capacity build will be bigger than 3. So just trying to understand a little bit whether you think there is a risk to the extent that there is no high-end smartphones ramping Gate-all-around in 25.
I think Gate-all-around is really driven mainly by data centers more than smartphones.
So the demand for that is very significant, and it dwarfs the 40K wafer stock per month that you're talking about. As we mentioned, 2025 will see significant investment in 2 nanometer in HBM, and we think this is going to continue in the future. Super.
T hank you so much.
Thank you, Didier.
The next question is from Timm Schulze-Melander , Redburn Atlantic. Please go ahead.
Yeah. Great. Thank you for taking my question. First question was for Hichem. So Molybdenum, congratulations on the insertion into volume manufacturing. How many customers do you expect to put Moly into high volume in 25, please, and then have a follow-up?
We're not going to really talk about the number of customers in production because of different factors. Some customers really have different technology.
For the same technology node, they have a different integration scheme whereby, okay, they will introduce Molybdenum in one generation, not the other one. So that's different from one customer to the other. But we see some customers who are more aggressive or for some integration reasons, they need Molybdenum to introduce Molybdenum in this first generation rather than the second generation. And when I talk about first generation and second generation, is that okay? When you look into the 2 nanometer, 2 nanometer is a big node. So 2 nanometer can be 2.0, can be 1.8, can be 1.6 nanometer. So there's a different sub-node. So as the sub-nodes continue, you see more and more Molybdenum being introduced in those layers. And the same thing is really we see the same thing happening in the other technology, be it 3D NAND or DRAM.
Okay.
It's more than one process tool of record. The others are not still development tool of records. They've moved on to PTOR, kind of level of conviction, I guess. Maybe a question for Paul just on the SAM uplift per 100,000 wafer starts a month. Could you just give us a kind of sense of scale of what the base is? Is that a $400 million uplift on a base of $100 or $400? Just some color there would be really helpful.
Thank you. Yeah. It's a very fair question, Tim, but actually, we've never quantified that externally, at least. It's compared to the FinFET node. So it's from an EPI and ALD combined. So that's the $400 million we have referred to. So unfortunately, I cannot give you more.
I don't even know the precise answer by heart, to be honest, but we have also not disclosed that. But yeah, it's compared to the FinFET node. It's more than double-digit. It's meaningful double-digit, as we always say, but that's maybe not very helpful. But that's as far as we, yeah, have given insight into the markets.
Okay. But it's not triple-digit, I guess, is also similarly helpful. Would that be fair?
No, it's not.
All right.
And the other thing is, as I said earlier in the call, of course, with every subsequent major node, there will be incremental SAM for us because of, yeah, shrinking and/or going 3D. It's, yeah, the perfect development for ASM.
All right. Thank you. Good luck.
Thanks, Tim.
The next question is from Tammy Qiu, Berenberg. Please go ahead. Hi. Thank you for taking my question.
So from GAA perspective, relating to ASMI's ALD penetration, does a different player ramping up in different volume make any difference for you, i.e., your market share is different between various GAA players? And also, how do you think about your ALD market share or TAM expansion when all the three main chip makers go into backside power relating to your ALD marketplace?
If you look into ALD, I think your first question was talking about ALD intensity in 2 nanometer. That's the first question. Am I right?
Yes. Yes.
I think what I can tell you there is that for, and which really was talked about earlier, one of your colleagues has mentioned that, okay, some customers use more deposition layer than others, etc., etc. That's true. Okay. I mean, it really depends on litho.
It depends on, okay, if you have more aggressive litho, then you have less patterning layers, etc., less patterning layers, so you have less ALD, but in the whole scheme of things, these delta changes between one customer to the other is small compared to the double-digit significant increase in ALD intensity going from FinFET to gate-all-around, so for us, it's really immaterial from that point of view. It's really very small from that point of view, so that's really the answer for your first question, and your second question, remind me of your second question,
so basically, how does your ALD market or TAM or market share change when all three mainstream makers start to use backside power?
Yeah, I think that what I can tell you is that our market share is good, strong for all three players.
I think in ALD, we are a leading player in ALD, and as I mentioned in my remarks, we are actually keeping our good market share in ALD from the transition to FinFET to gate-all-around.
Okay. Thank you.
Thanks, Tammy.
The next question is from Nigel van Putten, Morgan Stanley. Please go ahead.
Hi. Thanks. A follow-up on DRAM in China. Maybe reading too much into this, but is your visibility of healthy demand into next year for DRAM and also saying that China will only be moderately lower in the first half of 2025, I mean, are those related? So are you seeing China memory business becoming more of a tailwind into next year? That would be my first question.
Yeah. Thanks for the question, Nigel. That is difficult to say. We have, of course, a certain level of visibility.
As I said, it's likely to be still, especially if you take a number of quarters. I mean, from quarter to quarter, there can of course be some phasing difference or lumpiness or whatever. But if you take a few quarters, memory is the relatively smaller part of our China business. So for us, most important is actually, at least in the last, whatever, six, seven, eight quarters, logic foundry. And before that, it was power and analog, and that might become number one again. Now that logic foundry is likely to come down somewhat or normalize, we don't know yet how it will play out next year. But in all scenarios that we're looking at today, memory is still the smallest part of our play in China.
Got it. And not any sort of view today that that could change into next year? There's no reason to believe that.
Yeah.
I don't want to speculate it, but no, the current, if you look, if you have our base case scenario, which you don't have, that's not what we assume. But having said that, there is a limited amount of visibility. I'm never going to say it's not possible, but I don't expect that. We don't expect that.
Got it. Maybe a second question on DRAM broadly. So I think for HBM, a lot of conversion going on of the installed base, some of it coming from non-High-K nodes into High-K nodes. So my question is really, is there a double benefit today that you're sort of getting conversion to the latest nodes, but in addition, there's the High-K? And as we maybe move into next year with High-K conversion, maybe at least partly completed, I think you've been going at it for a while now.
Is that maybe not a headwind, but could that slow sales as you're sort of normalizing the demand drivers?
Well, I mean, I'll answ` er this question then, Nigel. I mean, the way I look into the HBM, I look into the GAA. If the GAA is strong, HBM is going to be strong. Because I mean, those two chips go hand in hand. What we have seen that the GAA is really very, very strong for 2025. So from that point of view, we see that HBMs do continue to be stronger in 2025.
Okay. Understood. Thank you.
The next question is from Mark Hesselink, ING. Please go ahead.
Thank you. My first question is actually a bit of a question on the pull-in of the orders. I think you discussed it before. It was partly China and then also, I think, gate-all-around.
I mean, is the Gate all-around part the biggest part? And the reason why I'm asking is that historically, actually, with these kind of technology transitions, whenever you saw those pooling of orders coming in, there was typically very positive signs for more pull-ins in the quarters to come. Just want to understand that dynamic a bit.
There's not one specific reason for the pooling. So you should not read too much in it, Mark. I mean, if we say pooling, it's based on, let's say, our expectation that we have based on customer forecasts. So although we don't guide on orders, we still have our own expectation what orders will be, of course, for the next quarters. And then when we see that some orders come in, let's say, earlier than what we had in our forecast that are basically customer forecasts, that's when we speak about poolings.
It's not one specific segment or one specific customer. Sometimes it can just be a few weeks earlier, and sometimes it can also be a few weeks later. There's no real hidden message in this remark. So yeah, that's basically what it is.
Okay. That's very cl ear. And then my second question is on your position in DRAM outside HBM. I think everyone has always in the past said, "Okay, we start with HBM," and then eventually, we also expect to win layers in the DDR part. I mean, can you maybe give a bit of an update there?
Yes. I will take this question. I think for DRAM, DDR is also still high-performance memory. So with this high-performance memory, you see High-K Metal Gate incorporation in DDR5. So you see more and more ALD layers incorporation.
And also, so we see better ALD layer and also more epitaxy also being implemented in DDR DRAM from this point of view.
Okay. So you're winning market share also taking out the HBM part?
I think once you go to High-K Metal Gate, we have a leading position in High-K Metal Gate from the many years we've been doing this in logic. So yes, we have quite a few customers running High-K Metal Gate, ASM High-K Metal Gate in. Sure.
Okay. Thanks.
Thank you, Mark. I'm afraid that we have run out of time. There are still a couple of callers in the queue. So if you didn't have the chance to ask your question, please contact us after the call.
So if there are no more questions, sir, I turn the conference back to you for any closing remarks.
On behalf of Paul and Victor, I would like to thank everyone for attending today's call. We hope to meet many of you in the upcoming conferences and other investor events. Thanks again. Stay safe and goodbye.
Ladies and gentlemen, thank you for joining. The conference is now over, and you may disconnect your telephones.