BE Semiconductor Industries NV

AEX:BESI

Earnings Call Analysis

Strong Q3 Growth Driven by Advanced Packaging for AI

In the third quarter of 2024, Besi reported significant growth in revenue, orders, and net income compared to the same period last year. Revenue reached EUR 156.6 million, up 27%, while orders totaled EUR 151.8 million, growing by 19.2%. This surge is primarily attributed to strong demand in AI applications, particularly in computing end-user markets such as hybrid bonding and photonics. However, this growth was tempered by challenges in the automotive sector and from clients in the Chinese market.

Boost in Net Income and Cash Position

Net income for the quarter rose to EUR 46.8 million, an increase of 33.7%, indicating operational efficiencies despite rising R&D costs. Furthermore, Besi's net cash position improved, reaching EUR 110.7 million, bolstered by recent senior note offerings, positioning the company well for future industry changes.

Market Forecast and Challenges Ahead

The overall market environment remains mixed, with continued growth in AI-related applications juxtaposed against ongoing weaknesses in mainstream assembly applications. TechInsights has revised its 2024 growth forecast down to 2.9% from an earlier estimate of 9.3%, suggesting a delayed market recovery expected in 2025-2026. Besi remains cautiously optimistic, noting that while the macroeconomic landscape poses challenges, its focus on advanced packaging continues to drive growth.

Capacity Expansion and Technological Advancements

Besi plans to expand its advanced packaging capacity significantly. In 2025, the company aims to double the clean room capacity at its Malaysian facilities and enhance R&D capabilities in Singapore. This strategic move reflects the company's commitment to meeting increasing demand for hybrid bonding systems, with advanced applications expected to grow significantly in the coming years.

Fourth Quarter Guidance: Flat Revenue Expectations

Looking ahead to Q4 2024, Besi forecasts flat revenue, with a range of plus or minus 10%, primarily impacted by customer shipment delays relating to hybrid bonding systems. The expected gross margins are forecasted to be between 63% and 65%, with operating expenses anticipated to remain flat or increase by 5%.

Customer Demand and Continued Growth in Hybrid Bonding

Total orders for hybrid bonding systems surpassed 100, with continued demand expected into Q4. Besi aims to fulfill a backlog of orders, reflecting growing adoption of its technologies across a diverse clientele. This expansion aligns with the rising trend towards high-performance computing demands in the semiconductor industry, ensuring Besi remains well-positioned for future growth.

Earnings Call Transcript 2024-Q3

[Operator]

Good morning, good afternoon, ladies and gentlemen, and welcome to Besi's Quarterly Conference Call and Audio Webcast to discuss the company's 2024 third quarter results. You can register for the conference call or log into the audio webcast via Besi's website, www.besi.com.

Joining us today are Mr. Richard Blickman, Chief Executive Officer; and Mrs. Andrea Kopp, Senior Vice President, Finance. [Operator Instructions] As a reminder, ladies and gentlemen, this conference is being recorded, and cannot be reproduced in whole or in part without written permission from the company.

I'd like to remind everyone that on today's call, management will be making forward-looking statements. All statements, other than statements of historical facts, may be forward-looking statements. Forward-looking statements reflect Besi's current views and assumptions regarding future events, many of which are, by nature, inherently uncertain and beyond Besi's control.

Actual results may differ materially from those in the forward-looking statements due to various risks and uncertainties, including, but not limited to, factors that are discussed in the company's most recent periodic and current reports filed with the AFM.

Such forward-looking statements, including guidance provided during today's call speak only as of this date, and Besi does not intend to update them in light of new information or future developments nor does Besi undertake any obligation to update the forward-looking statements.

I would now like to turn the call over to Mr. Richard Blickman.

[Richard Blickman, executive]

Thank you. Thank you for today's call. We'd like to review the key highlights for our third quarter and 9 months ended September 30, 2024, and update you on the market, our strategy and outlook.

First, some overall thoughts on the third quarter. Besi reported significant growth in revenue, orders and net income in the third quarter of this year versus the comparable period last year, as we continue to benefit from strength in our advanced packaging product portfolio for AI applications, despite continued headwinds in mainstream and Chinese assembly equipment markets.

For the quarter, revenue of EUR 156.6 million and orders of EUR 151.8 million grew by 27% and 19.2%, respectively, versus the third quarter last year, due primarily to strong growth by computing end-user markets, including hybrid bonding, photonics and other AI applications. Such growth was partially offset by weakness in automotive and Chinese end-user markets, continuing trends we've experienced this year.

Net income of EUR 46.8 million grew by EUR 11.8 million or 33.7%, reflecting a number of favorable trends, including increased Advanced Packaging Systems revenue, increased gross margins related thereto and better than forecasted operating expense levels despite continued growth in R&D spending for next-generation hybrid bonding and TCB systems.

For the first 9 months of this year, revenue of EUR 454.1 million and orders of EUR 464.8 million increased by 8.3% and 21.7%, respectively. Growth was due to significantly higher demand by computing end user markets, particularly for AI-related hybrid bonding and photonics applications and from Taiwanese and Korean subcontractors.

Net income of, in total, EUR 122.7 million was approximately equal to year-to-date 2023, as higher revenue and gross margins this year were offset by higher R&D spending in support of wafer-level assembly development and share-based compensation expense.

Our financial position improved as well in the third quarter of this year, with net cash increasing to EUR 110.7 million at the end of the quarter, an improvement of EUR 36.3 million versus the second quarter of this year, and EUR 20.5 million versus the third quarter last year, despite increased share buyback activity.

Total cash and deposits at quarter end grew to EUR 637.4 million, including net proceeds from our senior note offering in July of this year, which positions us favorably for anticipated growth in the next market up cycle.

Next, I'd like to speak a little bit about the current market environment and our strategy. The primary trends affecting the assembly equipment market in the first half of this year continued in the third quarter, namely strong growth in AI-related applications were offset by ongoing weakness in mainstream assembly applications. As a result, TechInsights further reduced its 2024 growth forecast to 2.9% from 9.3% previously, with a strong cyclical upturn pushed out to 2025 and '26.

As we've noted previously, assembly equipment market trends have diversed from front-end markets for the past 2 years, primarily given the absence of geopolitical trade restrictions in our market segment.

We are pleased with our strategic progress in 2024 in a challenging market environment. Besi continues to expand its development spending in support of the industry's leading advanced packaging portfolio, and has seen significant growth in both hybrid bonding orders and revenue this year.

During the third quarter this year, we received substantial orders for hybrid bonding systems from existing and new customers, and anticipate additional orders in the fourth quarter this year from a variety of customers, as adoption continues to expand globally.

We've also received increased interest for Besi's TCB Next system from leading logic and memory customers, which positions us favorably for anticipated growth in the next generation 2.5D and HBM applications. As such, we have taken steps recently to expand our advanced packaging production capacity in anticipation of future growth. In 2025, we intend to approximately double the clean room capacity of our Malaysian facilities and increase R&D and process development for hybrid bonding and TCB capabilities and customer support at our Singapore facility.

Now a few words about our guidance. Looking forward to the fourth quarter, we expect expanded adoption for hybrid bonding applications to be mitigated by ongoing weakness in mainstream assembly markets.

For the fourth quarter this year, we forecast that revenue will be flat, plus or minus 10%, versus the third quarter, partially due to shipment delays by a customer for certain hybrid bonding systems scheduled for delivery, which were scheduled for delivery in the fourth quarter of this year. In addition, gross margins are anticipated to range between 63% and 65% based on our projected product mix. Aggregate operating expenses are forecasted to be flat to up 5% versus the third quarter this year.

This concludes my prepared remarks. Before we begin Q&A, I would like to remind everyone to limit your questions to 2 at a time, so all participants have an opportunity to ask questions. Operator?

[Operator]

[Operator Instructions] We will take the first question from line Ruben Devos from Kepler Cheuvreux.

[Ruben Devos, analyst]

Yes. Just the first one is related to the hybrid bonding orders just in Q3. Could you give us a bit more detail on those that you've received the past quarter? I was just looking for a bit more granularity, like did a new account join or the orders somewhat in line with your expectations? And I think for the first 9 months of the year, how many units have you effectively shipped? I think at the end of 2023, you talked about an installed base of 40 revenue producing units, but curious where you are today? That's my first question.

[Richard Blickman, executive]

Well, that's a lot of information. Anyway, in the third quarter, we received continued orders for hybrid bonders, primarily for volume production at a major customer, which we have been serving from the beginning. At the same time, we added new customers to the hybrid bonding technology development.

As we said in the press release, we now have received over 100 orders for hybrid bonding systems. And as we have a certain backlog, and one of them clearly has to do with the order received in Q2, where we said that certain part will be shipped in Q4 and the rest in the first half of next year, and we repeated that comment because it also has an impact on our revenue guidance for Q4.

So we also added in the comments that we expect continued orders in Q4. It is always difficult to forecast precise numbers, also because the industry has a tendency to order machines simply based on end market demand. And that's not always the same quarter-by-quarter. We don't disclose further detailed information about customer specifics, products and number of machines.

[Ruben Devos, analyst]

Okay. Yes, it's fair enough. I just had an additional one on hybrid memory. I think you've been talking before about how the Korean makers are planning to use both PCB and hybrid bonding phase before. It seems like one is being more explicit about hybrid bonding adoption than the other, but both are leaving sort of the final decision to their customers based on varying requirements. Could you help us understand how that works in practice? What are sort of the trade-offs in the more specific scenarios that memory chip makers and their customers are considering when deciding which technology to go with?

[Richard Blickman, executive]

Excellent. Well, it has to do with performance. So far, the performance of hybrid bonded devices clearly demonstrate that, that technology offers higher performance, faster circuitry, less energy required. So that positive part of hybrid technology, apart from the fact that you can connect smaller geometries, then with reflow processes, is what is driving the continued adoption of hybrid bonding over, you can say, conventional reflow based processes.

So what the world explains in different ways, the 3 big memory manufacturers is that, in one way or the other, they expect a side-by-side development of using this technology for HBM 4 but certainly HBM 5, where hybrid bonding becomes to be expected more critical as dimensions decrease.

We are involved in this wonderful challenge with both technologies. So on the one hand, the TCB, so the reflow process. And on the other hand, the hybrid bonding. And time will tell, and especially in '25, it should become more clear which direction will prevail and with how many systems in percentage or, let's say, capacity in percentage. Probably for HBM 4, it will only be a very high-end range using hybrid. That's how it looks today, but that can also change due to the simple reason that the performance in the end determines which technology to use.

[Operator]

We will take the next question from line Robert Sanders from Deutsche Bank.

[Robert Sanders, analyst]

I just had a question about TCB. So you've talked about increasing capacity in high rebounding. But in TCB, we don't really have an idea of how large the capacity corridor is in terms of units per month or anything like that? Can you give us some kind of estimate of that? And I have a follow-up.

[Richard Blickman, executive]

Yes. In a similar way, although the systems are different, but from a manufacturing, let's say, technology, they are not that different. So the number of hours required to assemble such a machine to test it, and they also require assembly and clean room environment. And that will lead to a similar number of systems per month, 15 is currently. But then with the extended capacity in Malaysia coming online second half of next year, we should be able to do more. As we said in the call, we intend to double that capacity.

But yes, from, let's say, a broader perspective, the number of DC machines for HBM applications can be significantly higher than HBM -- than hybrid bonding for HBM, simply because the HBM 4, as it looks today, is produced with percentage-wise more reflow DC process than with hybrid bonding. But again, your question, we can build a similar number of machines as compared to hybrid bonders.

[Robert Sanders, analyst]

Got it. And just as a follow-up, when you are doing these kind of vendor selection assessments on TCB for -- you talk about bond pad pitch and all of those sort of things. Is it [indiscernible] that is the incumbent typically? Are they the kind of competitor to be? Or is it also ASMPT, K&S, others, just so we understand?

[Richard Blickman, executive]

Well, the key is, as you said, bond pad pitch. So the design, the architecture of the device, can you connect that with a reflow process? Or is it -- are the dimensions too small to say it in simple terms? Everyone is looking at that same challenge.

We have the advantage that our TCB Next has been designed, in particular, for below 20-micron bond-pad pitch. So it is much more accurate. We also are already producing data for below 10-micron bond-pad pitch. So you come very close to where the crossover point to hybrid bonding is.

Yes, we always benchmark versus competition. That's what our customers do every single day. And in the end, it's a matter of cost of ownership. So which system produces the most attractive cost of ownership, that's a matter of yield, so repeatability of process and at the same time, the throughput. And that combination and then with big calculation with cost of footprint and energy results in a cost of ownership model, and that's how we compare to any competitor.

[Operator]

We will take the next question from line Madeleine Jenkins from UBS.

[Madeleine Jenkins, analyst]

Richard, can I get an update on your Generation 3 hybrid bonding tool? I was just wondering when now you think that's going to come to market and kind of roughly what [indiscernible]?

[Richard Blickman, executive]

Well, that's a great question. First of all, the Generation 1 plus, which we don't call 2 because it's on the same -- exactly the same platform. From the 100 -- over 100 systems which are now on order, more than half is this Generation 1 plus. So the industry is moving to ever tighter specifications.

And the 50 nanometer, which is Generation 2, we call it Generation 2, not 3, that's just in the name, should be ready by the mid of next year. Already today, we can demonstrate that we're very close to that 50. So the system architecture is different. It needs more dynamic balancing, and we have succeeded also with this concept to achieve a higher throughput. And that will be very important for this next round.

So 2 major aspects, one being the accuracy. So remember, we started 150, 200 nanometers, Generation 1; Generation 1 plus, 100 nanometer, which is today the -- yes, let's say, the sweet spot. And then the next is 50 as of the second half of next year, and combining that with a concept, which offers a higher throughput.

[Madeleine Jenkins, analyst]

Perfect. And then just wanted to follow up on the capacity expansion. What kind of triggered this decision? Was it an indication from a customer of higher volumes? Or is it more just an internal decision, I guess, based on the long-term growth potential of the technologies?

[Richard Blickman, executive]

No. We don't invest because we like to satisfy ourselves. No. It's clearly driven by customers, customers who continuously confirm in one way or the other. So one way is the logic. The adoption in logic is becoming ever more, yes, clear and demonstrated.

Then we have the HBM question. As discussed in an earlier question, that could -- when that happens, lead to more orders than for logic. The combination is always more on logic than on -- let's say, more on memory than on logic.

Then we have the broader adoption for high-end smartphones. So the AI, the HAI introduction, that also could lead to significantly more machines. So we have to be ready for them.

And with the decision based on customer investment models, we should be ready second half of next year to be able to offer that capacity. Hopefully, they need it. If not, then it may take a bit longer. But anyway, it's based on customer programs.

[Operator]

We will take the next question from line Charles Shi from Needham & Company.

[Yu Shi, analyst]

Richard, I want to start with the TCB. We've been hearing your 2 competitors who are working on a TCB qualification at the leading foundry in Taiwan, one of them might have seen some big challenges recently and might have failed qualifications that the folks I spoke with investors, I spoke with, we're looking at the other vendor, whether they can step up.

But my question for you is this, does this open up the opportunity for Besi to position TCB Next at this leading foundry? And if it does, when do you think you can shift a qualification system? And just give us some thoughts at this point.

[Richard Blickman, executive]

Well, of course, it offers opportunities, and it's directly linked to the anticipated smaller geometries moving -- bond-pad pitches moving below 20-micron, 18 is the next standard, where our system initially has been developed for.

And qualifications take typically a year, so one should expect that in these next rounds, that they should come yes, let's say, in the second half of next year. But it definitely offers us opportunities. And yes, in particular, for those applications where the bond-pad pitches become more critical.

[Yu Shi, analyst]

Got it. The other question I have, Richard, the hybrid bounding system, you talked about a little bit delay. Not sure if you actually gave the color, but is it delayed, let's say, to next quarter? Or is it more delayed a little bit more into the future? Can you give us a sense kind of latest by how much? Yes.

[Richard Blickman, executive]

No, let me make that -- try to make that more clear. We announced this order, remember, in May, we announced a big order. And we also mentioned that it would be partly shipped this year and next year.

And to be a bit more clear, we anticipated more shipments out of the total this year and then the remaining part early next year. Well, the split has been slightly different. So in a sense, about half will be shipped before the end of this year, and the other half in the first half next year, first quarter, maybe April, that's the current planning.

But it's not a message that it is delayed by no means. There's an enormous customer drive behind this part, as we also shared earlier, is installed in the U.S., another part is installed in Asia, and that is causing some logistics and people have to be trained. So don't misread this message that the delay is caused by a delay in adoption.

[Operator]

We will take the next question from the line Marc Hesselink from ING.

[Marc Hesselink, analyst]

Yes. I would like to put that number of the 100 machines into perspective with your scenarios. It seems a bit that you're then trending toward the mid case of this area for '24, '25. I think you always said that in the period, it's predominantly logic. So could you maybe link that to how you see the logic adoption? Is it like your mid case? Or is there maybe even a bit better than you initially had?

[Richard Blickman, executive]

Well, excellent, Marc. Well, as always these questions, looking forward, are difficult to answer precisely. But the trend is very clear. As I said earlier, the adoption for high-end logic is ever more confirmed. So it started initially, as we all know, in Taiwan. And then quarter-by-quarter, it increased, and more customers were engaged in also qualifying this technology.

The next adoption in larger volume with this U.S. customer, increased adoption with a broader customer range in Taiwan. Also at certain subcontractors, they are preparing to offer this technology to the market.

So the adoption so far is -- and we've said that every quarter, I think, is according to the model we shared some 3 years ago, where, yes, you have an initial phase and you could expect a major mainstream market adoption as of '25, '26, and then 3 specific applications, one logic and then we have the HBM where it's still unclear where the inroad will be, whether it be at the high end of 4 or will it be only at 5. And then we have the adoption for chiplets and especially the chiplet supplied in high-end smartphone applications.

And logic looks today that it follows what we put out as a model and having hit no roadblocks so far. So it follows nicely and maybe somewhat better, but that's definitely the case.

That's what it is. So far, continued adoption, broader customer base, adding new customers, all preparing to own this technology eventually for mainstream applications, that's how you can best characterize what's happening. I hope that answers your question, Marc.

[Marc Hesselink, analyst]

Yes. Yes, that's clear. And the second question is actually in the 100 that you now communicated for, let's call, the volume production. Is there also already the OSATs in there? Or is this -- or is this more -- are these just testing at this stage?

[Richard Blickman, executive]

As I tried to explain, there -- many are evaluating testing samples in preparation. There's one OSAT who was early on. But then I should add the center of excellence in Singapore, applied center of excellence, where we have this hybrid bonding application ongoing, same in our own hybrid bonding center in Singapore, they're next to each other, and we sort of split the work because of the quantity of work, and that is continuously increasing.

So that's a center where all those customers do test for a variety of devices. And not only for logic, memory and -- but especially those chiplets, where they expect a use in many end applications, which simply use the smaller design technology.

[Operator]

We will take the next question from line Adithya Metuku from HSBC.

[Adithya Metuku, analyst]

So firstly, I just wanted to understand about this capacity deployment. Now one of your peers in Korea, they're talking about building capacity for [ 4 20 ] TCB bonders a year by 2025. And this has led some to conclude that the leading HBM vendor will stick with TCB in the HBM for much longer than many are anticipating. So I just wanted to think about how you think about the risk of ending up with overcapacity, with the capacity deployments you're talking about today? And then I've got a follow-up.

[Richard Blickman, executive]

Well, that's an interesting question. Yes, it's always possible that competitors take the entire market. So then our capacity is surplus, can also be the other way around. [indiscernible] detail.

[Adithya Metuku, analyst]

Understood. And maybe just tied to that, there's been recent articles talking about Samsung using hybrid bonders from its subsidiary, semis, to show 16 high HBM. And within the Taiwanese ecosystem, there's a company called Soltec, making a lot of seems like progress on hybrid bonding tools. So I just wanted to again hear your thoughts on how you think about your market share trending within hybrid bonding and the resulting impact on profitability as you look forward?

[Richard Blickman, executive]

Well, also, this question is hard to answer. Is this a conference call about our competitors? Or is it about us? I don't know your information.

[Adithya Metuku, analyst]

Sorry, I'm just trying to understand your prospects by assessing your competition. But yes, if you can't answer, then it is what is. I've got a few more questions, but I'll rejoin the queue given I've asked my 2 questions.

[Operator]

We will take the next question from the line Andrew Gardiner from Citi.

[Andrew Gardiner, analyst]

I wanted to come back to combining some of the answers you've given on TCB and hybrid bonding as well as the capacity expansion. Can I take it from your answers that really the visibility or I suppose the customer conversations you're having, that is giving you confidence to expand that capacity clearly coming more from the hybrid bonding side than the TCB side, particularly given what you described as the TCB orders so the qualification process has begun, but it's going to take better part of a year. And so really you don't have -- you're hopeful, but you don't have firm visibility there. So is it more for hybrid than it is for TCB that you have that flexibility depending on the market mix come this time next year, the year after, the year after that?

[Richard Blickman, executive]

Exactly. That's how it is. So driven by the ongoing, you could say, success of hybrid bonding demonstrated by orders received and also many installed, so applications are becoming ever more clear and being ready for, yes, let's say, more sizable orders than received so far. And at the same time being prepared for TC applications, which are also becoming more clear.

As indicated earlier, we have now demonstrated not only to be able to produce TC devices with 20-micron bond-pad pitch, but even successfully demonstrated below 10-micron bond-pad pitch. So there are many applications, which will be using either or. And so, yes, in a broader answer, it's for both.

[Andrew Gardiner, analyst]

Understood. And I suppose related to that, the point on TCB, where you're describing the qualification process, is it -- we shouldn't therefore be expecting much in the way of order announcements from you regarding TC until, I mean, maybe second quarter or more like the third quarter next year? Or could something happen sooner?

[Richard Blickman, executive]

Well, that's always the question. So the data I referred to is customer data. And that could well translate in orders. But the timing is always, yes, let's say, hard to -- but that's the same for hybrid bonding and the same for the conventional business. But it looks ever more solid. So time will tell.

[Operator]

We will take the next question from line Martin Marandon-Carlhian from ODDO.

[Martin Marandon-Carlhian, analyst]

My first question is on hybrid bonding adoption in the following quarters. I know it's a hard question to answer, but since your U.S. customer already ordered a significant amount in Q2, I guess it would be more driven by Taiwan now in the next following quarters. So what do you think will drive more capacity expansion in Taiwan now? Is it mostly about still AMD and the MI300, for instance? Or do you expect most of the capacity expansion in Taiwan now to be more driven by new customers?

[Richard Blickman, executive]

Combination, and it's an excellent question. And it's also -- one must remember, this Taiwan customer does not share any customer data. So either we hear that from the customer itself, and AMD is pretty open in sharing with the world the type of devices and also the advantages using hybrid technology, but there are also many other customers at that same Taiwanese customer qualifying for different applications.

And we can only, let's say, derive from the number of machines installed that, that is -- and also for the type of devices produced on those machines, that it can be a much more broader audience using this technology. So there is this tendency that the adoption is broadening, and that should then lead going forward also to a larger market using this technology.

[Martin Marandon-Carlhian, analyst]

Okay. That's very helpful. And if I have a follow-up. I also have a question on the system coming in 2025 with 50-nanometer accuracy, so what we should call Gen 2. Do you see your Gen 2 system would capture most of the hybrid bonding demand for 2-nanometer and below logic applications? And second element, is it fair to assume that this Gen 2 system will not be used for memory since the accuracy requirements seems to be lower?

[Richard Blickman, executive]

Yes. You're also very right in your assumptions. Memory is less critical in that sense than logic. So it tends to be more likely to use in next-generation logic below 2-nanometer design geometry. That is sort of what we hear.

Second half next year, the first system, which will certainly take 6 months, 9 months to qualify, being ready for that next generation, there's a lot of customer, let's say, pool. They want it early as possible. So that definitely is tied to this next-generation below 2 nanometers. They simply want to be ready. But then for logic, memory may follow quite some time thereafter.

[Operator]

We will take the next question from the line Martin Jungfleisch from BNB Paribas.

[Martin Jungfleisch, analyst]

I have 2 questions, please. The first one is just to come back on these delayed shipments. So just to confirm, you will still ship half of the big Q2 order in Q4 and the other half in the first half of next year. That's right, right?

[Richard Blickman, executive]

That's correct.

[Martin Jungfleisch, analyst]

Okay. Great. And that would imply that the mainstream assembly tool shipments or other tools like TCB would be down quarter-on-quarter in Q4. Is that also right to assume?

[Richard Blickman, executive]

Well, it's roughly, as we indicated, flat compared to Q3, plus/minus 10%, and that is the, let's say, uncertainty in current market. And that plus/minus is more related to the conventional market, then to the new technologies.

I hope that specifies it. So the uncertainty is not around the hybrid bonding.

[Martin Jungfleisch, analyst]

Okay. Great. And then the second question is on glass packaging. I mean there's been a lot of news flow using glass or glass panels as a substrate in packaging in the future. Just wondering how this would impact Besi, if this would require a type of bonder generation. How you're positioned there?

[Richard Blickman, executive]

Well, we're also definitely engaged in the development using glass substrates. It has many advantages. Glass, known for a long time, and it could become very important in these 2.5, 3D device technologies, where glass basically replaces organic substrate or interposer material. And it has, of course, an impact on how chips are mounted and also packaged. And any change is an opportunity. So in any case, very promising development for the future.

[Operator]

We will take the next question from Michael Roeg from Degroof Petercam.

[Michael Roeg, analyst]

Yes, Mr. Blickman, I have a question about the AMP market as a whole. You are typically so kind to share the latest TechInsights estimates, and you also warn every time that they're always in motion, and we see that again this quarter. But in the past 2 years and especially this year, I've noticed huge swings in the segments within the market.

For instance, dicing has been gaining considerable share, wire bonding has been coming down tremendously and die attach is somewhere in the middle. If you look at the estimates for '25, strong growth. And logic dictates that the segment with the weakest space will grow strongest, so that would be wire bonding, dicing would grow slowly and die attach probably again in the middle. Does this make sense? Or do we have extra insights from the segmentations of TechInsights to help us out a bit?

[Richard Blickman, executive]

Well, I'm trying already for nearly 40 years to understand market indicators, whether this is TechInsights, Future Horizons in the past, Gartner, but there also others should not forget [ Yole ]. Their models are based on overall market inputs, digested in different ways, tested with suppliers like ourselves.

And yes, there are always, by definition, incorrect, but the trends are very good to follow. So for next year, it could well be that wirebond has a strong recovery. And why? Because wirebond CapEx for -- as we know, the past many quarters, has been very soft. So that typically is a cyclical behavior. And that could well be for dicing.

On the other hand, dicing becomes ever more critical because dies are ever more complex, thinner. So that is directly linked to the Advanced Packaging segment. Same. If you take die attach, part of that world is conventional. And the other part is advanced packaging. So the conventional part is like wire bond, very slow already for quarters. So that could show a recovery.

But the focus is always on next-generation technology. And it's difficult to have that, let's say, segmented in an overall market view. It's only very recently that TechInsights is now also presenting data on advanced packaging. It used to be all in one market segment. But then, again, be careful, we very closely with our customers, as we explained many times, align our strategy and also the growth or the other way around, the stagnation directly with customers. Our top 10 represent 50% to 60% of revenue, and you should be clearly ready to service those customers in ups and downs.

[Michael Roeg, analyst]

Yes. Yes. No, I understand the problems. And the thing is, based on the midpoint of your guidance for Q4, you will be growing about 5% for the full year. That's slightly higher than the A&P forecasts from TechInsights, so we're gaining a little bit market share. But if I strip out the Japanese company that has no downturn at all, then you're growing much faster than your closest competitors. That is probably because of your hybrid bonding business this year, correct, that they don't have?

[Richard Blickman, executive]

Yes.

[Michael Roeg, analyst]

Okay. So actually, if that...yes, please?

[Richard Blickman, executive]

Yes. Sorry, please complete it.

[Michael Roeg, analyst]

So basically, if that recovery in mainstream, which has been taking so long, and I don't see it happening in Q1, perhaps not even in Q2 next year. Now you will probably continue to gain market share because of your focus on advanced.

[Richard Blickman, executive]

Maybe, but it's not only market share. The key is to be in the fastest-growing segments and volumes and the margins. So our focus is very clear, on the advanced packaging, on the mainstream applications, in high-end communication devices, smartphones, also high-end computing and also for that reason for high-end devices in automotive. And that is what drives the business. And in the end, the market share is a calculation on a model which is presented for a total by somebody like TechInsights. But anyway, those are statistics.

[Michael Roeg, analyst]

Okay. Then a final quick question. Your expansion next year, the clean rooms and so on, should we pencil in around EUR 20 million in CapEx in 2025?

[Richard Blickman, executive]

No, no. Less. Should be around EUR 10 million.

[Operator]

We will take the next question from line Alexander Duval from Goldman Sachs.

[Alexander Duval, analyst]

Richard, just a couple of quick questions. Firstly, just to clarify, you talked about substantial hybrid bonding orders in the quarter. Just wanted to understand what substantial means in comparison to the second quarter order of hybrid bonding, which is around 29 units. Did you see orders for a double-digit number of hybrid bonding units in the quarter?

And also just to clarify, could the order intake broaden in terms of customers in the fourth quarter beyond the 2 logic players we know about?

And then finally, as a second question, to what degree do you think the mainstream assembly market has now seen a bottom? To the degree that you do, what would underpin that just given there seems some uncertainty in areas like auto and industrial?

[Richard Blickman, executive]

Well, excellent. So 2 questions. One -- and we provide sort of a rhyme, a riddle. We indicated that we now have sold over 100 hybrid bonders. And as we indicated in earlier calls, in Taiwan by year-end, we should reach that magic number getting close to 50. And then in the U.S., we are only starting in volume. So the rest are customers with smaller numbers, and some are based for demand for early next year.

So your question in the third quarter, substantial means the largest quarter number of orders was 29 systems in Q2, a substantial means also substantial, not more than Q2 but double digit, and I can illustrate that in a way that you have to make a quick math to get to this over 100. So it must be more than 10 and maybe close to 20.

Anyway, so a substantial number and broader, ever broader applications testing for HBM stacking memory, for chiplet applications, so connecting multiple dies, both 2D and 2.5, 3D. So that is ever broader.

The next question, the mainstream bottom. It's with all the negative indicators right now and simply take our customers as a reference who have provided data in the past couple of weeks, and more will come in the next few weeks with quarter results. Don't yet demonstrate and, let's say, expected recovery soon.

Statistically, one would, let's say, conclude that there must be, at some point, a turn of tide because we're 8 quarters well into a correction. But as we said in the press release, we -- otherwise, we wouldn't formulate it differently, but we are still cautious in this next quarter.

All the outside, as mentioned earlier, TechInsights, but also Future Horizons and others, expect '25 to show then a recovery. We also expect this for '24, not anyway. That's as good as the information as we can get. I hope this answers your questions.

Any further questions?

[Operator]

We will take the last question from line [indiscernible].

[Unknown Analyst]

Richard, most of the difficult questions have already been asked. So ending this call with a simple question. Just on the recent news about the reduced or delayed front-end CapEx by 2 foundry players not being TSMC, perhaps you can just clarify whether the delay -- the slight delay in Besi's HB deliveries in Q4 is related to that? And maybe more in general, whether you expect any impact of these CapEx delays on your TCB or HP business next year?

[Richard Blickman, executive]

Well, the shipment, what we indicated for Q4 and early part of next year for hybrid bonders is not linked to -- and that was an earlier question. It's simply logistics. It is not caused by delay of front end.

Whether that delay has an impact further down the road, is not, let's say, clear. It's hard to estimate. But in this industry, the forefront of technology always continues. And that is how customers survive. So the efforts will be ever more steered towards either maintaining lead in the next-generation technology or catching up.

So you have to, as a supplier, always look at that part. So it can very well be that the delay or, let's say, the reduction in investments can be affecting less forefront technology capacities. But anyway, as I said earlier, for us, they don't share those details to such an extent. But it's full speed ahead on hybrid bonding and TC next-generation technologies.

[Operator]

There are no further questions. We will hand it back over to your host for closing remarks.

[Richard Blickman, executive]

Thank you very much all. And any further questions, don't hesitate to contact us. Thank you. Bye-bye.

[Operator]

Thank you for joining today's call. You may now disconnect.