BE Semiconductor Industries NV

AEX:BESI

Earnings Call Transcript 2024-Q1

[Operator]

Good morning and good afternoon, ladies and gentlemen, and welcome to Besi's quarterly conference call and audio webcast to discuss the company's 2024 first quarter results. You can log in to the audio webcast via Besi's website, www.besi.com. Joining us today are Mr. Richard Blickman, Chief Executive Officer; and Mr. Leon Verweijen, 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 would now like to turn the call over to Mr. Richard Blickman.

[Richard Blickman, executive]

Thank you. Thank you for joining us today. We will begin by making a few comments in connection with the press release issued earlier today and then take your questions.I'd like to remind everyone that on today's call, management will be making forward-looking statements. All statements, other than statements of the 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.For today's call, we'd like to review the key financial highlights for our first quarter ended March 31 and update you on the markets, our strategy and outlook.First, some overall thoughts on the first quarter. Besi delivered solid first quarter results in an extended assembly market downturn. Revenue of EUR 146.3 million was above the midpoint of prior guidance and represented a 9.7% increase versus the first quarter of 2023. Year-over-year revenue growth reflected strength in both 2.5D and 3D AI-related applications, partially offset by continued weakness in mobile and automotive markets.Order trends in the first quarter reflected a number of crosscurrents affecting assembly equipment markets currently. For the quarter, orders decreased by 10.1% versus Q1 last year and by 23.3% sequentially. Mainstream assembly markets continue to be soft, particularly for smartphone and automotive applications, despite increasing utilization rates generally. For smartphone applications, it reflects both ongoing weakness in Chinese markets and limited new product innovation this year. For automotive applications, it reflects excess assembly capacity after periods of strong growth over the past 2 years. We also noted a pause in advanced packaging order development this quarter, particularly for 2.5D and 3D applications after a strong ramp in the second half of last year, as customers installed new incremental capacity. We expect these orders to revive in the second quarter. Orders for photonics applications continued to be strong in the first quarter.Profit for the quarter was above expectations. Net income, adjusted for share-based compensation, rose to EUR 49.5 million, an increase of 15.1% versus the first quarter last year, with adjusted net margins increasing to 33.8% versus 32.2%. Profit growth was primarily attributable to increased revenue, combined with a 3.0 point increase in gross margins to 67.2% associated with a more favorable product mix and net ForEx benefits.Baseline OpEx were also on target relative to guidance, reflecting overhead cost controls and increased R&D spending to support next-generation product development.Our financial position continued to improve with net cash increasing by 60.1% from year-end to reach EUR 180.9 million due to strong cash flow generation and the conversion into equity of certain convertible notes. Since year-end, a total of EUR 127.7 million of Besi's convertible notes due 2024 and '27 have been converted into a total of 2.6 million shares through April 25. As of such date, their aggregate principal amount decreased to EUR 200.5 million.Next, I'd like to speak a little bit about the current market environment and our strategy. It appears that the recovery of the assembly equipment market in 2024 is progressing more slowly than previously anticipated due to continued excess capacity conditions in a number of our end-user markets. We are in an extended downturn for many applications lasting 8 to 9 quarters, including the Chinese market. Industry analysts now expect the upturn in mainstream assembly applications to be more second half 2024 focused. Similarly, TechInsights has recently reduced its assembly market growth estimates for 2024 to 16% from 31% in December. Growth is now more back-end loaded for the year. However, it also increased its estimates for '25 and '26 and expects the next market peak in '26 at USD 7.6 billion, up 73% from 2023 levels.Our principal strategic focus currently is the expansion of Besi's advanced packaging market share in 2.5D, 3D and photonics applications to meet strong growth anticipated for AI-related applications over the next decade. Our advanced packaging prospects continue to be favorable based on customer investment plans for such applications, particularly in the areas of hybrid bonding, CoWoS and photonics assembly. We anticipate orders for 25 to 35 hybrid bonding systems in Q2 from multiple customers, substantially all of which are for Besi's most advanced 100-nanometer accuracy generation. As such, we are increasing our R&D investment in each of these assembly processes to take advantage of growth anticipated in the period 2024, '25, '27.We have also expanded our resource commitment to next-generation TCB systems. We see a parallel path pursued by leading memory customers for the adoption of both hybrid bonding and next-generation TCB assembly processes in order to meet the significant demand for high-bandwidth memory necessary to support AI-related capacity growth. In addition, we received a follow-on order for our new in-line flip chip system for CoWoS applications, shipped a TCB Next system for evaluation to a second customer, and received indications of interest for additional systems for multiple customers.Now, a few words about the guidance. For the second quarter 2024, we forecast that revenue will be flat, plus or minus 5%, versus the first quarter of this year, with gross margins between 63% and 65%, based on our projected current product mix. Aggregate operating expenses are forecasted to decrease by 15% to 20% versus the first quarter due to a reduction in share-based compensation expense. At the midpoint of guidance, we see H1 '24 revenue relatively flat versus H1 '23.That ends our prepared remarks. I would like to open the call for questions. Operator?

[Operator]

[Operator Instructions] The first question comes from the line of Alexander Duval from Goldman Sachs.

[Alexander Duval, analyst]

You talked about hybrid bonding recovery in the second quarter in terms of orders. Just wondered if you could help us understand, should we be expecting orders for the third quarter as well as the second quarter, and should that be at a similar level? If we look at consensus for this year in terms of shipments, people seem to be expecting around 50 hybrid bonding shipments for the year, and your expected second quarter order intake would cover well over half of that. So I'm just trying to understand to what extent the second quarter bookings for hybrid bonding are a one-off or part of a sustainable trend.And then, secondly, I'd like to ask about the memory application of hybrid bonding. We've seen some positive commentary from some peers talking about TCB being used on HBM4. I wondered if you could give us your updated expectations on usage of hybrid bonding for HBM5, and whether speculation that it won't be used on HBM4 actually has any bearing on usage for HBM5 further down the line.

[Richard Blickman, executive]

Excellent. Thank you, Alexander. First, the indication of orders to be received in this quarter is not a one-off. That is part of a program, establishing a capacity for logic devices mainly, which follow an investment program, which we have shared with the investment community since basically end of '21. So, to explain that again a bit, as we all know, Taiwan was the first where we're now in the mid-30s installed base. The U.S. followed a bit later, some delay, as we know. We did receive the first batch in the middle of last year, and most of what we indicated is expanding that capacity. And that then brings the installed base to slightly over halfway from the indicated 50. So what we see is a pattern at the 2 major customers so far for logic is building up a capacity to 50 machines, 50 systems. There's only a difference significantly between both approaches. Taiwan is a stand-alone environment. In the U.S., the concept is more integrated lines. So, that's slightly a different approach. The reason behind that is probably to achieve a more, let's say, highest level cleanliness process environment. But again, this is -- we are at the beginning of the adoption of hybrid bonding in an ever-broader environment. So we have logic. We have, tied to logic, other applications being tested, and then we have the HBM, the next step. So, if we take the graph from our investor deck, where we see a trajectory for the coming years until 2030, the first phase is logic. In the middle is the adoption for hybrid bonding -- for HBM, so more about that in a second. And then, we have also the next application, more in [ chip flip ] solutions for high-end smartphone communication. That trajectory still is followed in every sense.On the second question on what's happening in the HBM environment with announcement recently that JEDEC may allow a higher stack, first of all, why is that requirement? That requirement is pretty easy to understand. This higher stack allows with the different processes used to move hybrid to have, on the firsthand, TCB extended for 12 layers, 16 layers maybe. And what is the effect of that is that probably there's a significant demand in the years to come, all related to AI, so that the industry can use both technologies. So what we see at the major memory manufacturers producing HBM, as we said in the press release, but also in the call, we see side-by-side, which is not new. Remember our comments on Micron orders in the previous quarters, and that is similar to the 2 Koreans evaluating the benefits of hybrid bonding in performance of stacked memories compared to stacked memories using TCB. There are certain advantages certainly using hybrid bonding. Number one is the performance of the memory pack and the energy consumption, and that is very important in applications in data centers, so for heat dissipation because the more energy you need, the more heat you have to dissipate. And along these 2 technologies over time, the current view is HBM4 and TCB early versions using hybrid bonding. And as of HBM5, the world today expects that to be overwhelming hybrid bonding. So there may be a small shift because of this increased height in using an HBM4 and TCB, but that doesn't take away the major advantages of using the hybrid bonding process. So that's, in a longer answer, where we are today.

[Operator]

The next question comes from the line of Charles Shi from Needham & Company.

[Yu Shi, analyst]

Richard, I want to ask a little bit more about the hybrid bounding orders. You're expecting, I think, 25 to 35 machines ordered in 1 quarter. That's probably a record for you guys. But wondering if you can provide more color, who is ordering those machines, for what kind of application? You kind of alluded to maybe it's the U.S., but I want to confirm. And when do you expect to recognize revenues on that 25 to 35 machines?And lastly, related to this, your Asia base, the competitor, I think yesterday, they talked about shipping 2 hybrid bonding systems in Q4 this year to an unnamed customer. Wonder if you can provide any thoughts on the competitive dynamics there. And what's your thought -- is there any threat from that competitor at this stage?

[Richard Blickman, executive]

Well, to start with the last, there's no threat. We cannot, let's say, confirm or we do not know where these 2 machines are going to be installed. So that's not at our customers today. Your first question, about half of that is shipped in Q3, Q4, but most is Q4, and also the other half is Q1 next year. But as I answered to the earlier question, this is part of a program, which -- or most of it is part of a program, which we have shared over the many quarters. So it's not a one-off situation. It simply fits into the projected capacity build over a period of 2 years. And there are other orders for other customers for development, and certain add-ons also in Asia, where our major capacity is established at this moment. So that's, in a nutshell, where this is all about.

[Yu Shi, analyst]

Yes. Richard, if I can follow up, so as part of the program, this is -- I assume this is an integrated product line, right?

[Richard Blickman, executive]

That's supposed to be -- a part of it is for integrated line applications. Part of it is also additional stand-alone machines. But the key message is that we are seeing continued expansion in different products, new product generations, which are using this hybrid bonding technology. That's the main message. And in high volume, and of course, there's continued development. The key question is, of course, when is the adoption in hybrid bonding for HBM?

[Yu Shi, analyst]

Okay, got it. So maybe my second question, I'll use that to ask you about HBM. So it's a little bit confusing to us. Hopefully, you can shed some light what your memory customers are trying to do here. I believe your U.S. memory customer, they have a hybrid bonding tool already. They are getting a TCB next from you -- they already got one from you as well. And then, I think they got -- they [ $16 million ] worth of foundry tools from HANMI, one of the Korean assembly equipment makers in HBM earlier this month. So it's a little bit confusing. That may be just 1 out of the 3 customers, but the whole memory industry, what exactly they are planning on in terms of HBM? Can you kind of walk us through once again? I know you probably answered this question from a different perspective earlier. What are they trying to do here? And what is Besi's strategy, specifically on TCB for high HBM applications?

[Richard Blickman, executive]

Excellent. Well, from our perspective, the recent announcement you referred to is for applications in the HBM3. And the development for the next HBM4, the higher end, and certainly 5, the key is to determine what I tried to explain earlier is what are -- what is the best process for both? So, is that possible to use a next-generation TCB, which we, as you confirmed, have delivered? And where -- and which products will be produced using hybrid bonding? And the key, we understand in the discussions, is simply the device performance. And although there's a higher cost supposedly using hybrid bonding compared to TCB, the end question and answer is the performance of this hybrid bonded pack is that significantly superior to a TC connected pack, and that should drive the demand one way or the other. Again, we have both systems available meeting the specifications in every way. So for us, it is, on the one hand, of course, interesting to see this go into hybrid bonding because that, to our view, and we are just confirming what customers tell us, that should be for HBM5 onwards, the mainstream, and in between, we have certain device families, which will use TCB. Does that help?

[Yu Shi, analyst]

Yes, very helpful.

[Operator]

The next question comes from the line of Madeleine Jenkins from UBS.

[Madeleine Jenkins, analyst]

I was just wondering what sort of timeline you're seeing for hybrid bonding adoption within smartphone. I know last quarter, you talked about a potential [ pull from ] Qualcomm. Is that still the case?

[Richard Blickman, executive]

Well, it continues along a similar, let's say, timeline. As we discussed in previous quarters, there's enormous development going on in certain versions. And there's also the simple question, does hybrid bonding in a certain chiplet content offer significant performance improvement versus a mass reflow or TCB process solution? There have not been changes in a way what happened in the HBM stacking. There is not, let's say, a different standards being development -- developed to extend conventional interconnect technologies like flip chip or TCB for that matter. So again, continuation on the same trajectory as we have reported in the past quarters.

[Madeleine Jenkins, analyst]

So is that sort of a '25, '26 time frame?

[Richard Blickman, executive]

Well, development ongoing in '25. As we indicated in the Capital Markets day last year, our view was '27 for a major mainstream adoption. There are some who more, yes, let's say, have views on earlier adoption in '26. That's still difficult to really understand. So this year will be critical to see what the progress is, which technologies are going to be used for next generation, and we can only say that we are involved in all of these developments, which ultimately will offer opportunities, either in continued mass reflow TC or into hybrid bonding. Don't forget, the cost factor is very important in those decisions for consumer end products. It's hard to really understand the delta between that cost. There are factors mentioned, [ factor 4 ], hybrid bonding more expensive than, if possible, using TCB. That certainly varies based on die size and which circuitry, but the performance is significantly better. Simply take the ongoing data sharing by AMD to the market on their experience on the MI300, which supposedly has a much improved performance compared to using older technologies. So, that is not much different using hybrid bonding for other processes than a certain logic device. But we'll share every quarter the progress.

[Madeleine Jenkins, analyst]

And then, just as my quick follow-up, could we have an update on your generation 3 hybrid bonding tool? When do you expect that to come to market? What sort of ASP, and what sort of applications? That would be useful.

[Richard Blickman, executive]

Well, it has to be ready for the market end of next year. That's an excellent question. That's also one of the major data points we all have to observe. The fact that there is a lot of pressure on having that available for the market by the end of next year also tells you that the hybrid bonding technology is critical for next-generation device architecture. Otherwise, that pressure would not be existing. So, launching that product end of the year. Price range, we indicated, is higher, of course, than generation 1 plus. We mentioned earlier between [ EUR 3 million and EUR 4 million ] based on what options are included in that system. But again, the -- let's say, in the past quarters, the need for below 100 nanometer becomes ever more clear, and that is very important also in where this industry is heading.

[Operator]

The next question comes from the line of Didier Scemama from Bank of America.

[Didier Scemama, analyst]

I'm going to be my annoying self. Since you very kindly gave us a guidance, which is I think a bit unusual, for your hybrid bonding unit bookings for Q2, can you just tell us a little bit about how many units you booked in Q1? And secondly, obviously, going back to an earlier question, do you expect that strength in bookings to carry on into Q3 and Q4? And I've got a follow-up.

[Richard Blickman, executive]

Well, we mentioned in the press release that the orders for production systems were not -- we did not receive orders in the first quarter. We did receive for evaluation and testing, but not for major production. You could say that the orders which are scheduled could have been placed in Q1 as well. But your question on what happens in the second half, if all goes according to the schedules we have shared in many previous quarters, the volumes are gradually built step by step, and that is often an order quantity of 5. Sometimes when it's, yes, more strategic, it is a batch of 10 systems. And so, as you already noticed, between 25 and 35 is a significant expansion and confirmation that this technology is used for devices, which will hit the market not long thereafter.Will this continue Q3, Q4? It certainly looks more promising as quarters pass by. I can't give you any numbers in terms of what we expect, but we will certainly share more about the market outlook at the Capital Markets Day on June 6, and you're most welcome to join in Radfeld, where we also show you the latest developments on the hybrid bonders, the progress on the 50 nanometer, and also our TC system, which we shipped the first one successful mid of last year. The second one is shipped recently and being installed. So, that should give all of us the opportunity to get more understanding of both technology directions, applications and also market sizes.

[Didier Scemama, analyst]

Excellent. So, on the Q2, can you just give us a little color on the sort of end customer? Not the names, of course, but whether it's coming from DRAM customers or foundry logic customers, those 25 to 35, just a rough split. I guess, the background to my question is to sort of gauge your level of confidence really for the adoption of hybrid bonding in '25 in HBM because by the looks of it, it's a bit more uncertain, and I think it's not new. That's been sort of discussed in the press for quite some time. But just to try to understand, are there a significant part of your bookings for Q2 coming from DRAM customers? And I've got another quick follow-up, if I may.

[Richard Blickman, executive]

No. As we answered earlier, the majority is for logic, which is not, let's say -- we have even -- or let's say, also before this JEDEC norm to increase the height, which has the, let's say, possibility for TC to be longer used in stacking memories. '25 is certainly -- or let's say differently, every -- each one of them is developing this technology side by side because there is a big advantage using hybrid bonding in terms of the performance, and that is definitely always the key to technology. On the other hand, TC has still a cost advantage over hybrid bonding. As we said, in many cases, we can influence that cost equation to a certain extent by developing ever-more mature systems, which, in the end, produce more devices per hour. That immediately impacts the cost. So we have moved, in the past 3 years, the needle gradually from below 1,500 UPH to 1,500 UPH, now beyond 1,500 towards 2,000 UPH. And there are also many other factors. Remember, it has to be ultra clean, so you need die preparation, you need all kinds of process requirements, which all have to become mainstream. TC is an extension of flip chip, where we built on an experience of over 30 years. But the disadvantage is simply the performance of the memory pack. So every customer, all 3 of them, are developing this side by side. And in the course of '24, and certainly in '25, it will become more clear which devices will be produced using one or the other. That's also, yes, fitting exactly our strategy. We are ready for that. Otherwise, customers would not want these machines. So, exciting times. And I can't tell you which volumes in which directions are to be expected. We can make models.But one of the very important things to answer from this direction, if we would not see continued orders and in major volumes for logic application in this hybrid process, that would also impact the adoption in the memory space. So every single day, 24/7, the experience in data from using this hybrid bonding is building. And that confidence is also supporting the case that it will be used in memory, plus also the adoption for high-end smartphone devices. And those road maps have not changed. So I've had many calls and discussions that, oh my God, with this JEDEC standard change, you can forget hybrid bonding. That's not at all the case. It extends the use of TC as an interconnect process for some more time. But the next step, and that's the earlier question, when are you introducing the 50-nanometer, the second generation, so after 100-nanometer, which is very much pressured to have that ready by the end of '25, and that simply confirms the need for this hybrid bonding going forward. Anyway, a longer answer, Didier, I hope...

[Didier Scemama, analyst]

Very much so. Maybe my final question. One of the sort of leading AI accelerator company has announced a product which is made of 2 chips, 2 dies, I should say. And I think it's not entirely certain whether you are sort of supporting that potential new product or that new product. Is there anything you can talk about this?

[Richard Blickman, executive]

Well, that's in the 2.5D, if I understand you correctly. We are certainly involved in that. And if you recall, we had significant orders in Q4, less in Q1, but that is simply installing that capacity. And we will see in the next quarter, a next round. But we are very much involved in that technology.

[Didier Scemama, analyst]

And it's a high-end flip chip system, is it?

[Richard Blickman, executive]

Yes. It's partly flip chip. There's some question about the need for a TC process because when they have a certain size, the warpage, so keeping the device flat, becomes an issue, and you can support that by using a thermal compression bonding. So a bit of compression. But anyway, cost is key. So if you can do that with a reflow process flip chip. Yes. And we did mention we received another order this quarter, but one, not as many as in Q4. But anyway, so that's also a very, yes, let's say, growth prospect in the years to come.

[Operator]

The next question comes from the line of Robert Sanders from Deutsche Bank.

[Robert Sanders, analyst]

My first question is just about SoIC-P, which is TSMC's TCB variant. I was just wondering if you thought that smartphone applications might use that TCB variant of 3D IC first, and then they would move to hybrid bonding later. And I have a follow-up.

[Richard Blickman, executive]

Yes. We also -- and this is also published in certain conferences. There's a conference coming up in 2 weeks in Thailand, which will share probably more information about this. We are also involved in the evaluation of both, so TCB and hybrid bonding for that application. And that is the answer to when will hybrid bonding be used in high-end smartphones.

[Robert Sanders, analyst]

I got it. And then just on the Taiwan customer, you said, I think, you've shipped 35 tools, hybrid bonding tools so far cumulatively versus the 50, the initial build-out. I was just wondering when you thought that you would see the kind of second wave in hybrid bonding at the Taiwan customer.

[Richard Blickman, executive]

Well, if we follow the public announcements and, let's say, the progress reported in these conferences, it could likely be a next round. The number is never really known in the second half of this year. So the bottom line is that the adoption is gaining confidence, yes, I would say, day by day, month by month, certainly. So that opens up other applications next to what is already produced right now.

[Robert Sanders, analyst]

If I could just sneak in one quick question. Your Q2 is normally seasonally up, if not this year. Q3 is normally a seasonally slow quarter just because of summer holidays, et cetera. I was just wondering, is there any particular reason why you wouldn't see the normal seasonal sort of softness in Q3 before the hybrid bonding orders land in Q4? Or is there something that is unusual about this year?

[Richard Blickman, executive]

The unusual -- well, it's not unusual. The industry is cyclical as we all know. The fact that there is a delay in recovery of the next cycle probably caused by overhang from the previous cycle and COVID. Don't forget, our revenue is still predominantly 80% is conventional business. So that's the answer.And the softness in high-end smartphones, that is an impact and automotive is also very slow. So the world expects the tide to turn somewhere in summer period for improvements in the second half. We've seen that historically as well. So that could -- with, yes, the continued business in hybrid bonding, yes, provide a model which is simply second half loaded or however you want to call it.But there's high uncertainty in a sense that we see utilization rates rising, we see disappointing end customers reporting apart from, yes, for instance, TI because it was a good message. Others are disappointing. So that tells you that the industry is still in a careful improvement. You see that in our numbers, improving year-over-year. But still -- and there's a lot more to come.

[Operator]

Your next question comes from the line of Nigel van Putten from Morgan Stanley.

[Nigel van Putten, analyst]

I have 2 questions on the order trends seen in the last quarter. First, more of a follow-up on the TCB Next system. So congrats on shipping to the second customer. But could you provide some more color around the indications of interest you received? What type of customers and what type of applications are these? And maybe a hint on time lines in terms of shipping another 2 for qualification to any of these customers? That's my first question.

[Richard Blickman, executive]

Well, what's very important in this year, and we have indicated this also last year, if all goes well, there's always, for memory, a higher volume application potential than for logic. By nature, there are more memory devices than logic. So the question is -- well, let me go one step back. Our TC Next has tighter specs than other systems currently available in the market. So it is the next generation. That's why it's called TC Next.And based upon the customer specifications, this year, it will become clear how much that inroad can offer us. You have 2 tracks. One is, yes, of course, the logic where we ship the first system, ready for a next generation with smaller [indiscernible], the other one for stacking memory with tighter specs. So those are our applications at logic customers and memory or you could say 2 memory customers, 1 U.S. and 1 Korean.

[Nigel van Putten, analyst]

Right. And when you say logic customers, that could also be 1 U.S. and 1 Asian foundry?

[Richard Blickman, executive]

Yes, in fact. And that is for other customers, of course, of that customer.

[Nigel van Putten, analyst]

Substantially 4 customers here. Okay, that's interesting. Second is also the orders in the first quarter or the absence of orders for 2.5D. I think 3D, we covered extensively now, but what about, yes, 2.5D, because in the press release, you sort of heavily hint on a resumption in this quarter. Could you already provide some color? Are you adding new customers? Is it a system or a customer like doing another round? And also...

[Richard Blickman, executive]

It's another round. So as I tried to explain to the earlier question, we have shipped several systems very successfully. And the question is, is the next round?

[Nigel van Putten, analyst]

Got it. And could you provide maybe some color on the potential size? I mean I'm trying to be greedy here, but it's very helpful sort of doing the math on the 3D side. Would this be comparable? Or should we not -- should we expect a lower overall number for 2.5D rather to the 3D?

[Richard Blickman, executive]

Well, I would guess similar than what we shared last year. But again, on the Capital Markets Day, we will share some more details, background about market sizes, potential market sizes. But for now, I think it's about similar.

[Nigel van Putten, analyst]

Got it. And then maybe a quick follow-up on the previous question by Rob. I think sort of getting the seasonality right would be helpful. Because in terms of getting the industry going again, I can see that there's more activity towards the end of the year typically. So would that also mean that you'd expect more meaningful shipments to sort of make that fourth quarter timeline?So it could be 3Q, but potentially more 4Q loaded. I realize you don't actually have the orders or the timelines, but would that sort of conceptually makes sense so we don't get carried away in terms of our modeling for the third quarter?

[Richard Blickman, executive]

Well, the best way to look at this is what does the industry expect for 2025? If that is a major, let's say, uptick, then ordering should start in Q3, Q4 of this year, certainly Q4. You can look at 2019, the last downturn. Also, the industry started to improve by Q3 and Q4, and that could be a similar pattern this time. And we had the abnormality of COVID hitting the world in February, March. But then it significantly increased also in a cyclical way. So that is a way to look at a potential expansion model.

[Operator]

We have time for one last question from the line of Ruben Devos from Kepler Cheuvreux.

[Ruben Devos, analyst]

I just wanted to continue on sort of the TCB versus hybrid bonding technology. Obviously, you've got the lead in hybrid bonding, and you're working hard to get that 50-nanometer tool out by the end of 2025. But I hear you saying also for TCB that you have basically the tighter specs than other systems currently available in the market.Many of the memory chip makers are doing this parallel development. I'm just trying to get a gauge of what's the opportunity for you in TCB, yes, apart from obviously the second customer ordering and much more interest. But what is basically also your capacity when you think about how many units of these you can deliver?

[Richard Blickman, executive]

Well, similar to units like hybrid bonding, the model for hybrid bonding as shared in the past with 15 systems per month, that is also capacity theoretically available for TCB. We also mentioned ASP-wise, there's not much difference. So theoretically, we could deliver in similar numbers. But that, of course, depends on how much will be produced using TC versus how much is produced using hybrid bonding. But that's [indiscernible].

[Ruben Devos, analyst]

Okay. And then just another totally unrelated question. I think in the past, you talked a bit about micro LED. And I think the assembly and transfer process can be quite a challenge in micro LEDs. There was some major industry news a few weeks ago of a U.S. consumer electronics supplier who terminated its project. I was just thinking is this becoming a bit of an opportunity for you, let's say, by the late '20s or how do you -- what are your thoughts on micro LED actually generally at this point?

[Richard Blickman, executive]

Well, we have been involved in the micro LED development from the beginning. We have been fortunate in a way that we have chosen to be more cautious in the requirements because -- well, it's a very good question. Why and what is Besi doing? We -- our strength is to build die attach systems. And we are not experts in other processes than die attach packaging and plating.So if you look at micro LED requirements, it needs -- it forces you to invest also in other technologies to produce a complete micro LED. We held off on that. So far, that looks like a good decision. On the other hand, there's a lot of technology benefits in moving micro LED technology forward, maybe in a different way. It's very important to understand that with our relationship with Applied Materials, we also have a better view on what is happening in that market. And once there are opportunities available where our technology can be used, then we certainly look at that seriously.But the simple conclusion is, and you mentioned that, that the attempt so far and to produce the micro LEDs with one by one is the most simple way to see that in the eyes of that major U.S. customer is currently not feasible. Are there other ways to skin this cat? Maybe and others are looking at that as well. You could think of a matrix as opposed to individual LEDs. So anyway, technology is always exciting. And especially when it becomes smaller, there are always opportunities.

[Ruben Devos, analyst]

Okay. Great. And then just if I can squeeze in a final one just on the R&D investment. I think you talked about it in the release that you want to see some of these opportunities with the various assembly processes, and you talk about '25, '27 time frame. I think if you think of the operating margin guidance in your business model, it's about 30% to 45%, which is quite a broad range, of course. But is it fair to say that maybe we could see a step-up in R&D and that basically, if you have some of these significant gross margins, which we've been seeing in the past few quarters that gives you more headroom to also increase your R&D spend?

[Richard Blickman, executive]

Absolutely. I think it's fair to say that with the margin development over the past 15 years or even more, we have developed a business model which is ever more self-sustaining, which is unique if you look at all the transitions to ever smaller geometries. So far, we have, I think, chosen the right technologies and the right customers for that. But as things become ever smaller, the amount of R&D to spend increases year by year.5 years ago, it was about EUR 40 million per year. Now it's EUR 65 million, supported by strong gross margins, a very good way to look at that. And that won't change. It will become more difficult if we don't sustain these gross margins. But anyway, so far, that all looks fine.

[Operator]

That is all the time we have for questions today. So I will now hand the call back over to Mr. Blickman for any closing remarks.

[Richard Blickman, executive]

Thanks, everyone, for attending the call and the questions. And if you have any more questions, don't hesitate to contact us. Goodbye.

[Operator]

Thank you for joining today's call, and you may now disconnect your lines.