BE Semiconductor Industries N.V. (BESI) Earnings Call Transcript & Summary

Good morning and good afternoon, ladies and gentlemen, and welcome to Besi's Quarterly Conference Call and Audio Webcast to discuss the company's 2023 First Quarter Results. [Operator Instructions] Joining us today are Mr. Richard Blickman, Chief Executive Officer; and Mr. Leon Verweijen, Senior Vice President of Finance. [Operator Instructions] I'd like to turn the call over to your host today, Mr. Richard Blickman. Please go ahead, sir.

Thank you, George. 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 would like to remind you that some of the comments made during this call and some of the answers by management in response to your questions may contain forward-looking statements. Such statements may involve uncertainties and risks as described in the earnings release and other reports filed with the AFM. For today's call, we'd like to review the key highlights for our first quarter ended March 31 and also update you on the market, our strategy and the outlook. First, some overall thoughts on the first quarter. Besi's first quarter operating results were within guidance and in a challenging industry environment. Revenue of EUR 133.4 million decreased by 3.1% versus the fourth quarter of last year, primarily reflecting general weakness in high-end and mainstream computing markets from both IDMs and Asian subcontractors and a temporary reduction in shipments for automotive applications. Such adverse influences were partially offset by increased shipments for high-end smartphone applications. On a year-over-year basis, revenue this quarter declined by 34.1%, reflecting the impact of the current downturn on Besi's business. In addition, demand from Chinese customers remained weak, representing approximately 27% of total revenue this quarter. Our trends in the first quarter reflect both the negative effects of the current industry downturn and the favorable prospects for Besi's wafer-level assembly portfolio. For the quarter, orders of EUR 142 million declined by 21.3% versus the fourth quarter, primarily due to the pull forward of some high-end smartphone bookings into the fourth quarter and, to a lesser extent, decreased orders for mainstream computing and hybrid bonding applications. Orders for smartphone applications for the most recent 2 quarters significantly exceeded those for the comparable period of the prior year but were less than previously anticipated. Demand for automotive applications improved in the first quarter versus both the fourth quarter last year and the first quarter last year. We received a EUR 10.5 million order from a Chinese subcontractor subsequent to quarter end for automotive applications, indicating ongoing strength in this end user market. Bookings this quarter also included orders from multiple hybrid bonding systems with follow-on orders anticipated in the second quarter from multiple customers, including 2 new customers. In addition, the first order for our TCB chip-to-wafer system was received subsequent to quarter end from a U.S. customer. Besi's profitability was solid in the first quarter with gross margin reaching 64.2% due primarily to the current product mix and cost control efforts. Operating income of EUR 41.7 million exceeded the midpoint of guidance. Excluding share-based compensation expense basis, net income and net margin were EUR 43 million and 32.2% this quarter versus EUR 42.3 million and 30.7% recorded in the fourth quarter last year despite lower revenue levels. As seen in this next chart, our financial performance has significantly improved at this point of the industry cycle as measured by a comparison of Besi's results for the last 12 months ending March 31 this year, versus the comparable period ending March 31, 2019, with revenue, orders and operating profit increasing by 44.7%, 66.5% and 82.7%, respectively. It is also a positive indicator for business target achievement in the next industry upturn. Besi has significantly increased its capital allocation in recent years. In 2022, we distributed EUR 416.3 million. To date, in this year 2023, we will have distributed almost EUR 300 million, including the dividend for fiscal 2022. Progress also continues on our EUR 300 million share repurchase program under which we spent EUR 77.8 million in the first quarter this year to purchase 1.1 million shares. This brings total purchases under the program to 2.9 million shares at an average price of EUR 59.33 for a total of EUR 174.1 million. Our financial position remained solid at quarter end, with cash and net cash of EUR 644.9 million and EUR 325.8 million, respectively. In addition, cash flow from operations rose strongly to EUR 61.4 million -- in this quarter, up 36.4% versus the first quarter last year. As a result of recent conversions, there remain outstanding only EUR 600,000 and EUR 13.4 million of our 2023 and 2024 convertible notes. Next, I'd like to speak a little bit about the current market environment and our strategy. As seen in this next chart, industry conditions have continued a downward trend since the second quarter last year, highlighted by slowing memory, mainstream computing and data center markets, continued weakness in China and CapEx reductions made by many of the largest semiconductor producers. The outlook for 2023 remains negative as we work through the current industry downturn. Analysts have further lowered market estimates for this year, while forecasting a renewed upturn starting in the second half of this year. We are prepared for various industry scenarios with a highly flexible production model and strategic initiatives in place to enhance profitability regardless of the industry direction. In the interim, we continue our investment in R&D programs and service/support in preparation for the next upturn and the significant projected growth of our wafer-level assembly portfolio over the next 3 to 5 years. We are currently updating Besi's strategic plan with the aid of an external consultant. At present, the principal focus is on R&D as we prepare our existing and wafer-level assembly portfolios for the next industry upturn. Progress in hybrid bonding and other wafer-level assembly adoption continues at pace with yields improving at customer sites and increased interest received from the development community and customers primarily for data center, mobile and memory applications. We are working closely with AMAT on the commercial development of an integrated production line, which shows significant promise at the first to customer locations. From an operational perspective, our principal focus currently is adjusting headcount, supply chain and production overhead to align with challenging industry conditions. In addition, we continue to work to complete new cleanroom space in Singapore and a new tooling facility in Vietnam by year end to support future growth opportunities outside China. Now a few words about our guidance. For the second quarter this year, we forecast that revenue will increase by 15% to 25% versus the first quarter this year, with gross margins remaining elevated at 62% to 64%, due primarily to our projected product mix. Operating expenses are forecasted to decrease by 0% to 5% versus the first quarter as lower sequential incentive compensation expense is partially offset by a one-time strategic consulting expense. That ends my prepared remarks. I would like to open the call for questions. Operator?

[Operator Instructions] Today's first question is coming from Madeleine Jenkins calling from UBS.

So my first is on hybrid bonding. I was just wondering if you could quantify the orders you received this quarter. And you also mentioned that the release 2 new customers that will order in Q2. So should we be modeling a kind of significant ramp-up from Q1? And also quickly, ASMPT mentioned in their release today that about a kind of first customer order for hybrid bonding. I just sort of wanted to hear your take on how that tool may compare to yours? And more generally, how you see your technology lead versus competitors? And I have a quick follow-up, if I may.

Yes, we had in this quarter, several follow-on orders already for those customers where systems are installed and then we have a very clear indication that we will receive in the next week or 2, orders from 2 new customers. So that tells you the proliferation of hybrid bonding is continuing. It's difficult in this phase to forecast exactly how many machines will be required because also in the high-end computing applications, it's a matter of qualifying processes and also expected market growth. But compared to 3 months ago, sorry, it's 2 months ago ended February, things have further improved. Also the yield on the installed base, the largest base in Taiwan, which is over 20 machines now, we reached yields of well above 99%. So that means that, that installed base is very capable to serve the mainstream. So that's what's happening on the hybrid front. On the TCB, we proudly announced to have received the first order. As explained earlier, this system for TCB processes, again, stretches the envelope tighter specifications, far less consumption of certain chemicals, gases in this case, higher accuracies as mentioned and that should be the next benchmark for these type of processes. So then we can offer the complete solution offering hybrid bonding, TCB for those processes where that is preferred. And then in addition, we have this bridge attach machine, which we also have already supplied several, which is the basic tool required to build chiplets. I hope these comments help you.

Yes, yes, that's helpful. And sorry, also on the ASMPT, they announced about the hybrid bonding tool and they had a first customer order. Could you talk a bit about how that compares to yours and your kind of technology lead there in hybrid bonding?

Well, first of all, it's very positive that compared to 2 years ago, when there was a significant, let's say, debate about how fast hybrid bonding would be adopted that now also we see competitors jumping on the bandwagon. The concept, as we know, is different. It is certainly not comparable in terms of mainstream high-volume capacity compared to ours as far as we know. It's still in a very early phase. One thing I should add, what I understood is they are marketing this system with 150 nanometers. We will ship in the next few days, the first 100-nanometer accuracy system, which is definitely required to expand the use of hybrid bonding into the next generation. So in the 3, even 2-nanometer chip design geometry. So we are definitely on track to further expand our market positions.

And then just quickly on the automotive end market. You talked in your release last quarter being weak, but then this quarter, you also flagged a large order from a Chinese subcontractor. I was just wondering if you could maybe give some more color on how you think this end market will develop as we go through this year?

Well, first of all, the automotive market has been very strong already for the past several quarters. There was -- yes, just by, I would say, coincidence less orders in the first quarter books, but then immediately after quarter end, we mentioned a major Chinese one because that's stuck out because of its size, but also all the other -- and you may have noticed that in their communication to the shareholders, big plans for expansion, whether it's Infineon or STM, but also in the U.S., same for China. There are major expansion plans underway for power devices, especially in the electronics for electrical and hybrid cars. So we are benefiting very much from that as opposed to still ongoing weakness in high-end computing. As we mentioned, less orders than expected for this year's round in high-end smartphones. So that's the picture.

The next question is coming from Charles Shi, calling from Needham & Company.

Richard, I want to start with the backlog. I think last quarter you disclosed that the backlog to be about EUR 270 million. And given that your book to bill is above -- was above Q1. It seems like a backlog grew a little bit. But if I compare the Q1 quarter and backlog with your Q2 guidance, it seems like your backlog coverage, which I calculate a backlog for the prior quarter, divide that with the revenue for the next quarter seems to be 1.7x, but historically, especially in downturn years, that ratio tends to be like 1.2x. So it seems like your backlog has still remained relatively high relative to historical. It's a little bit unusual in a downturn here. Do you kind of expect that ratio to go back down to somewhere around 1.2x or maybe closer to 1x, maybe by the end of the year? Or what's the expectation there in terms of the backlog?

Well, Charles, we are comparing, let's say, a different product portfolio in this downturn to the previous one. Don't forget that for the hybrid bonding orders, lead times are longer. And we mentioned also end of February that the orders we received are not for typical, let's say, 8 to 12 weeks turnaround in conventional assembly equipment, but they have 6, 9 months lead times. Also when we talk about systems to be integrated in automated lines, the lead times are even a bit longer than that, even 10 months. So they -- out of that 270 plus, there is a substantial amount for Q3 and some for Q4. And even a few for Q1, but that has to do with that new customers, simply program to adopt this hybrid bonding. So it's a different -- it's not apples-to-apples comparison. We also book, of course, orders. We had a good April month more to come in May, June. Needless to say, there is some uncertainty in the market in this current environment. So one should always be more on the conservative side in guidance than to expect that things should develop more positively. But one never knows. Still, it is expected that somewhere in the second half of this year, certain markets will need, again, capacity. Although many of the utilization rates are somewhere around 70%, some are bit higher, also some lower still. So we're not yet up in the 80s, which are typically turning cycles. That may well happen in the second half. That's what VLSI, TechInsights today. They all expect this type of pattern. But for us, this backlog is a very, very good one. Does that answer your...

Yes, excellent, Rich, always appreciate the color you provide to the investment community. Maybe a quick follow-up on the smartphone side of the business. I think you talked about high-end smartphone order seems to be a little bit lower than you previously expected. But I want to ask overall smartphone revenue, I know there's a high-end, there's a mainstream, but combining those 2, high-end seems to be doing a little bit better than last year, but I don't hear you talk about the mainstream side. Do you expect the total smartphone revenue? I think you called out mobile Internet in your annual report. Is that going to be flat or stay up or slightly down? What's the expectation for 2023?

Well, it can be either one. There are certain new applications under development for next years, but that already has to be established in a certain initial volume towards the second half of this year. On the Internet, you can find many articles about configurations expected for 2024. And that could mean, yes, the uptick in '23 for a strong year '24. For '23 itself, yes, well, it's flat or slightly more, but I can cover it may be a bit different. It looked like in the second half of last year, a more, let's say, higher growth expecting, let's say, model for '23 than it currently looks like. But we've had this in previous years as well. It's always hard to tell the marketing decisions, which are made at those brands, how much they will, yes, let's say, order in terms of the capacity available. But I'm just telling you the situation as it is right now.

We'll now go to Rolf Bulk calling from New Street Research.

Yes, I wanted to ask about the memory market. Specifically, SK Hynix this morning mentioned that it expects the high bandwidth memory market to grow around 50% this year with a lot of that growth driven by generative AI? And my question is, of course, hybrid bonding is a candidate for adoption in HVM in coming years. And I was hoping you could give us an update on how your discussions with the main memory manufacturers is progressing and also whether the interest in hybrid bonding is still so consistent despite the CapEx pullback we see in memory?

The 2 customers, which we mentioned where we expect the adoption in the next couple of weeks is exactly for this HVM-5. And also we are continuing successfully to stack memories on top of CPU. We now have certain stacks with 8 and also working on 12, which are then hybrid bonded onto one another and that is definitely picking up pace. So from the start in CPU, we are now entering into memory applications and then combined in the end with stacked on to CPUs.

You also mentioned for my follow-up question, that orders for hybrid bonding have decreased sequentially. How should we think about this? Does any of this relate to a slowdown in adoption of the technology in a particular end market? Or is this just regular lumpiness with some of your customers placing large orders in particular quarters that might not necessarily repeat in subsequent quarters?

That's a combination. One is, of course, yes, the nature is always lumpiness. But then, as I mentioned earlier, yields of our bonders have increased significantly. So the volume available as -- remember, a year ago, we were at yields of about 20% and we have accomplished in this 12 months that these machines are now running well above 99% yield. So the output is significantly higher. And supposedly, that covers today's demand. Further proliferation is underway. New devices are qualified like the MI-300, then we have a suite of products from Intel, which are being tooled up. So it's not that hybrid bonding adoption is slowing. It is more a matter of, yes, a combination of our machines output and maybe end markets, yes, let's say, the adoption of, yes, new devices that sometimes takes longer time. But the adoption is well on track. Any next question?

[Operator Instructions] Next question is coming from Mr. Robert Sanders from Deutsche Bank.

My first question would just be about optics, in particular, the periscope zoom lens. I was just wondering how big an opportunity that could be for you as it's rolled out at your largest customer? And can you give us an idea of this is a larger opportunity relative to prior camera innovations like dual cam and trio cam?

It certainly is a major opportunity. And as I tried to explain in an answer to an earlier question, the qualifications should be, let's say, reached in the second half of this year. And we will see in the number of machines which are connected to that first round. How big the rollout next year is expected to be, so a significant next round.

Just my follow-up would be on TCB. You sound like you've been -- you've received your first order from a U.S. company, which sounds like it might be Intel. I mean, as I understand it, the TCB market is going to remain large for quite a while. So does this mean that you have an opportunity to recover share here against ASMPT? And how long do you think the TCB market will be larger than hybrid bonding?

That's an excellent question. Number one, I think we answered some calls ago, we have developed this tool simply because that customer has requested us to help to do that. So their current installed base is not able to follow the roadmaps and specifically down to 1 micron accuracy. And well, that is, of course, tied to a certain expected market volume. Otherwise, we would not develop that machine for that customer specifically. There are also other customers interested in this tool. It looks very positive. As we shared in the Capital Markets Day and we will share it in the next Capital Markets Day and again in more detail, for definitely the next 5 years, you will have a mix of TCB type of connect, whether or not fluxes and then you will have the hybrid bonding. And the combination is a very strong, let's say, support to those customers who are especially developing chiplet architecture devices. So with this first benchmark, the first machine qualified also ordered and accepted that offers us a very strong next product line.

Next question is coming from Mr. Marc Hesselink from ING.

Yes. First question also on hybrid bonding. What are your discussions currently with your largest client? I think you just said it should be capable for mainstream with the current yields and the current shrinkage. So is it likely that, that will happen soon going into mainstream? I think I would have to read that going into the smartphone devices.

Yes. That's again, an excellent question. We don't know for certain, but we are read of what is happening. They want to move in an accelerated fashion to the 100-nanometer spec. And this is likely to be linked to a next processor for a high-end smartphone application, which then should hit the market in the model 2024. So qualifying this year, multiple systems installed in the first half of next year in time for volume application in the models launched in Q3 2024. Maybe it is first for the high-end tablets and the year thereafter smartphone, no one really knows that. It also depends on other factors than the factors we can influence. But that's our current read. Clear?

Yes, sure. That's clear. And then the other question is actually more on the second half for this year. And if I read correctly, I think as the -- probably a bit more hybrid bonding and automotive continues to be strong and your backlog is still very well filled. Is it fair to assume that the second half of the year will be a bit stronger versus the first half of the year than your normal seasonality?

Well, the best comparison is 2019. That is the -- the bottom in '19 was at exactly April, May, June '19, and we saw orders pick up in Q4. After let's say, the industry tide turned early 2018, a similar pattern. The only thing which is different is that the peak in '21 was significantly higher than the peak in '17. How much of that is AI related or other applications related is difficult to tell. What's also different is the global economy, how much that will impact semiconductor demand, who knows. The China situation, on the other hand, we recognize now rapidly building capacities outside China. Before year-end we'll have certain new programs also for high-end smartphones in India. We are preparing organizations to support that locally, same with Vietnam. So that is gaining traction. That could lead to orders in the second half of this year. But still always sort of rule of thumb, the first half is stronger than the second half apart from 2019. And as I mentioned earlier, we track very closely the utilization rates at customers. It's varies per region and China is a bit lower than Taiwan and also Southeast Asia. Southeast Asia is also typically more automotive related. But it's somewhere around 60, 70, some areas slightly above 70. So we're working through this correction phase. Whether the tide really will turn in the second half, it's anyone's best guess. But we have some unique new technologies, which could accelerate as well. So very interesting and very positive.

Clear. And then the third question, follow-up question is on your investment levels. Obviously, you always adjusted for the output levels, but you're also now strictly investing a bit more to prepare for [indiscernible] hybrid bonding. What should we expect for the coming quarters?

You mean CapEx or...

Sorry. Sorry, the -- sorry, more of the OpEx side, so within the combination of R&D and the more structural SG&A as we're not -- the near-term with more of the things that you're building up because of hybrid bonding?

Well, the current levels support our plans as they are. There is no significant -- there is no increase of any materiality. The organization is in place. The support is organized. So I would guess on the OpEx side that we would stay at similar levels. And if you take out one-time effect always in Q1 share-based payments, yes, those are difficult levels which are sustainable.

We'll move on now to Mr. Didier Scemama calling from Bank of America.

Thanks for the update and the thoughts, Richard, high-end smartphone adoption for hybrid bonding in 2024, I think that was really my -- one of my questions. I wanted to ask you also about 2024 on hybrid bonding. Which one do you think is going to be bigger for you? Do you think it could be high-end smartphones? Or do you think that the generative AI, high performance compute sort of thematic could be a bigger driver for hybrid bonding adoption? And specifically, you mentioned high bandwidth memory and CPU, which is a no-brainer for hybrid bonding. I'm wondering whether you think the GPU market and training algorithm could also benefit from hybrid bonding perhaps next year?

Yes. GPU market, certainly. But if you ask us today, as I also answered to an earlier question, we would guess that the high-end tablets is the first let's say application for a process using the hybrid bonding and probably chiplet architecture to some extent as opposed to high-end smartphone, but we could be wrong. But it's definitely the case that they are developing with a lot of effort to be able to use the 3 nanometer, maybe even 2 geometry in that next generation. So a lot of effort is being invested. That's from a mainstream application and volume. I would guess more '25, more likely. But that is for all let's say major volumes hybrid bonding. '24 for high-end definitely. And the first, let's say, product families from Intel processors and big volumes to be expected '25, '26. The infrastructure is being organized. These will all be established on integrated lines. But then putting me on the spot for '24 is still too difficult to answer with a definite answer. But we will update everyone again on the Capital Markets Day, June 6, in some more details. Yes, that's as much as I can share right now.

All right. Well, that's great color. Just also being a bit more down to earth on the second half. Is there any particular reason why, let's say, historical seasonality would not play out in Q3 and Q4 above and beyond the current macro uncertainty?

Well, what I mentioned earlier, we did not expect in '19 the tide to turn positively in Q4. That was, to us, all of a sudden, yes, the tide turned. And then we had sort of a hiccup in March due to COVID, but then immediately it picked up further in April. But the rules of thumb 6 to 8 quarters up and then 6 to 8 quarters down. Yes, by Q4, we will have reached definitely 6 to 8 quarters. We have been -- although nobody believed that we have been careful since Q2 last year. We saw definitely signals of, yes, let's say, peaks past. It's also hard for us to understand what is the impact of GDP in this case? Will next year, interest rates be low, I don't know.

Next, we go to Ruben Devos calling from Kepler Cheuvreux.

Yes. I've got 2 questions, both on the hybrid bonding. The first one is something you mentioned in the annual report, you sort of provide a rundown of the potential hybrid bonding applications, which is impressively wide range. In this call, we've talked mostly about computing AI and smartphones. But in the annual report, you also mentioned autonomous driving, medical, gaming. So I was curious how far out is development for this type of applications? And to what degree is maybe Besi working towards potentially addressing these other markets at some point? That's my first question.

Well, the gaming certainly is already part of the, let's say, AMD first product range. It's not -- I mean we're talking as you rightly say about computing and high-end smartphones, which is expected. Certainly the high-end smartphones when that is adopted that is really a game changer in terms of volumes. And you need many machines for that. Computing a bit less, the high-end computing. So from a numbers of machines, the biggest impact will be in positive sense, the adoption for mainstream high-end smartphones. That's the expectation as it stands. Automotive, yes, it's more autonomous driving, of course, you need more intelligence. But that's currently not seen as the imminent driver. Same for medical. But gaming, yes. I should add to that, the chiplet architecture element. So it's not just connecting single devices, but the architecture of multiple devices to create an electronic circuit. For many end products is what you see in conferences also when you simply look online, what are the objectives to be reached of designing chiplet architecture using hybrid bonding and also TCB for that matter.

All right. And then maybe a second question. I believe you provided an update in the prior release that you shipped about 35 units to date, of which, 4 bonders incorporated in integrated production lines. I was wondering whether you could offer an update as to shipments, but also which type of systems you currently have in the backlog? Yes, in the previous answer, you talked about a clear direction into integrated lines. But how should we think about the incremental traction for let's say integrated bonders relative to standalone tools?

Well, there are several questions. We certainly have shipped a few more bonders Q1, but as we indicated also 2 months ago, the quantity and orders we received will be shipped Q3, Q4. So not so much in Q1 and Q2. The other question, and I mentioned that several times, this whole concept of integrated lines is gaining more and more traction because of the simple facts yields impact by particles. So an integrated line offers less particles. And the qualification of the first line in Taiwan is as we speak, same in the U.S., and those should become the standards you could say, workforce for hybrid bonding in this industry. But again, in qualification phase as it is right now. So over time, we will ship more bonders to be integrated into integrated lines than standalone bonders.

We'll now go to Nigel van Putten calling from Morgan Stanley.

Just some clarifications really. First on the high-end smartphone optics or periscopic lens. I think there was some assumption that, that would be maybe due for shipment this year, maybe already in your backlog from an earlier question, I got the impression that the qualification still needs to be finalized, and you see this more of a next year development. Is that correct?

Partly, the orders for multiple systems should still happen this year or early next year. So those orders were not yet booked.

Understood. That's clear. Then another question. I think your temporary FTE count in Asia this quarter was 232. That compares to 60 in the fourth quarter on similar revenue levels. I know you manage that very carefully. So does that imply that there is a lot more work now going on, on converting the backlog into the next quarter? Is that flagging some optimism into the second half of the year perhaps? How should we read that?

The biggest is of course the ramp to go in the midpoint from 133 to 160. And certainly, some -- if you take the backlog also into Q3. And we have to, yes, let's say, carefully anticipate that next year we might go into a higher gears, and you have to train people. So the best you train them once you have volume. Anyway, so part of it is that. But as you said, we are very close tight ship, look at the margin. By the way, nobody asked about the margins. We've had many questions last time and supposedly that's the industry standard now 64% gross margin. Anyway, you only achieve that when you run a tight ship.

Understood. As a follow-up to ask a question about the gross margin. So going forward, how should we think of that? Is that now another step up slightly, 62, 64, which is not continue? Or was there any specifics in the mix? I think you already gave the answer, but still want to provide you the opportunity to provide more color.

Well, there are several things. First of all, don't forget we had a wonderful dollar in the second half of last year, nearly parity to the EUR o. And we've also been able to increase prices early for inflation and other issues. And then, of course, the ongoing development of new machines with successful applications, that all helps to bring that margin again at higher levels. Despite the fact that we have less effect of serial production, volumes are nearly 40% lower than peak, were about 40% lower than peak, so it's very inefficient. So you have several factors which help to further increase the margins, but may also with a different currency model have another impact. But needless to say that once you have reached these levels, and especially in downturns, that tells you that things are developing continuously in a better margin control.

Yes. Understood. Very strong, indeed. Maybe one quick last question. In the order book combined sort of the statements you made before, but also today about a large U.S. IDM and that lands into perhaps already in the second half of the year. If I can sort of read you correct that is not yet in the order book, but there's quite high confidence that those orders will still materialize in the second half of the year as flagged before. Is that correct?

Yes. That definitely we have mentioned that also in previous calls, you -- there is a phase, which is an initial adoption and qualification. And then you have sort of an initial production capacity, which is around 20 machines at a minimum, but now these machines also start to say that again, run at very acceptable yields. So that offers a mainstream adoption from an end market perspective. And once that really becomes, yes, let's say, yields definitely above 99.9%, then customers can start scaling. And then you need multiple systems. And then the minimum picture is typically 50. We're somewhere in between the low 20s and 50 in Taiwan. We are still below 10 in the U.S., but with definitely a big potential this year that, that could move up to what you just suggested. So the adoption is continuing. End markets are a bit -- you could say, look at it, be an Intel less, yes, at this moment, but there are definitely signs that should turn at a minimum next year.

We still have a couple of more questions. And next one will be coming from Johannes Ries calling from Apus Capital GmbH.

Yes. Some follow-on questions to what we discussed before. First, generative AI, how much you have realized since last months, because my feeling is it really spread around everywhere. And it's -- maybe it's just at the beginning, how much this maybe changed the plans maybe of your customers, especially by the adoption you mentioned of chiplet technology and so forth, could help you going forward with all we have to offer bridge attach and also bonding technologies?

Well, as I tried to explain earlier, more and more the industry is moving in this direction. There are many indications, broader adoption. Remember, 2 years ago we only had 1 customer in Taiwan driving the bus and then in the U.S., but slower, and now also Korea, also memory in the U.S. And then more and more clear examples of designs used in those end market applications. So yes, there is an ongoing adoption. And with our portfolio now both hybrid and TCB and bridge attach, don't forget our flip chip. We are able to cover, yes, most of that spectrum, and certainly at a very high-end.

And this AI boom is also maybe further -- maybe driving the adoption of just chiplet technology?

Yes, that's what everyone tells us.

Maybe on the 100-nanometer hybrid bonding, the next generation. If I got it right, so that maybe also maybe the generation, which is used for this high-end smartphones solution maybe start next year. Remind us a surprising difference between just generation first and generation second, so compared to the 150 nanometers?

Well, this first round from 150-nanometer to 100-nanometer moves depending upon which loading automation is used and whether it's connected into a line. We call it, Generation 1 plus and it's not yet Generation 2, because it's based on the same system platform. We moved from between 1.5 million to 2.5 million from 2 million to 3 million. That's where it currently stands. But this is early, simply confirming the capability of 100 nanometers. You would need more automation around such a system so you have the cleaning, wet cleaning, dry cleaning in the automated line. Yes, the price of the bonder is again, it touch higher because of the interface.

Great. And so that's not the end because in the longer-term we even expect -- we discussed it in last call, move to 50 nanometers, but that's sometime out into the future, but...

Yes, that's 3 years away. But as I mentioned, we are in a prototype phase developing such a concept. It's moving along because there's definitely customer pressure, but that will not be ready before 2026. Maybe in prototype '25, but that's what the market also expects.

Coming back to this memory topic. How big could be this memory market opportunity for hybrid bonding compared to the logic part?

Well, it could be much larger than the CPU, because you have more variety and how many memories 2Ds, 3Ds you use. But the biggest, let's say, volume upside is this chiplet architecture. So if you combine more devices, every combination needs a bonding step. So we explained that also in the Capital Markets Day, that is the real let's say game changer in hybrid bonding volume. Not so much the difference between CPU and memory.

Bringing all this together on one device that's a big driver, it's clear. Totally, as a topic, a big boom in compound semiconductors silicon carpet, gallium nitride -- silicon carbide gallium nitride. Is this anything also you're benefiting from?

Yes, because they also need to be connected. It need to be packaged. So that's also definitely helping, especially in the automotive applications, power modules. And big [indiscernible] plants in Germany, Bosch, in particular. So yes, that's definitely a growth segment.

Great. Then on this geopolitical investments which are made now in U.S. and in Europe, now say, definitely now has just started, maybe the front-end, things are coming. But in the end, also maybe it's after some time also they have to invest in back-end solutions. When you expect from these investments also maybe a further push in orders? Is it maybe the end of next year? Or is it 2025 topic?

'25 topic. Before you have a front-end fab and it takes you 2 years. And don't forget the Chip Act is still -- in the political environment that also takes time. So we don't count with that before '25.

But nevertheless, [ it's also far away ] another growth driver. And finally, on China, we hear also from the front-end guys that in the -- not in the high-end segment, but in the middle-end or even basic mature nodes, there is a huge investment going on in China. You mentioned U.S. China business had quite been weak. But could you also see an uptick going forward if these front-end things also need to -- need for back-end?

No, we mentioned already that in automotive we had a big order in the first week of April, which for China was a typical major order. And that is for people also expect could support a next investment round in the second half of this year. But anyway, many more questions. We are past the hour.

So I come out of those things.

And call me -- in case you have further questions, we can set up a call, and look very much forward to have a longer discussion.

Yes. Mr. Blickman, the last question has been withdrawn and we do not have any further questions at this time, sir.

Thank you very much. Thank you all for calling. And please, in case you need more details, don't hesitate to contact us. Thank you. Bye-bye.

Thank you, gentlemen. Ladies and gentlemen, that will conclude today's conference. Thank you very much for your attendance. You may now disconnect. Have a good day and goodbye.