BE Semiconductor Industries N.V. (OTCPK:BESVF) Q2 2024 Results Conference Call July 25, 2024 10:00 AM ET

Company Participants

Richard Blickman – Chief Executive Officer
Leon Verweijen – Senior Vice President, Finance

Conference Call Participants

Madeleine Jenkins – UBS
Didier Scemama – Bank of America
Robert Sanders – Deutsche Bank
Simon Coles – Barclays
Andrew Gardiner – Citi
Martin Marandon – ODDO BHF
Martin Jungfleisch – BNP Paribas
Johannes Ries – Apus Capital
Charles Shi – Needham & Company
Ruben Devos – Kepler Cheuvreux

Operator

Good morning, good afternoon, ladies and gentlemen, and welcome to BE Semiconductor Industries Quarterly Conference Call and Audio Webcast to Discuss the Company’s 2024 Second Quarter Results. You can register for the conference call or login to the audio webcast via Besi’s website, www.besi.com.

Joining us today are Mr. Richard Blickman, CEO; and Mr. Leon Verweijen, Senior Vice President, Finance. Currently, all participants are in a listen-only mode. Later, we will conduct the question-and-answer session and instructions will follow at that time. 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 like to remind everyone that on today’s call, management will be making forward-looking statements. All statements, other than statements of historical facts might be forward-looking statements. Forward-looking statements reflect Besi’s current views and assumptions regarding future events, many of which are by nature, in heavy claims, uncertain and beyond basis control.

Actual results might 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 report filed with the AFM.

Such forward-looking statements, including guidance provided during today’s call, speak only as of this date and Besi doesn’t 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 like now to turn the call over to Mr. Richard Blickman. Thank you.

Richard Blickman

Thank you. For today’s call, we’d like to review the key highlights of our second quarter and six months ended June 30, 2024. And update you on the market, our strategy and outlook.

First, some overall thoughts on the second quarter and first half year ’24. Besi reported second quarter revenue, gross margin and operating profit at the high end of guidance with significant order growth realized for hybrid bonding and other AI-related applications. For the quarter, revenue of €151.2 million and net income of €41.9 million increased by 3.3% and 23.2%, respectively, versus the first quarter of 2024.

The revenue increase was due primarily to higher shipments for photonics and 2.5D assembly applications. Order trends this first half year highlighted increased demand for Besi systems used in AI and other advanced packaging applications. Bookings of €185.2 million in the second quarter, of this year and €313 million in the first half 2024, represented increases of 64.5% and 22.9%, respectively, versus prior year periods.

Moreover, we estimate that approximately 50% of our orders over the past 12 months were AI-related. Sequential and year-over-year comparisons highlight contrast and growth trends for AI and mainstream assembly equipment markets in recent quarters. Since second half 2023, we have seen significant order growth for high-performance computing applications, including 2.5D, 3D and photonics assembly solutions, in support of a broad-based expansion of generative AR demand.

Such growth has been partially offset by a slower recovery as expected in mainstream assembly markets and in China, particularly for high-end smartphones, automotive and industrial applications. In general, post-pandemic inventory levels at semiconductor producers still remain elevated despite gradually increasing utilization rates.

Basic continues to navigate an extended assembly downturn at high levels of profitability as a result of increased 2.5D and 3D order momentum with gross and net margins realized of 65% and 27.7%, respectively, in the second quarter of ’24. For the first half year 2024 gross margins improved to 66.1% versus 65% in the first half of 2023. The reduction in our net margin to 25.5% this first half year, primarily reflected a 24% increase in development spending and increased share-based compensation versus the first half of 2023.

Our financial position is healthy with net cash of €74.4 million at quarter end, equal to 13% of the latest 12-month revenue post the capital allocation of €186.3 million in the form of dividends and share repurchases during the quarter and the conversion of €89.9 million of convertible notes into ordinary shares. On July 17, 2024, we successfully completed an offering of €350 million of 4.5% senior notes due 2031 to further solidify our capital base and helped fund anticipated growth over the next decade at attractive terms.

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 slowly due to pockets of excess capacity still remaining in certain consumer markets including automotive, industrial and smartphones.

As such, the mainstream assembly downturn, including the Chinese market has extended to 8 to 10 quarters. As a result, tech insights has reduced its assembly market growth estimates for 2024 to 9% from 16% in March and 31% at the start of this year. It also increased its growth estimates for 2025 and ’26 and expect a next market peak in 2016 at €7.4 billion up 80% from 2023 levels.

However, it also forecasts that Besi’s addressable market and, in particular, the advanced placement market will grow at rates significantly above the overall assembly equipment market over the next five years due to increasing AI adoption in all end markets. We are encouraged about Besi’s programs given expanded hybrid bonding adoption for both logic and HBM applications. traction gained in the marketplace by our next-generation TCB system and continued demand growth for our flip chip and multi-module die attach for 2.5D applications.

During the quarter, we received orders for 29 hybrid landing systems from two customers for estimated delivery in Q4 ’24 and Q1 2025, further highlighting increased market adoption of this new process technology. Substantially, all such orders were for our latest-generation 100-nanometer QFC systems to be used in 3D logic applications. We anticipate additional orders in the second half of this year as customers ramp capacity for high-volume manufacturing in 2025. In addition, we received an important second quarter this quarter for our TC next system.

To support our product strategy base continues to increase R&D spending for the next-generation hybrid on development, targeting 100-nanometer placement accuracy the ongoing build-out of hybrid bonding and TCB capabilities in anticipation of expanded logic and memory adoption and enhancements to our current product portfolio for the next up cycle. We anticipate additional share gains in the next market actor as node sizes shrink further and placement accuracy increases. All of these trends play to the strength of Besi’s core competencies. Now a few words about our guidance.

For the third quarter ’24, we forecast that revenue will be flat plus or minus 5% versus the second quarter of this year, with gross margins ranging between 64% and 66% based on our projected current product mix. The projected third quarter ’24 revenue increase reflects a typical seasonal growth due primarily to a weaker smartphone capacity built this year. Q3 ’24 aggregate operating expenses are forecasted to decrease by 0% to 5% versus the second quarter of this year. This concludes my prepared remarks.

Before we begin Q&A, I would like to remind everyone to limit your questions to two at a time to all participants have an opportunity to ask questions. Operator.

Question-and-Answer Session

Operator

[Operator Instructions] the first question comes from the line of Madeleine Jenkins calling from UBS. Please go ahead.

Madeleine Jenkins

I was just wondering on the traction of your kind of TCB tools was the order this quarter from a memory or logic customer? And also, are you still seeing good traction for kind of the Korean memory players for HBM?

Richard Blickman

Well, two separate questions. First of all, this order is for the development in logic. And number two, in Korea, we see and also outside Korea, continuous development side by side of TC and also hybrid donning for stacked HBM devices.

Madeleine Jenkins

Okay. And we’ve also had some supply chain feedback that TSMC might change the dielectric material used in bonding for the SIC from silicon dioxide to silicon carbon nitrate, I think. I was just wondering if this is something that you guys had also heard of. And if that’s the case, do you need a new configuration of hybrid bonding tool? Or is the current installed base that’s still okay?

Richard Blickman

Well, that’s a very general question. There’s a constant development of new, let’s say, hopefully, better materials underway also in our systems, they are different processes. So far, the mainstream, and that’s evidenced by ongoing orders is using our current technology. So, there is, to our knowledge, no imminent change.

Operator

The next question comes from the line of Didier Scemama calling from Bank of America. Please go ahead.

Didier Scemama

Good to hear from you, Richard. A couple of questions, as always, quite basic for me. If you could give us an update on how many hybrid bonding systems you shipped actually in Q2? And also on the fourth quarter, assuming you don’t want to give us a guidance, but I think you mentioned that half of the Q2 order intake from hybrid bonding will be delivered or broadly speaking in Q4, the other half in Q1. Should we expect a material increase sequentially in Q4 on the back of that? Or is it still early to say, given the uncertainty on the flip chip business?

Richard Blickman

Well, also, thank you very much. Good questions. We do not disclose the exact number we site. But if you look at the revenue guidance and what we accomplished, then we shipped what we were scheduled to ship in Q2. with the orders received in early May or end of April, it was already clear that, that would come our way. there’s a typical lead time of about nine months to build these systems, especially testing is highly critical.

So, as you pointed out, whether it’s exactly 50% or around that number will be shipped in Q4 and Q1. As we also mentioned, we expect further orders to be received in the second half of this year. And they are also because of the lead time typically then for Q1 shipments, maybe some ready for Q2. But in general, the numbers are increasing. So that is, as we said, a clear evidence of further adoption.

Didier Scemama

All right. I appreciate that. It’s a hard question to answer, but the stock, what is the consensus in the market is for about 80 hybrid bonding system shipments in calendar year ’25, and we can see why those numbers are justifiable perhaps given the backlog that you have. I estimate you’ve got 57 units in your backlog at the end of Q2. So, it’d be interesting to see if you validate that number. But are you prepared to endorse that 80-unit shipments next year, even if that includes the TCB and flavors of advanced packaging, because that’s clearly what’s underpinning the stock price. And I think it would be good if we could have your opinion as to whether HBM developments are warranting those sort of volumes increasing in ’25 whether you see a pull-in of smartphone adoption or, in fact, on the logic side, you’ve got obviously multiple server CPU customers that are worked on — sorry, on hybrid bonding, whether those orders would choose to justify you would be comfortable with AT unit next year?

Richard Blickman

Well, those are precise numbers. And as we clearly always indicate, there are many factors, which cause variation. And certainly, on year outlook term. But if you — and that is a comment confirmation to the adoption which is gradually increasing in logic ever more broadly, new families introduced and also on the road maps for ’25, ’26. So, for logic, there are very clear road maps on the adoption in HBM. We all know that, that is still in a very critical development phase. The outcome is hard to forecast.

But — and more specifically, how much will be TC bonded and how much will be hybrid bonded, where will the change will it be 16? Or will there be still 20, nobody has a clear decision about that. Sometimes at conferences, certain technologies are explained and models shared. So to come to clear numbers is default. But again, the adoption is definitely continuing. This quarter was very important this last quarter. with not only the major order for ’26, but also the additional two and then more in the forecast.

You also referred to the adoption in high-end smartphones. We all know that there’s a lot of development going on in creating the major advantages using a chiplet architecture also for the application in high-end smartphones we’re very much engaged in that. That’s also why R&D costs go up. When exactly that will be introduced in a next-generation device it’s hard to forecast. But the — yes, the major message is that the effort the industry is investing in demonstrating the superiority of hybrid bonding technology to other interconnect technologies is increasing gradually.

Didier Scemama

Okay. I completely understand my question. It’s a bit hard to answer, but maybe on something you’ve got a better grasp on is your confidence on H2 order intake for hybrid bonding. Should we expect an order intake for hybrid bonding that is consistent with the level we’ve seen in Q2? Or is it going to sort of come back to, let’s say, a more normalized level?

Richard Blickman

That’s so hard to fulcrum. It looks, otherwise, we wouldn’t make the statement. For Q3, we know already that certain orders are to be expected. How much the total will be is also hard to forecast. But again, the adoption at the ones who are the first movers with this technology is increasing gradually, continuously, so that would then lead to more orders anode. So, a positive development.

Operator

The next question comes from the line of Robert Sanders calling from Deutsche Bank. Please go ahead.

Robert Sanders

I guess my first one is just on the statement about 50% of the last 12-month orders being AI-related. I get that hired bonding is roughly half of that 24%. I’d love to just break down the remaining other half. I get that you’re in the OS debt and you’ve got business with ASC on cohort. It’d be great just to sort of — if you could just try and pass it out. Obviously, you’ve done some calculation to get to that 50%. So, I’d be interested to sort of understand a bit more.

Richard Blickman

Yes. Part is CoWoS and especially the OS part and also, to some extent, the COW, and then Photonics. So, an ongoing trend in repairing that technology introduced in those 2.5D modules. So, if you all add that up, then you’re at 50%.

Robert Sanders

It’s a great number. And then just on this SOIC DASH X, the high bonding version of SOIC at Apple. I mean, that customer seems to be introducing it for a very low-volume part, the server part. How quickly after that server part do you think they would be in a position to move it into a notebook type part, if that was to be the plan.

Richard Blickman

Well, it’s in the road map, as explained also publicly for ’25. So we should see that then, yes, with orders first half and second half ’25, maybe ’26. But there is a lot of development efforts, as I answered earlier, ongoing which points towards anticipated adoption either next year or the year thereafter.

Operator

Next question comes from Simon Coles from Barclays. Simon, please go ahead. Please go ahead.

Simon Coles

Richard. TSMC last week, talked about for N2, they think increasingly, customers will have to adopt chiplets. I was just wondering how you think about that for your hybrid bonding adoption? And if they’ve given you more color on that ramp as well?

Richard Blickman

Well, that is also to the earlier asked question. It simply confirms that this technology is on, you could say, the brink of being used in the interconnect of this next generation. So that makes the case more clear that, that could happen in ’25. And as I mentioned earlier, if not in ’26 for us to add one comment, it’s not always possible to understand for which interpretation a machine is used.

So we have, in the meantime, a quite significant number of hybrid bonding installed in Taiwan. And we know for certain devices in AMD for a long time, which was a start. But then there’s ongoing development going on for other customers with other applications. We are not informed in detail, which is understandable because if we would know, we would not be able to disclose that information. But what we share is what we hear at conferences. Those are typically public events. So, I can’t add many of those questions in details.

Operator

[Operator Instructions] The next question comes from the line of Andrew Gardiner calling from Citi. Please go ahead.

Andrew Gardiner

Richard. Another one on hybrid bonding, if I may. You took in 29, two orders in the quarter. That was right in the middle of the range that you told us about back in April of 25 to 35 and again from the two customers. But is in early May, you announced 26 of those orders with one customer. So clearly very skewed towards one customer with high-volume aspirations and another one, at least in the quarter with much lower volume, was that what you had anticipated? Was it sort of on HBM and one perhaps more development? Or have there been some orders that perhaps slipped into 3Q, and that’s giving you that better visibility in the second half of the year?

Richard Blickman

Well, the latter. So, it’s part. Again, it’s very hard to forecast when orders will be placed. So, you know programs, you know bigger plans. So, as we have said, since hybrid bonding started, the model used in Taiwan was 15 machines, the infrastructure in place to add that number of machines. Well, we’re getting closer to that. But the orders have always been in increments of five or eight.

And I think once also 12. It depends very much on fine-tuning in the meantime, the investment at the customer in Taiwan, but also, the other major volume customer has started, kicked off with ’26. They already had a certain number of machines, so that it’s a larger installed base once it’s all delivered.

There will be a next round which may be ordered towards the end of this year or the beginning of next year. That depends on the success of those first devices introduced to the end markets. Those are all factors which cause let’s say, not a fixed time frame. But for the rest of the business, that’s also always the case. These are capital goods. Customers have programs, whether they execute that in full at once is very rare. Usually, it is increments. So one can’t be too precise.

That’s why, by the way, production planning at Besi, we adjust our pipeline planning every Monday, Wednesday and Friday. And the reason we do that is because there is no certainty for any of the products until they are ordered. So, we make the best assessment. That’s also why there’s always a range, but that’s the nature of the business.

Operator

The next question comes from the line of Martin Marandon calling from ODDO BHF. Please go ahead.

Martin Marandon

Could you maybe make a quick update on the new hybrid bonding system coming with 50-nanometer accuracy, I believe, for the end 2025. And help us understand what kind of application could use this tool. And also since today, most of the hybrid bonding demand seems to be driven by the large generation of 2, 3, 100-nanometer accuracy. Do you think it would be the same when you release 50-nanometer accuracy tools and that if we kept most of hybrid bonding demand?

Richard Blickman

Well, that’s an excellent question. Number one, the 50-nano is full in development, and we should deliver a first prototype, you can call it, by the end of ’25. That has not changed. And the reason for that is clearly the design geometry moving into the 2-nanometer space. And for that, the expected omit for logic devices, requires then an accuracy in placement below 100-nanometer. So that has then been defined as 50.

And your next question is also a very good one. The 100-nanometer, supposedly the two major customers today driving the adoption of hybrid bonding, feel that the 100-nanometer is more fitted to this current generation and the next. So that is in 3-nanometer design geometry for logic devices is what is the most, yes, let’s say, highest yield let’s say, possibility. But that’s also not an exact answer. Maybe generation 1 would suffice to 150- to 200-nanometer.

But that also is part of the whole development in — it could very well be that for certain families the generation 1 is best cost of ownership for those applications because there’s a price difference. There’s 500,000 cost increase for the 100-nanometer. The 50 will be more. We don’t know yet the precise number, but they are yes, more precise cameras, motors, all kinds of developments which are not yet completed, of course, which lead to a higher cost of that platform.

Martin Marandon

Okay. Very clear. And just last question on the TCB side. When do you expect the industry to start mass adoption of TCB systems with buses below 25, 30 microns like the TC next fullness. And is there a difference there of timing for adoption between logic and memory.

Richard Blickman

Well, many of you know that RTC next, we call it TC next was developed for a major customer for exactly the application below 25, 20-micron bump pad pitch and also with a fluxless process. The current generation of devices can, let’s say, use 35-micron on bump pad pitch. So, our system is today not yet required, but the next generation is very clearly on the horizon. And we have seen in this quarter, a significant increase in activity to make this system ready for exactly those devices below 20 nano, that starts micro, that starts with logic devices.

And at the same time, the memory, we also shared that we shipped the first system from memory to a U.S. memory manufacturer, and they are already stacking 12, even 16, even higher numbers in the development lab. So, to technologies can be used for this HBM stacking, whether it’s a TC process and at the same time, side-by-side, it is being developed using the hybrid technology. Our system is running constantly demos multiple devices, different architectures, different sizes, different processes. So once technology moves to this below 20-micron bump pad pitch, it is ready to be used in mainstream applications.

Operator

Your next question comes from the line of Martin Jungfleisch calling from BNP Paribas. Please go ahead.

Martin Jungfleisch

Maybe I joined the call late, so apologies if this has been asked. But in the press release this morning, you mentioned that prospects remain encouraging also for higher bonding adoption HBM. So, does that mean that your systems are not qualified already and being used in upcoming HBM production? That’s the first question.

Richard Blickman

Well, as I answered to the question before, there’s constant development already for, I would say, now three years in developing a hybrid bonded stack of memory devices and side-by-side with using a reflow process and mass reflow process in case of [indiscernible]. And at some point, there will be a mix. So, if you follow closely the three big ones in memory, their road maps simply forecast that they will be using both technology for different device applications. So that’s also why the statement of ongoing let’s say, further development and increased development is pointing towards, at some point, the adoption of hybrid bonding in fact memories and HBM applications.

Martin Jungfleisch

And maybe if you go back to the mainstream markets. I think in orders, excluding the AI-related stuff has been flat. And you mentioned that the assembly market remains relatively softer smartphone auto in industry. If you look at data points on the smartphone segment, it’s been somewhat more positive lately. Would you say that this positive development comes to you with a delay, so you would expect orders on the smartphone side to improve in Q3 and Q4? Or is that not the case?

Richard Blickman

Well, everyone is looking for positive signs in the mainstream. But as we stated in the situation in smartphones, automotive and industrial, simply remained remain soft, that doesn’t mean that there are no orders, but there is no increase. There are new devices always develop them also in the engine 1, as we call it environment. But also, to be more specific on the high-end smartphones, there’s an ongoing development in new or new upgrades.

One of them is the processes, of course, but also the camera modules, also other devices inside those — and there’s a lot of publicity about announcements, both in the iOS and in the Android camp. And also, the expectation that HAI at some point, that was also an earlier question. would have a major impact on the tablet architecture in the processor, but also the memory side of those devices.

But clearly, in our statement, we don’t see hard orders for major volume ramps so far. When they are expected to come, it’s hard to forecast. But there’s a lot of development, which is always prior to market adoption. So, it will be very important to see what happens in Q3 and Q4. Will there be new modules adopted for a generation iron smartphone to be launched in Q3 next year, then we should see more visibility.

Operator

The next question comes from the line of Johannes Ries coming from Apus Capital. Please go ahead.

Johannes Ries

Yes. Richard, maybe two short questions. First, regarding the throughput of 100-nanometer machine. Is it the same like 150-nanometer machine? Or is there any reason that is lower? That’s the first question.

Richard Blickman

I can barely hear you. Can you speak up a little bit?

Johannes Ries

Can you hear me now better?

Richard Blickman

This is much better.

Johannes Ries

Okay. Sorry, I repeat the question. The throughput of 100-nanometer machine, is it the same like 150 or is it for any reason, lower end will stay lower than sales generation first?

Richard Blickman

Accuracy and speed are enemies was the question. The time accuracy, the slower the machine, but in the past three years, we have made a lot of progress in aligning this accuracy and speed in generation 1. So, the throughput is increasing from around 1,000 and even below that four years ago to well over 1,500 for generation 1, close to 2,000 depends on the size of the die and also the pattern on the die and for generation 1 plus, we are not that far off from generation 1.

But again, that also depends on which dies to handle you easy to imagine the better you can recognize to be placed the die, the faster you can place the die. But at the same time, you have these dynamic factors. So moving a die from one place to the next creates vibration in the [Volmer]. And the way to handle that is also a science.

But anyway, the step today between generation 1 and 1+ is not that large. What is very important to understand once we go down to 50, the dynamic forces are becoming a greater impact. So, then you can expect a somewhat slower throughput, which in the end with similar volume would result in more machine.

Johannes Ries

Second question, for photonic, you mentioned it in 1/2 answer. Can you give us a little bit maybe perspective you see in the next 12 to 18 months for your photonic business?

Richard Blickman

Well since early part of last year, the photonics application has gained significant traction. There are multiple customers now, and that’s especially for the high-end data center processors where this photonic technology is used in driving that circuitry and that is expected to continue. So, depending upon the AI adoption and volume installation, you can expect ongoing business. And we have definitely currently an excellent product range fitting those applications, which I also explained in earlier calls dates back to more than 10 years in the past. We’ve been engaged in this since early days. And that finally, I could say is coming to major mainstream applications.

Johannes Ries

And do you see any competitor who was close to you in this area?

Richard Blickman

Well, it’s all about also here accurate placement. And so many die [Volmer] companies, they are able to place something accurate and — but again, it’s the combination of accuracy and speed. So, the cost of ownership determines finally who is the preferred supplier, and we have an excellent combination. And you can see that currently in the progress of our business. But we are not alone.

Johannes Ries

Final question regarding your partnership with applied materials, a feeling comparably quite, you didn’t say anything about it at your Capital Markets Day, even nowhere today. Can you give us an update about your partnership and maybe the combined machine systems when they come in the market and so on.

Richard Blickman

Well, that’s definitely a very important question to put right. We’re working every single day very closely with materials in developing this technology. And their part is the preparation of the dies and the cleaning, the — also the preparation of the copper bump pads. And we do that jointly in the center of excellence in Singapore, also in our own laboratory. The key person has also joined Besi’s technology advisory board in February. So, I should take this message very well. that we use to explain more of what we are doing. So, the health of applied is very important in the adoption of hybrid bonding and it’s working perfectly.

Operator

The next question comes from the line of Charles Shi calling from Needham & Company. Please go ahead.

Charles Shi

Richard, maybe I want to ask a couple of questions about the TCB next. Maybe I missed it, but can you tell us what this new shipment? Where is it going to? Because I recall the first system probably, we to the U.S. logic customer second went to the U.S. memory customer. But the third one, where is it going to?

Richard Blickman

Well, there’s several developments. One has gone to Taiwan, another one to the U.S. and another one in Europe.

Charles Shi

All right. Then I want to ask a bit more about the overall TC next strategy for I’m sure you probably at least get some briefing from your team about the earnings call of your peers a couple of nights ago. Yes. And I think my takeaway is in HBM thermal compression bonding is a pretty close to market in Korea, meaning your Korean customers when they do HBM bonding, they tend to want to use the local vendor there. I mean up apparently, they get two tools qualified, but that they were not able to ramp up production involving with the customer. So, I wonder what’s your strategy there in that part of the HBM market. Again, you said you shipped Taiwan to one of the memory customer, which is not Korean, right? But the other two are pretty big players in HBM. Any thoughts on that?

Richard Blickman

Well, first of all, one side is the HBM. And the question will be in the future, how much will be TC driven and how much will be HBM driven. And again, the side-by-side these developments are currently with a lot of effort ongoing. And then we have the logic side. So there, we move below, let’s say, 20-micron bump pad pitch. And it’s clear that the data we are producing is simply confirming that for that segment, you have a major, yes, let’s say, open market for the next generations.

And that is not only U.S. but also Taiwan and then for many customers who use Taiwan. And there also, you have the picture of both technologies side by side. So, you have the TC in Korea to complete the picture, there’s one process which uses the mass refill. Still as a technology, and then you have the TC, but again, it’s not yet clear how much will be using either of those technologies. It is certainly the case that so far, Korea has been very much self-supporting.

But there is, of course, the question whether those suppliers can follow that same road map. Time will tell. We are heavily involved in the development. Of course, with hybrid bonding in the center of excellence in Singapore. Same for TC. But then currently, that’s done in Europe in field with many advantages over what is currently available. So again, Q2 was a very important quarter to understand that our TC next has major advantages over other solutions available in the market.

Operator

[Operator Instructions] The next question comes from the line of Didier Scemama again calling from Bank of America. Please go ahead.

Didier Scemama

Sorry, Richard, just a follow-up on — I think you mentioned earlier that Taiwan has reached, so it’s close to reaching the 50 unit installed base. I think that was a sort of similar target for the U.S. logic customer. Just if you could confirm that? And then related to that, maybe what’s your — what’s your best guess of what the installed base for HBM for Korean or U.S. customer might be at sort of peak capacity.

Richard Blickman

The first is easy to confirm. That stands and even our numbers beyond that 50. simply following the successful adoption. The second question, we’ve always said that in volume theoretical assessment, you could say 4x memory follows a logic. So that would mean a market significantly larger using hybrid filming for HBM than for single logic devices. However, with different architectures, you also — well, every die needs to be placed. So, it’s a rough calculation.

There’s something else which is important to realize. Logic has far more IOs contacts than memory and also the spacing is more for that reason, more critical and logic than it is in memory devices. So, the question then also, but this is also very, let’s say, old — in logic, the opportunities let’s say, historically are much more than on the HBM side. The memory side has always been a less attractive segment of the market.

But with this HBM with the stacking because that’s another complexity. Remember, 12 years ago, we delivered the first TC solutions, which are still running for a major U.S. memory manufacturer very successfully. With through silicon via stacking those dies even up to above 20. But to give you exact numbers, it’s difficult. I would answer that potentially, there’s a larger number of machines for HBM hybrid bonding stacking required then for logic.

Didier Scemama

That’s helpful. And are you still sort of your view, your current view our best guess is that the intersection of hybrid bonding HBM 6-layer stack in HBM 4? Or has that changed?

Richard Blickman

Well, that goes back and forth. Some are pulling it forward because of performance, because of heat, because of all the issues which supposedly and also with test confirmed are better using hybrid bonding, but at the cost. So, one of the key issues is, again, the speed, the throughput of the machine. And with slightly less accuracy, you can have these machines run faster.

So, the gap between using hybrid and TC is for that spec part less, but you don’t need full clean room for a TC process. So, the infrastructure cost is using TC lower than the hybrid bonding. But these are all information shared at conferences and there’s also an ongoing debate. And what is more important, the performance of the device or the cost and the only conclusion every single day is be very much active in that whole development.

Operator

And the last question comes from the line of Ruben Devos calling from Kepler Cheuvreux. Please go ahead.

Ruben Devos

Yes. Just a few questions, a follow-on basically on TC still. Just a clarification, I think, so far, you’ve mostly supply the TC next systems, right? So, I think at the Investor Day, you also introduced the TC fluxless system, but did I hear it correctly that TC next is already a fluxless process?

Richard Blickman

Well, you — let’s say, the world today is still using flux. The disadvantage of flux is that you have an additional cleaning step. And you also need more space. On the other hand, the fluxless process is not simply done. So you need a different process altogether. There are different technologies used. I won’t go into detail, but that’s all not settled.

So optimum would be to find a process which does not need flux. We have, yes, let’s say, developed with customers a certain process, which is operating. And the direction is certainly to see whether we can accomplish that in high volume. So, the direction and you can hear that in general in the market, is to, yes, let’s say, be successful in offering a fluxless process.

Ruben Devos

Okay. And that’s very much still in development. You have not yet delivered prototypes of the TC fluxless system?

Richard Blickman

Yes, we have the prototype of that. That’s also developing very positively. That’s also why we received an additional order and also ongoing development, especially in the U.S., both in logic and in memory application. So, it looks promising, but we’re in early stages.

Ruben Devos

Okay. And then just apart from, let’s say, the logical candidates in memory. Is it fair to say that also for logic, you would have multiple customers interested for the TC fluxless system?

Richard Blickman

Yes. Yes, certainly. In the U.S., Taiwan and Thailand from many other customers. So, the best way to look at it, if you can offer this fluxless in a reliable, high yield, let’s say, process qualification that is the ideal solution. You still have a full bag using flux. But again, the fluxless has disadvantages.

Every way you look at that, you have a continuous development increasing in what is possible using TC. Where do we have to move to hybrid and we are in the middle of this enormous development effort in the industry.

Operator

There are no further questions, so I will hand you back to your host to conclude today’s conference. Thank you.

Richard Blickman

Thank you very much, everyone, for attending this call and your questions. If you have any further questions, don’t hesitate to call us.

Thank you. Bye-bye.

Operator

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