Asmpt Limited Unsp/Adr

Hi, good morning. Ladies and gentlemen, this is Ben Po, the head of investor relations at ASMPT. And today I'll be moderating the call for the first time. On behalf of ASMPT Limited, welcome to our third quarter 2025 investor conference call. Thank you all for your interest and continued support. Please note that all participants will be in listen-only mode during the presentation by the management. We will start the Q&A session after the presentation. During the Q&A session, priority will be given to the covering analysts. Before we start, let me go through our disclaimer. Please note that there may be forward-looking statements about the company's business and finances during this call. Such forward-looking statements could involve known and unknown uncertainties and risks that could cause actual results. performance, and events to differ materially from those expressed or implied during this conference call. On the call, unless stated otherwise, all references to gross profit or margin, operating profit, segment profit, and net profit are on adjusted basis as described in our MD&A. For your reference, the investor relations presentation on our recent results is available on our website. On today's call, we have the Group Chief Executive Officer, Mr. Robin Ng, and the Group Chief Financial Officer, Ms. Katie Hsu. Robin will cover the Group's key highlights for the third quarter, guidance and outlook for the next quarter, while Katie will provide details on the financial performance for the third quarter. Now, I will hand it over to our Group Chief Executive Officer, Robin.

Thank you, Benjamin. Good morning and good evening to everyone today. It is a pleasure to have you all on our earnings conference call for the third quarter of 2025. Now, let's start with the key highlights of the third quarter. This quarter, we continue to experience strong momentum driven by AI. the group's advanced packaging and mainstream businesses continued to benefit from sustained AI adoption. The group's strong advanced packaging momentum has been driven by thermal compression bondings or TCP. We remain dominant in advanced logic have made rapid inroads into high-bandwidth memory, or HBM, and more recently, have a first-mover advantage in HBM4. At the same time, AI infrastructure, comprising data centers, data transmission, and power management, contributed to demand in mainstream businesses. In China, demand was also driven by EV and high factory utilization across all sets. Now, let me talk about our technology leadership in TCB. We have further solidified our leadership in HBM. The group's HBM-TCB solution have achieved better years versus the competition. And as I said above, we are leading in the transition to HBM4. In addition, a proprietary fluxless active oxide removal technology provides superior scalability for HPM69 and above with the lowest cost of transition. In logic, the group's ultra-fine pitch TCB for chip to wafer with Plasma AOR solution has successfully passed final qualifications for quality and reliability at the leading foundry and is ready for high volume manufacturing Notably Plasma based technology has been endorsed by this leading foundry underscoring its technological advantage over other processors Turning to TCB orders Encouragingly, the group achieved recurring orders from both memory and logic customers in the third quarter. In memory, our PCB solutions for HBM4-125 became the first to secure orders from multiple HBM players. We expect to remain as a primary supplier demonstrating our technology leadership in the rapid transition to HBM4. In logic, the group continued to win orders as a process of record for cheap to substrate applications of key customers. As the market transitioned to a larger compound size, we are well positioned to secure sizable orders in Q4 2025 and beyond from the OSEC partners of the leading foundry. As a business, we remain confident in the outlook for TCV demand. As to the other updates in hybrid bonding, the group continues to ship hybrid bonding tools in Q3 2025. Our second-generation hybrid bonding solutions are competitive in alignment precision, bonding accuracy, footprint efficiency, and units per hour. In photonics, we continue to dominate the optical transceiver market, reinforcing our leadership as a key supplier of 800G transceivers while also actively engaging industry players on next-generation 1.6T photonics solutions Moving to SMT bookings were better than expected in the third quarter demonstrating signs of recovery in the business SMT's AP solutions achieved strong bookings year-on-year growth in the third quarter and won sizable system-implemented orders from IDMs and OSACs for RF modules for base station to support AI growth. SMT also continued to win orders for the next generation chip, SMT2, in advanced logic smartphone applications from the leading foundry and OSAC partners. In our mainstream SMT business, the demand came mainly from EVs, where we remain the leading player in China. Before I conclude this section, I want to highlight that we have delivered a profitable quarter, excluding the strategic restructuring costs from the voluntary liquidation of the Shenzhen ABC plant as announced in August. The decision was made to optimize the Group's global supply chain to better align it with the evolving market dynamics and customer needs. As said in the announcement, this move is expected to improve the cost competitiveness, agility and resilience of the Group's global manufacturing operation for its key products and solutions. With those highlights, let me now pass over the time to Katie, who will talk about our group and segment performance.

Thank you, Robin. Good morning and good evening, everyone. Let me take you through the group financials. This slide covers the group's key financial metrics for the third quarter of 2025. The group delivered revenue of $468.0 million, representing an increase of 7.6% quarter-on-quarter and 9.5% year-on-year, largely driven by growth in SMT. In the third quarter, the group recorded bookings of $462.5 million, driven by AI momentum. We recorded recurring TCV orders in memory and logic, and SMT bookings were also better than expected. This marks the sixth consecutive quarter that we have achieved year-on-year growth. The group had an isolated bookings cancellation in the third quarter for its panel deposition tools from a leading high-density substrate manufacturer in response to a slower-than-expected digestion of its existing capacity. This is a one-off occurrence. And excluding this cancellation, the group's bookings in the third quarter would have been $486.6 million, 1.5% higher quarter-on-quarter and 20.1% higher year-on-year. The group achieved a book-to-bill ratio of 1.04 for the quarter, maintaining a ratio above 1 since Q1 2025. SMT posted a robust ratio of 1.12, while CEMI's ratio was at 0.96. The group closed the quarter with a backlog of $867.7 million. The group's adjusted growth margin for the third quarter was 37.7%, which is lower than our typical level. It was impacted by a larger contribution from SMT and the lower semi-gross margin, which I will explain in the next slide. I would like to note that the group's year-to-date adjusted gross margin remained healthy at approximately 40%. The group's operating expenses were up 6.2% Q on Q and 5.3% year on year. As expected, high OPEX was largely due to strategic R&D and infrastructure investments and foreign exchange impact. They were partially offset by prudent spending control and some benefits from restructuring. The group's adjusted operating profit was HK$124.4 million, down 26.6% quarter-on-quarter and 30.3% year-on-year, due to lower gross margin and higher operating expenses. Group adjusted net profit was 101.9 million Hong Kong dollars, down 24.4% quote-unquote, but up 245.2% year-on-year. The quote-unquote adjusted net profit, which included fee collected from the order cancellation mentioned above, was offset by the absence of tax credits recorded in the previous quarter. The year-on-year increase in adjusted net profit was driven by the fee collected from the order cancellation and the lesser negative impact from foreign exchange. The adjusted earnings per share was HK$0.24. Now moving on to the semiconductor solution segment for the third quarter of 2025. Semi's revenue was US$240.5 million, down 6.5% quarter-on-quarter but up 5.0% year-on-year. The year-on-year revenue increase was driven by stronger demand for Y-bonders and Y-bonders due to the increased needs for power management across multiple applications. Quarter-on-quarter revenue decline was due to the timing of key customers' AI technology roadmaps, which impacted AP demand this quarter. There was also some shipment disruption caused by a typhoon in September in China. Sammy's bookings of $207.8 million were down by 1.7% quarter-on-quarter and a 12.4% year-on-year. Excluding the booking cancellation explained above, Sammy's Q3 2025 bookings would have been $231.9 million, 9.6% higher quarter-on-quarter and a slightly lower year-on-year. Sammy recorded quarter-on-quarter year-on-year growth in Y-bonders and Y-bonders. TCB orders were up quarter on quarter but remained at a lower level due to the impact of IEP demand as mentioned above. As I said earlier, CEMI's adjusted gross margin was lower than normal at 41.3% for Q3 2025. Q on Q decline was due to a higher contribution from Y bonders, lower TCB revenue, and a relatively lower manufacturing utilization in Q3 2025. Year-on-year decline was due to high base effect from TCB manufacturing ramp in Q3 2024 and a high contribution from Y-bonders this quarter. Encouragingly, year-to-date semi-adjusted gross margin has stayed in the mid-40s, and AP margins have remained stable. The semi-adjusted segment profit was 82.6 million Hong Kong dollars in Q3 2025, down 52.8% quarter-on-quarter and 41.5% year-on-year, mainly due to lower gross margin and higher operating expenses as mentioned in the previous slide. Next, the SMT solution segment of our business. SMT delivered strong revenue of about 227.5 million U.S. dollars, up 28% quarter-on-quarter and 14.6% year-on-year. This was due to a robust performance in Asian markets driven by AI servers, EVs in China, and the delivery of a smartphone bulk order booked in the previous quarter. However, contributions from automotive outside China and industrial remained soft. SMT registered Q3 2025 bookings of $254.7 million, down 5% quarter-on-quarter, but up 51.8% year-on-year. Marginally lower quarter-on-quarter bookings were due to a high base effect from the Q2 smartphone bulk order, while the year-on-year increase was driven by strong momentum across both AP and China mainstream markets. AP bookings were supported by demand from IDMs and OSETs for telecom base stations and AI servers. China's mainstream business recorded strong year-on-year growth due to demand from EVs. SMT delivered a gross margin of 33.9% this quarter, up 136 basis points quarter-on-quarter and 163 basis points year-on-year. And the segment profit was HK$163.0 million. Up 205% quarter-on-quarter and 65.6% year-on-year. Both were driven by higher volume effects. With that, let me now pass the time back to Robin for Q4 revenue guidance.

Thank you, Katie. Now to Q4 revenue guidance. The group expects Q4 2025 revenue to be between U.S. $470 million and U.S. $530 million. This is up by 6.8% quarter-on-quarter and 14.3% year-on-year at the midpoint, which is above market consensus. This growth will be supported by momentum in both SEMI and SMT. Looking ahead, the group's GCB temp has the potential to go beyond US$1 billion in 2027, supported by recent news about investments in the AI ecosystem. AI data centers will continue to drive demand for AP, particularly TCP for HBM4 and Advanced Logic, where the group has technology leadership. The group's mainstream business will be supported by global investment in AI infrastructure, and stable demand from China, while visibility for automotive and industrial end markets recovery remains low. While the group has not experienced any material impact from tariff policies, it acknowledges that uncertainties remain. The group's global presence will provide flexibility to navigate any potential impact, and it will continue to monitor the situation closely and adapt as needed. This concludes our third quarter 2025 presentation. Thank you. And we're now ready for Q&A. Let me pass the time back to Ben to facilitate.

Thank you, Robin. Ladies and gentlemen, we will now begin the Q&A session. To ask a question, please raise hand on Zoom and I'll request you to unmute. Please limit yourself to two questions each time. With that, may I request Gokul to unmute?

Hi. Good morning, Robin, Katie, and team. Thanks for the opportunity to ask questions. First question I had is on the HPM4 commentary from you, Robin. And you mentioned that you are leading this transition to HPM4. Could you explain a little bit more what exactly that is indicating? Do you think that you would have higher market share in HPM4-based CCB compared to the incumbent Korean vendor? And also, your updated view on when does the fluxless CCB insertion happen for HPM? Is it happening for HPM4 or are we waiting for HPM4E for this migration to happen? Thanks, Gokul.

Gokul, have you finished? Yes, that's my first question. Thank you. Yeah, I think the first question is on the HBM4, right? The leading trustee shares in HBM4. That's right. Yeah. I don't know, as you mentioned, you know, MD&A, you know, we believe we have established ourselves as a primary supplier for the HBM4 market. I mean, we have a conviction because we are probably the first to have won the HPM4 orders, but not just one, but two major HPM players. Now, I think the second question is on fluxless. We believe that at some point, as the industry continues to stack higher and higher and moves from HPM4 to 4A to 5, In our opinion, it is quite inevitable that they have to move to a fluxless solution because the number of IOs will continue to increase, the pitch will probably narrow down, the chip gap will get smaller. So all this means that fluxless will be a better solution compared to a flux-based TCP solution.

So just to clarify, when you talk about two HPM vendors, does it include the biggest market share player? Because I thought they are still using the incumbent vendor, right?

Yes, of course. Of course, as I said, we have one order from two of the three. So definitely, yeah, we are talking to the leading one. Got it.

Understood. Maybe next question is, I think you observed some pause in AP and TCP in Q3. What is the reason for that? And given your guidance for 7% Q1Q growth for Q4, could you talk a little bit about how semis overall and AP and TCP within that will be growing? That'd be outgrowing that 7% or it'll be growing slower than that 7%. Thank you.

I think in terms of when we talk about pause, actually it's really largely driven by the timing of key customers' technology roadmap. So we're confident that when they launch the new architecture, we will get the order. So it's a matter of timing, in our opinion. Now, so this CB demand, whether it's in terms of booking or billing, will actually align with this timing as far as you're concerned. So it tends to be a bit lumpy here.

Is that more about logic or is it more about HPM? And also, any indications on, like, segment-wise, how are we thinking?

uh google i'll say both because the the technology roadmap will drive both hvm as well as on the logic side as well yeah okay and at q4 any thoughts i think in terms of if you if you're alluding to booking um maybe maybe it's time for me to give you some color on booking for q4 right i'm sure this question will pop up including the conference call as well. Now, the way we look at Q4 booking color, on the group wise in Q4, group wise booking in Q4, we expect bookings to be kind of flattish compared to Q3 reported number. Q3 reported number was US dollar about $462 million. So, going forward in Q4, we expect that to be kind of flattish on the group basis. However, we do expect that this Q4 booking for the group will be the seventh consecutive quarter of year-on-year booking growth since Q2 2024. So it's encouraging to note that we have been growing our bookings for seven consecutive quarters. Encouragingly, we see semi-bookings are expected to increase by meetings, Q&Q, mainly due to TCA. So when I say Q&Q meetings, I'm still comparing against the reported number. So that's for the semi-bookings, expected to increase, meetings, Q&Q, mainly due to TCA. on the Q2 basis compared to Q3. And for SMT, we expect SMT to decline Q on Q due to the high base effect in the prior quarter. Now, going into the Q4 bucket for SEMI, comparing the POR for Gen1 substrate application, and as the market moves towards larger compound dive because of higher computing power requirement we are confident of achieving a accessible tcb order for os application in q4 from the leading boundary offset partners and these orders will be lightly built in early part of 2026 which will be definitely gross margin accretive. So I think to sum it all in terms of Q4 and certainly beyond Q4 in terms of booking color, we remain confident that the strong AI tier wins, including the recent news regarding investment in the entire ecosystem for AI will continue to drive demand for AP. In particular, our TCP technology leadership will position us strongly going into Okay, that's very clear.

Thank you.

Thank you, Goku. And next, I would like to request Donnie to unmute.

Thank you, Robin and Benjamin, for taking my question. My first question is the housekeeping question. So, considering we have disposed the AEC operation in China, wondering if Katie can give us some colors on how should we estimate the OPEX? or OPEX ratio in the coming quarters as we have seen the OPEX ratio has been pretty high for the past few quarters so wondering if it would be coming down after the disposal of the AEC operation and also some cost control management and my second question is regarding to the TCB so My understanding is that despite of we have some progress in fluxless TCB, but the real volume shipment remains small into maybe fourth quarter this year. So I just wondering if you can give us a timeline when exactly the fluxless TCB for memories and for leading foundries can ramp up more significantly. in the future and also some comment on the progress in china will be also appreciated as you know that china has been aggressively increasing their ai production capability including hbms as well thank you

I think I will take the first portion, Donnie. So you asked a question about AEC liquidation. I just want to make a correction. For AEC liquidation, as we announced, the savings was going to be R&B of $150 million each year. Majority of that saving actually would be benefiting COGS, not OPEX. There would just be a little bit of factories and G&A that would be part of OPEX. So that – so AEC – let me just spend a quick minute. The liquidation took – sorry, the announcement took place in August, and the project's been progressing pretty well. And we do expect that the savings will – benefit us going forward. Now, on the OPEX ratio, and specifically on OPEX, there's actually no change. At the beginning of this year, we announced that we'll be investing incrementally $350 million Hong Kong dollars in R&D, especially AP and the infrastructure of the company. So every quarter, we are on the path of the investment. And because of that incremental investment, we've mentioned in prior quarters that this year's OPEX will be similar to prior year with some marginal increase. And that narrative has not changed and will not change for the year.

Okay, I'll take on the second question, Donny. In terms of TCP fluxes application, as mentioned in MD&A, we have made very good progress in terms of fluxes application TCP for a logic side, chip on wafer. I think plasma technology has been endorsed by the leading foundry. And also just to recap only, we have been saying already in the past, but it's good for a recap, that G1 wafer demand this year, even if we have won the technology battle, the G1 wafer demand will not be significant this year. We are looking into 2026 for an inflection point in terms of G1 wafer application for logic, you know, TCP fluxless. Now, I think that's your question, if I'm not wrong.

Yeah, this question, the timeline for the memory and reading foundry shipments.

Yeah, I think in terms of fluxes, I answered the first question to Goku already. So I think it all depends on when they will adopt the fluxless TCB for memory. As I said, in our opinion, as the industry continues to stack higher, the chip cap gets smaller. More IOs, in our opinion, at some point, it's quite inevitable that they have to move towards a fluxless TCB solution, even for HPM.

Thank you. Any color on China's adoption of TCP or opportunities there?

I think we have been saying we supply to the global customer base. I think in terms of volume, obviously the rest of the world's volume in TCP is huge. And for sure, China is the ambition to really step up in terms of advanced packaging.

Okay. Thank you, Tony. And next, I will request Kevin to unmute.

Hi. Thank you, Benjamin, for taking my question. My first question is on the TCP outlook. As mentioned, on the logic side, we are already passing the qualification, right? So I was wondering, how should we think about the potential business opportunity on the chip to wafer part as compared to a chip to substrate? As mentioned, most of the contribution will be coming from next year. And when is it likely the timing of this contribution will start? And also on the memory side, I think we just mentioned that HBM4, we are securing order from multiple customers, right? So just wondering for the customer, are these for simple tool or for production already? Thank you.

I can answer your first question first, Kevin, in terms of Cheeto Wafer. Quite similar answers to Tony. Cheeto Wafer, in terms of volume, compared to substrate because substrate I think the whole industry has moved almost the whole industry has moved to TCP solution whereas for chip2wave at the moment is only the leading company leading in terms of using a TCP for critical end customer So if more end customers adopt TCP, then you will see chip on wafer, TCP solution fluxes will increase. Otherwise, it's just one customer. I think the volume will still be smaller than the substrate volume. Now, in terms of HPM4, I would say they are already into some kind of a small volume production, already using our tools for HPM4 production for the two customers that we talked about.

Thank you, Robin. My next question is on the hybrid bundler side. So, I was wondering, How competitive are we in our Gen 2 hybrid boundary, which, according to our announcement, we are already shipping? And what kind of chip or process are these for? Or this is going to be for mainly on the logic side or for the memory side?

Kevin, we are shipping HP Hybrid Bonding solution for both logic and memory. As we speak, we are actively collaborating with other key logic and memory players. and we're making good progress and all these projects are at different stages of evaluation. So we are hopeful that at some point when the hybrid bonding market takes off, we are there to compete with the incumbent.

Okay, so we have already, are we securing order from

this customer already or this is just uh right now still in the evaluation process uh uh yes stay in the version for some of these very key logic and memory players we are engaging them very actively as we speak yeah okay thank you robin okay thank you kevin and uh next i would like to request sunny to unmute

Good morning. Could you hear me okay?

Yes, everyone. Yes, please.

Thank you very much. So my first question is on a high-level direction lead. How should we think about the recovery of mainstream semi-solution from here? I wonder in the last few months, now given more magical impact on tariff, do you think the overall client sentiment is improving or not much change for 2026?

Thanks, Sandy. I think in terms of mainstream, I would say quite encouraging because mainstream are now also, I mean, AI also contribute to the mainstream I think mainstream, as you are probably aware, China is a significant portion. So we see China's volume has been picking up for the last few quarters. So that's supporting the mainstream quite a fair bit for both semi as well as SMT. Now, in terms of tariff, I think in the initial part of the year, in the initial period of the year, I think the tariff situation definitely has some impact on the sentiment of our customers. Now, I think with the tariff situation a little bit more stable, I think customers are now a little bit more confident, I would say, in terms of placing orders. That's why we're also seeing You know, we have a good orders coming from A-stream, Y-Bone, Di-Bone, and SMT are also seeing a mainstream application for putting chips on larger PCB boards for base stations and all that. So all these are also partly driven by the AI adoption. So in general, we see mainstream certainly coming up from the bottom, but going forward, we see mainstream stable. kind of provides that kind of stability for mainstream.

Got it. Thank you very much. And then I have questions on PCB. Maybe if you could remind us the lead time for you to make TCB tools nowadays. And in terms of orders, should you expect the inception point to potentially come maybe in first half or second half of 2026 for Logic and for HBMs?

uh i think for for for logic i think if we win that uh sensible orders for the chip on substrate for large larger compound i uh will will most likely realize that the revenue in the early part of uh 2026 uh for hbm um It all depends again on the timing of our key customers' technology roadmap. So if they accelerate, we will see a preview earlier for HPM. If there's a further delay, then our timing will also align accordingly. now in terms of tcp lead time actually internally we are efficient we don't take a long time to assemble a tcp machine it all boils down to material supply right so if we if customers give us more visibility we can you know order materials earlier then the lead time will be shorter So I think that's the dynamic of the TCP lead time at this point in time.

Thank you. Sorry, maybe quick follow-ups. So for logic, so on chip-on-wafer, any view on when the lean foundry may start to migrate to KCB? Maybe would that be in second half of next year or early 2027? And therefore, assuming if your lead time is about like two quarters, should we see orders starting to come through maybe from first half of next year?

We are hoping others will come sooner, but again, as I said, it all depends on the timing of the roadmap. We are confident that chip to wafer will have a delay rate of shipment in 2026. I don't think it will delay to 2027.

Got it. Thank you very much.

Thank you, Sunny. Next, I would like to request Daisy to unmute.

Hello, Robin and Katie. My first question is for Katie regarding the semi-solution gross margin. Katie, you previously mentioned that the closure of AEC will have a positive impact of the cost of goods sold going forward. Yeah, so how we should think about the near-term and the long-term gross margin for the semi-solution segment?

Daisy, assuming you're kind of talking about basically the gross margin going forward, right? Yes, yes. Okay, so first on the AEC point, yes, correct. We would expect the savings to come in gradually in Q4 and then full-fledged in next year. Now, in terms of the overall semi-Q4 gross margin, we do not provide guidance, but, you know, just some kind of direction pointers. The guide of the Q4 revenue probably could tell that the TCB contribution, revenue contribution will continue to be lower, but with some hypotonics then. but wide bond momentum will be sustained. So, therefore, we expect a slight margin accretion for semis margin in Q4. And then, you know, and look at the group level then, you know, if semi and SMT mix stays similar and SMT experiences a stable margin, then we expect basically a slight margin accretion for the group in Q4. Now, of course, we always, you know, caveat, right, it's really depending on the mix going forward, especially in the midterm, you know, kind of longer run, we... The technology leadership in HBM and advanced logic, with those leadership, we expect the TCV order in Q4 and beyond, I'm talking about mostly in the mid-term now, would actually provide support to SEMI's gross margin. And with this liquidation that you mentioned earlier, we do expect that the SEMI gross margin will come back to the kind of the mid-40s level.

Thank you. It's clear. And the second question is for Robin on the Hyperbond. So you are at the evaluation stage for the leading Foundry and HBM customers. So for the HBM user Hyperbond, do you see that it will happen in 16 high or 20 high?

Since we are a dominant TCP player, we hope that they can continue to use TCP even up to turning high. But nevertheless, we are prepared that if they have to switch to hybrid bonding, we will be there also to provide competitive solution for hybrid bonding for HPM turning high and above.

Yeah, and also quick follow-up for your leading foundry customer. Your European peer has been a dominant supplier for hybrid bond at that leading foundry customer. So how do you see your hybrid bond opportunity at this leading foundry customer?

Yeah, we will be relentlessly knocking on their doors for sure. But I think having said that, we also have been saying that, you know, because we are not the leading player in hybrid bonding, I think the advantage is that, you know, we know there are pain points, existing pain points, right? So with that, you know, coming in from behind, We are relentlessly and diligently working with all these leading logic and memory players, asking them what are the current pain points so that we can incorporate features, engineering innovations to mitigate or totally eliminate those pain points. So I think we are confident that our Gen 2 and future Gen 3 should be able to address all these pinpoints and give us an entry point into all these leading key logic and memory pairs.

Thank you, Robin. And my final follow-up. So in Goku's question, you said that you are the primary supplier of the HBM4 market, and the first company won the HBM order at two key customers. So is it the fluxless TCB or the flux TCB?

It's still the flux TCB at this point, yeah.

A flux one.

Yeah, a flux one, yeah.

Okay, thank you. That's all my question.

Thank you, Daisy. Yeah, next I will request Le Ping to unmute.

Yes, thank you for taking my question. I have another question about the TCB. So what are the current customer concentration level of your TCB equipment now and what may it look like in the future? So is it mainly still concentrate on the top three memory maker and the leading foundry or you also see some broadening of your customer to other OSET or other foundries in the market? Yeah.

you know our tcp you know fan out to to not just leading foundry they're all sets hbm and and also you know globally as well uh so so we're pretty uh you know pretty engaged with all top ai customers needing uh requiring or requiring tcp solution i hope i answer your question okay yeah

Yeah.

Okay.

The second question is about this. You have the deposition equipment cancellation. So is it due to some the roadmap change of the in the advanced packaging? Also, I remember, is it due to the you have a company subsidiary called Next? It is from that subsidiary?

It is from Next. It is from Next. I think it's a case of digestion of capacity, right? So there was a build-up of a sizable capacity maybe about two years ago, right? So the customers take time to digest. And these particular customers decided to give it up, you know, and pay us a cancellation fee.

Okay, thank you.

Okay, thank you, Levin. Next, I will request Alex to unmute.

Thank you for taking my question. First question is about your margin on S&P solution. Seems like your quarterly revenue level already increased to the level similar to 4Q. about 23 or 24 so I see the margin is still like a low 30s. Is this the normalized margin going forward or you expect margin can return to high 30 level sometime in the future?

Yeah, Alexis, Katie, thanks for the question. You're kind of comparing to a few years ago where actually the SMTs and markets composition were quite different. A few years ago, actually automotive and industrial were running really, really strong, and their contributions to SMTs revenue were much larger. And this is where we actually could commend a relatively higher margin So currently, as we mentioned, the automotive and industrial end markets are relatively muted, and that's why the margins are sitting in the, you call it, low 30s. To me, unless the end market composition changes, this kind of level will be sustained in terms of margin percentage.

Thank you. Another follow-up question on TCP. You mentioned the TAM would reach like $1 billion in 2027. Do you have probably a rough split between the logic versus memory and also split between C2W applications? Thank you.

Yeah, I think it's dynamic. I would say, Alex, it's very dynamic. Again, it all depends on the customer roadmap and all that. But generally speaking, if you take, you know, really looking further into the future, it's just intuitive that the HBM, TCB demand or size or time will be larger than logic. because of the number of stacks and also as the industry migrate from one architecture to the next generation, they require more HBM stacks per chip, right? So naturally, I think, you know, HBM demand over time, not in a particular year, not in a particular quarter, but over time, HBM demand for TCP will be larger than object.

Got it. So what is the company's target market share for each, you know, application?

We don't go down to that kind of granular level, you know, HBM market share or logic. But overall, I think last year we put up, when we put up the temp, you know, for TCB, our aspiration is to hit 35% to 40% market share in the entire TCB temp.

Got it. Thank you.

Thank you, Alex. And next, I would like to request Arthur. Arthur to Amir, please.

Hi, Alex. Can you hear me?

Yes, Arthur.

Hi, thanks, Robin, Katie. Take my question. So the first one, Robin, if you can, can you share with us a high-level ballpark figure on the revenue contribution from the A.I.? ?

This is a difficult question. We don't share, but also this is a difficult question because we talk about AI benefiting both AP and mainstream. Well, we have better visibility on how AI benefits AP, but in terms of mainstream, it's a little bit tricky because, you know, wire-borne, die-borne, you know, they are quite fungible. You know, today, customers may say, okay, I use it for AI-related packaging. Tomorrow, they use it for others. So, it's a bit difficult to really nail down. Sorry.

Yeah, no problem. Because you just mentioned that you saw some power application, they stopped to come back and the driver is from the AI. So that's why I want to, you know, get this high-level Bopar figure. Maybe we can discuss it next quarter when we have a visibility. The question number two is on the uh cancellation from the high density subject and i think leping already touched base a little bit um so my question is a is the key component of the equipment fungible can you uh you know give it to the other uh substring customer yes it is the short answer yes it is there's no inventory related issue relating to this cancellation Yeah, thanks. Because if we look into the AI business of the REC and also the key component, actually, we heard more and more PCB, HDI, substrate shortage at this moment. So I'm kind of wondering, was the client based in Japan or in Taiwan or China?

None of this actually, none of this. I mean, this is the next business where I'm saying that they supply to few key players, high intensity substrate players. It just happened that this, as I said, I repeat again, it just happened that there was a big capacity ramble in the last two years and this particular customer just said, okay, I'd rather not keep you holding on to all these orders. I decided to cancel it. So, So I think that, in short, it's that kind of circumstances.

So in the future, when we look back, this could be an isolated event. So do we think this demand for the other customer will return? No, no.

you are thinking, is this AI related? I wouldn't say this is AI related. Yeah, this is, they are serving a particular IDM, you know, which use, you know, all this equipment, you know, for RDL and all that. So it's a particular application, I would say it's not related to AI. So don't link this cancellation please don't link this translation with AI that we have been talking about. Dealing with these two, please, Arthur.

Okay, okay. Thank you.

Nothing to do with this translation has nothing to do with AI.

Thank you, Arthur. I think we have time for one last question. I think we have Goku here. So, Goku, could you please unmute?

Thanks, Ben, for taking my question. So my question is more on the margins and operating leverage. I think we are having pretty good momentum both now in mainstream and in TCP. Margin still seems to be a little bit sluggish. I think, let's say, next two, three quarters, TCB revenues will come through given all these orders, bookings into revenues. What does it do to gross margins? Like, is TCB still accretive to group gross margins right now or is it kind of similar to group gross margins? Second part of the question, again to Katie, is on operating leverage. Because now that we are back to some degree of revenue growth, we're still not yet seeing meaningful operating leverage come through. I think because speed expectations are for very big operating leverage to kick in for next year. I think revenue growth of 10% or 15% contributing to doubling of your operating profit is what a lot of the Bloomberg estimates are looking at. So just wanted to understand what is the extent of operating leverage that we can expect? I think we have seen operating margin go back to high teens to 20% at really, really peak kind of levels back in 2021. In the recent past, we've not really seen operating margin really get beyond the mid-single-digit levels. So just want to understand what is the extent of operating leverage we can expect as we start some of these RAM pubs for TCP and other products.

I appreciate the question. First thing, TCB, I just want to make it very clear that TCB margins being stable and is accretive to semi-business. Now, overall, when you say operating leverage, volume has come back up, but not quite at the super cycle level. And within the volume, like we'll always say, there are a few mixes that actually impact margin. One is a segment mix. So far as you can tell, like in Q3, for example, the SMT contribution to the group is at about 50%, right? SMT naturally has lower gross margin. Therefore, the segment mix could be higher. different based on the contribution from the two businesses. The other thing is on product mix. Within SAMI, for example, it really depends on the product mix, say, between TCB and Y-bound. In Q3 and As we guided for Q4, if you look at that product mix, when we have less TCB revenue but more Y bond revenue coming from mainstream applications, the gross margin side would not – would be under certain pressure. But having said that, in the long run, as a few of you asked earlier, we do expect that our semi-business will continue to enjoy the accretive margin contribution from applications like CCB. And with the AEC liquidation we mentioned, we should have savings from operation efficiency, et cetera. So I think our conviction for semi-gross margin to stay in the mid-40s And then gradually going up has not changed. And then so at the group level, we've been talking about the 40%, right? I think, again, I'm talking about in the long run, not a specific given quarter. I think we are comfortable that the group's gross margin will be at that level and gradually improve as we go.

Got it. So just on the OPEX side, because that's something that you can control, revenue order to control, especially on mainstream. Are we going to stay around this roughly 5 billion Hong Kong kind of level going into next year? Or we still see that OPEX will keep growing even we are investing in some of these newer technologies?

Yeah, so, Gogo, I actually cannot answer your question very well right now. Maybe give us a quarter because the organization actually is going through the budget process. But directionally, as we have talked about before, the OPEX has been running at $4.7 billion in the last few years. And this year, you know, with the R&D and infrastructure investment, we have communicated that it will be marginally higher. Though the investment is at $350 million, we are doing certain restructuring projects and the cost-saving projects that you probably have seen the last few years on trying to bring it down. So this year, I think you guys can do the math, right? It's about $2.8 billion. So that's kind of where we are. I think going forward, we're not going to change our commitment in R&D investment. As you guys were talking about, TCP, Hyperbond, all that, that side of the commission has not changed. We'll continue to do the right investment. On the other front, for the overall efficiency and productivity of OPEX, we'll continue to look into any opportunities we can find. and trying to contain that. So again, I cannot give you a specific number. We'll probably share with you more, but I think our strategy or our thinking on OPEX has not changed.

Got it. Thank you very much. Thank you, Koko. That will be all for the last questions, and I will now pass the time back to Robin for his closing remarks.

Thank you, Benjamin. Just a couple of pointers before we officially close the call. The group maintains from business momentum this quarter, our AP and mainstream business will continue to benefit from sustained AI adoption. TCP solution will secure repeat orders in both memory and logic, reflecting ongoing technology leadership, particularly in HPM4 and advanced logic. Like what Kelly said, we are in the midst of really finalizing our budget for 2026, but certainly at this juncture, we can give you some direction or some color of how we look at 2026. We expect a growth year in 2026, largely driven by an AP because of AI, and underpinned by the sustained momentum of our mainstream business. And finally, we remain confident in the total investable market 1 billion in 2027. So thank you. With that, I will close the call and see you next quarter.