Asmpt Limited Unsp/Adr

Good morning and good evening, ladies and gentlemen. This is Leonard Li from the ASMPT RR team, and I'll be moderating today's call. On behalf of ASMPT, welcome to our 2025 Second Forward Investor Conference call. Thank you all for your interest and continued support. Please note that all participants will be in listen-only mode while the management is presenting. 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 do 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 should differ materially from those expressed or implied during this conference call. For reference, the investor relations presentation for our recent results is available on our website. On today's call, we have our Group CEO, Mr. Robin Eng, and our Group CFO, Ms. Katie Hsu. Robin will cover the Group's highlights, outlook, and next quarter's performance, while Katie will provide details on the financial performance. With this, let me now hand this over to Robin.

Thank you, Leonard. Good morning and good evening to everyone today. It is a pleasure to have you all on our earnings conference call for the second quarter of and the first half of 2025. Now let's start with the key highlights of the first half. Let me begin by saying that the strong demand continues to be driven by the AI tailwinds across our AP and increasingly the mainstream as well. For the first half of 2025, we achieved better than expected bookings and our revenue guidance for Q3 is above market consensus. The group's advanced packaging continued to grow with AP revenue contributing significantly to group revenue in the first half of 2025. This growth was primarily driven by the ongoing demand for thermal compression bonding or TCB tools. In the first half, the group secured repeat orders for TCB tools in both memory and logic applications, maintaining the largest TCB in stock base by surpassing 500 tools worldwide. In the mainstream business, the group is beginning to benefit from AI tailwinds. AI data center demand has driven Booking's growth on new power management capabilities. The group also experienced strong Booking growth in China, driven by electric vehicles and consumer end markets. I'm also pleased to say that we have maintained gross margin above 40% despite foreign exchange headwinds in the first half of 2025. With that overview, let me go into more detail about advanced packaging, a business that is growing and we remain confident will continue to do so. In the first half, our AP business increases revenue contribution to around 39% of the group's revenue. or approximately US$326 million, driven by strong AI tailwinds. TCB has continued to be the largest AP revenue contributor and remain a key growth driver. Orders in the first half were up 50% year-on-year as it gained further traction with customers, including major AI players. The group's leadership position in TCB across both logic and high bandwidth memory, or HBM, supply chains continued to strengthen, supported by the expansion of our AP customer base. During the first half, the group secured TCB orders from various HBM players, further reinforcing our leadership position in this market. The group successfully installed the bulk order of TCB tools for the leading HBM customer, fully meeting their high-volume manufacturing requirements for HBM-3E-12H. These tools have demonstrated outstanding performance, delivering industry-leading production yields and exceptional interconnect quality. Additionally, another key HBM customer began low-volume manufacturing for HBM-4-12H before TCP. In the HBM4 market and beyond, the group continues to maintain its technological advantage due to its Active Oxide Remover or AOR technology. These innovative capabilities enable us to support customer as they transition to next generation HBM and beyond. AOR is a key differentiator, facilitating the demanding requirements of HBM4 and beyond. This includes higher input-output connections, more challenging die-bomb layouts with finer-bound pitches, thinner dies, and a high number of die stacks. Promisingly, the group is currently engaged in HBM4 AOR sampling builds for multiple customers. Turning now to chip-to-substrate, or C2S-TCB, The group secured additional orders in the first half of 2025 for C2S solutions at a leading foundry's OSAP partner. The group also delivered several high volume shipments of these TCB tools in the first half of 2025, serving as a sole supplier for chip to substrate. Meanwhile, our joint development of ultra-fine-pitched chip to wafer for C2W logic applications for next-generation AOR-TCB with the leading foundry is progressing from pilot production to volume production. Moving on to hybrid bonding, the group expects hybrid bonding to coexist with other packaging technologies and its adoption will be gradual. We continue to see progress with our both first and second generation hybrid bonding tools with various customers actively engaged at different stages of setup qualification and shipment. Notably, our second generation hybrid bonding tools feature competitive capabilities in terms of alignment and bonding accuracy, hook-free and UPH. As previously announced, we expect to ship this second generation tool to an HBN customer in Q3. In addition, There is also continued collaboration with a leading IBM, a leading research institution, and the leading foundry on our tool capabilities. Now turning to photonics and co-optic package or CPO. Rapid AI growth continues to increase data center bandwidth requirements and boost demand for high bandwidth optical transceivers and co-optic package CPO applications. Our photonic tools are able to package this high bandwidth transceivers, especially 800G and above. Due to a clear market relationship, we expect continued order momentum from global transceivers market serving all major AI players. While the CPO market is still in an early phase, we are actively working closely with leading CPO players around the world. In the first half of 2025, we had a major win with a leading IDM and a well positioned to grow our market share. Finally, the system in package of SAP business within AP, SMT won orders in the first half of the year from the leading global high-end smartphone players for radio frequency modules and wearables. In addition, SMT has been gaining traction with its next generation chip, SMT2, in several areas, including AI-related applications, with shipments leading to leading foundry and OSAP players. Next, I will turn to our mainstream business. During the first half of 2025, as I mentioned earlier, AI tailwinds are beginning to benefit the group's mainstream business. Demand for AI data centers has driven increased needs for new power management capabilities among all major AI players. AI growth requires more power-efficient data center racks to meet the shift towards 800-volt high-voltage DC power distribution architecture. This has driven increased demand for semi, wire and die bonders, and SMT placement tools. In addition, in the first half, the group achieved a strong half-and-half and year-on-year bookings growth in China. For SMT, the growth was primarily supported by AI and EVs, where we continue to be the leading EV player in China. Meanwhile, SAMIC saw increased utilization across OSEP providers, serving both consumer and EVM markets. With that, Let me now pass the time over to Katie, who will talk about our group and segment financial performance.

Thank you, Robin. Good morning and good evening, everyone. This slide covers the group's key financial metrics for the first half of 2025. The group delivered revenue of $837.6 million. Semi delivered strong revenue growth, 31.7% year-on-year and 6.1% half-on-half. while SMT experienced revenue declines year-on-year and half-on-half. Group's booking reached $912.8 million, which was better than expected, showing 10.5% growth half-on-half and 12.4% growth year-on-year. The group continues to build backlog, with two quarters of bookings above one. In the first half, the group's gross margin was 40.3%, up 121 basis points half-on-half, but down 65 basis points year-on-year. The half-on-half improvement was primarily due to segment mix, but the year-on-year decline was mainly due to an unfavorable product mix in SMTs. The group's operating expenses reduced by 6.3% half-on-half, but went up 1% year-on-year. The half-on-half OPEX reduction was due to the group's prudent spending controls and restructuring benefits, despite strategic R&D and IT infrastructure investments. The group's operating profit reached HK$329.3 million, showing 79.5% half-on-half growth, but a 12.2% young year decline. The half-on-half improvement was driven by gross margin improvements and OPEX reduction. As a result, adjusted net profit was HK$218.1 million, up 95.7% half-and-half. Our half-and-half improvements were driven by tax credits from R&T centers in Europe and Asia, but partially offset by unfavorable foreign exchange translation from a weakened U.S. dollar, despite the group's hedging facilities. Similarly, the year-on-year decline was also due to this unfavorable foreign exchange translation, partially mitigated by favorable tax credits. We have an existing dividend policy of distributing about 50% of the annual profits as dividends. Therefore, for the first half of 2025, with EPS at 52 cents in Hong Kong dollar, the Board has recommended a dividend of 26 cents in Hong Kong dollar per share in line with this policy. Our business remains focused on increasing shareholder value and continue to evaluate options to return excess capital to shareholders. Most of the financials on this slide are covered on the previous page. I will not go into the details, but I will touch on revenue by end markets. Computers became the largest contributor to the group's revenue, supported by strong growth driven by AI. Automotive is the second largest contributor supported by EV demand in China. The next contributor is communication, supported by demand from photonics and high-end smartphones. This is followed by consumer and industrial end markets. Now, let me move on to Group's Q2 financial results. We delivered revenue at approximately the midpoint of the revenue guidance, totaling $436.1 million U.S., an increase of 8.9% quarter-on-quarter and 1.8% year-on-year. The quarter-on-quarter improvement was mainly due to growth in SMT, but semi remained flat. The group's booking reached $481.6 million, which was better than expected for the second quarter in a row, showing 11.9% growth quarter-on-quarter and a 20.2% growth year-on-year. These increases were mainly due to the growth in SMT, Future book-to-bill ratio was 1.1, and as I mentioned earlier, has now been above 1 for two quarters. In the second quarter, the group's gross margin was 39.7%, down 119 basis points quarter-on-quarter and 30 basis points year-on-year. The quarter-on-quarter decline was mainly due to a decline of 161 basis points in semi, while SMT improved by 108 basis points. However, Q2 gross margin would have been above 40% using Q1 2025 foreign exchange rates. The group's operating expenditure was HK$1.18 billion, indicating a 5.7% quote-on-quote increase and a 1.8% year-on-year reduction. This was largely due to strategic R&D and IT infrastructure investments and the foreign exchange impact. although partially mitigated by prudent spending control and restructuring benefits. The group's operating profit reached 169.4 million Hong Kong dollars, showing 5.9% growth quarter-on-quarter and a 25.4% year-on-year. Q1Q was mainly due to volume effects, while year-on-year improvements were due to OPEX reduction and higher volume effects. As a result, adjusted net profit was HK$134.9 million, up 62.1% quarter-on-quarter, but declined 1.6% year-on-year. The quarter-on-quarter improvements were mainly driven by better operating profits and tax credits mentioned earlier. Moving on to the semiconductor solution segment performance. For the second quarter of 2025, semi-revenue grew to $257.6 million, up 1% quote-on-quarter and a 20.9% year-on-year. This segment contributed about 59% of the group's revenue. TCB2's photologic and memory solutions were our largest revenue drivers in Q2. Y-bounders and Y-bounders showed quote-on-quote and year-on-year growth. This was supported by shipments to major IDMs focused on AI-related power management applications, as well as to China customers, especially OSETs. Semi-bookings were $212.5 million, down 4.5% quote-on-quote and 4.6% year-on-year. In Q2, both quote-on-quote and year-on-year experienced wide-bound and dive-bound growth. while TCB orders were down due to uneven AP order flow. Semi's gross margin of 44.7% for Q2 2025 was down 161 basis points quarter-on-quarter and up 19 basis points year-on-year. The quarter-on-quarter decline was mainly driven by product mix. Lastly, Semi's profit was HK$174.9 million in Q2 2025 a decline of 25.9% quote-on-quarter, but up 99.8% year-on-year. The quote-on-quarter decline was mainly due to lower gross margin and high operating expenses arising from strategic R&D investments. Year-on-year improvement was driven largely by volume effects. Next, on to SMT solution segment. SMT delivered revenue of $178.5 million in the second quarter of 2025, an increase of 22.6% quarter-on-quarter, but a decline of 17.2% year-on-year. The growth was mainly due to stronger revenue in China and AP, partially offset by continued softness in overall automotive and industrial end markets. SMT bookings of $269.1 million were up 29.4% quote-on-quote, largely driven by a bulk order to meet the supply chain diversification needs of a leading smartphone ad customer, as well as all the wins in the AI server market. Additionally, SMT's gross margin of 32.5% for the quarter improved by 108 basis points quarter-on-quarter, but declined by 311 basis points year-on-year. The quarter-on-quarter improvement was due to higher volume effects, partially offset by product mix and foreign exchange impact. Year-on-year decline was largely due to lower volume and product mix. This slide highlights ASMPGIS management's best estimates of revenue breakdown by end markets for the first half of 2025 compared with the first half of 2024. This highlights our exposure to diverse end markets. The computer end market was the highest contributor to group revenue, accounting for 30%. Strong revenue growth was mainly driven by continued demand for AI-related applications in both memory and logic. The automotive end market was the second highest contributor at 15% supported by EV demand in China. The communication end market contributed 13% to group revenue, with demand in photonics and high-end smartphone-related applications, continuing to support this end market. The consumer end market contributed 12% of group revenue, driven by semi-mainstream solutions, particularly for China. Lastly, the industrial end market contribute 8% of group revenue, in line with soft market conditions. As you can see from this slide, we're a truly global business, partnering with customers around the world. China registered year-on-year revenue growth, increasing to 36.7% of group revenue. AI demand supported growth in revenue from Korea to 13.6% and Taiwan to 10.6%. Revenue share from Europe and America declined year on year, mainly due to market softness in SMT, with Europe's share down to 11.4% and America's to 12.3%. The group maintained a diversified customer base, with the top five customers accounting for approximately 24.8% of total revenue in the first half of 2025. I will now pass the time back to Robin

Thank you, Katie. Looking ahead to Q3 2025, the group expects revenue to be between US$445 million and US$505 million, up 10.8% year-on-year and up 8.9% quarter-on-quarter at midpoint, which is above market consensus. We are confident of sustained AP revenue and expect SMT revenue to improve. The group remains confident that AP will continue to grow, benefiting from the strong AI tailwinds and our technological leadership in the market. We reiterate our TCB total addressable market projection of US dollar 1 billion in 2027 and remain focused on solidifying our TCB market leadership in both memory and logic applications. The group's mainstream business will be supported by momentum in China and opportunities driven by the emerging demand for AI data centers. However, the automotive and industrial end markets will remain soft in the near term. While the group has not experienced negative impact from tariff policies, it acknowledges that uncertainties remain The group's global presence provides flexibility to navigate any potential impact, and we will continue to monitor the situation closely and adapt as needed. This concludes our second quarter and the first half of 2025 presentation. Thank you, and we are now ready for Q&A. Let me pass the time back to Leonard to facilitate it.

Thank you, Robin. We will now proceed with the Q&A session. If you would like to pose any questions, please raise your hand and we will unmute you. Please be reminded that we can take a maximum of two questions each time. Hi, Donnie. Please unmute yourself and go ahead with your question.

Thank you, Leonard. Thank you, Robin. Thank you, Katie, for taking my question. My first question is as usual, the housekeeping question. I'm wondering if you could kindly give us some colors on the booking trend into the third quarter across the different businesses. And secondly, the SMT booking has been recovering strongly in the second quarter. And as you just mentioned, it's part of the reason driven by leading smartphone vendors' capacity diversification across the world. So just wondering, will this trend continue into the second half driven by the tariff uncertainty or it has been majority that have happened in the first half already? Thank you.

Thank you, Donglin. I'll take your question. The first question is on the booking credit for Q3. Yes, I think the way we look at Q3 booking, Q and Q, for Q3 versus Q2, we think it's going to be slightly down by, say, single digit percentage. And then on a year-on-year basis, we expect the bookings in Q3 to be up double digit percentage. So that kind of range we are kind of expecting for bookings in Q3. Now in terms, if you drill down a little bit more, QNQ declined largely due to SMT because due to the absence of a big order that we have in Q2 SMT. So I think that caused the Q2 SMT booking to come down. But for SMT, Semic, it will be in terms of the AI will continue to be strong. We believe the momentum is there for AI in Q3. So that would drive TCB booking as well. And as we have said earlier, you know, MD&A as well, we see emerging also opportunities in terms of AI data center, the field of AI data center, increasing our semi-mainstream demand as well as our 70 placement tools. The semi-demand will come mostly from the normal dive ponders and the wild ponders. I think probably last thing to know about year-on-year group bookings for Q3, we see that the momentum will continue. If you look back, our year-on-year bookings have been on the rise for six quarters already. So I think this is something probably worth noting. So I think that will answer your question number one. Now, in question number two, you're asking about SMT booking recovery from Q2. Yes, indeed. So we want this bulk order from two customers. This is part of our end customer. We suspect it's part of the ultimate end customer's diversification, what you call that drive, right? to have manufacturing capacity diversified in other parts of Asia. You're asking whether this will repeat in the second half. We don't exactly can tell, but we're hopeful that there will be another order coming in the second half. But even if that happens, we will be not as material as the one that we have got in Q2. So I hope I answered your question, Tom.

Thank you, Robin. Just a quick follow-up. So you mentioned about the semi-bookings in Q2, Q3 should be improving. And you mentioned about die-bound or wire-bound recovery driven by maybe AI-related applications. So Just want to clarify, so for third quarter outlook, it's like conventional packaging like die-bounder, wire-bounder may be growing faster than advanced packaging?

Not really. As I said earlier, the Q3 booking, we see the PCB booking should increase QonQ. I don't have the rate of growth, but I will say AP will continue, the momentum will be there. For the semi-tools at the die-bond and the wire-bond, yes, indeed, we are seeing growth, driven by two factors. One is the AI data center, because a new power management requirement, and two, also driven by China. We see increasingly the momentum in China mainstream, for both mainstream, SAMI and SMT is on a good track.

Okay, understood. Thank you, Robbie.

Thank you, Donnie, and may I now ask Sunny to unmute yourself?

Hi, Leonard. Could you hear me okay?

Yes, yes, we can.

Thank you very much. So my first question is on TCB. If we could start from HBM. How should we think about maybe in the coming, like, 12 to 18 months, your order opportunities within this big player competition seems to be intensifying? And earlier in the year, you mentioned that you – We expect maybe a second bulk order to come through, maybe at some point second half or early 2026. So any update there? And how should we think about the competition into 2026?

Thanks, Sunny. I will answer the question, Sunny. The first part is how do we view the 12 to 18 months order opportunity for HBM? So let me start by probably recapping where we are right now. So in terms of HBM, as you are probably aware, you know, we have shipped already and installed, you know, the bug order that we received last year for a leading HBM player. that is for HBM3E25 that has gone into volume production using our tools and certainly we are pleased to announce that we have been performing very well in terms of the tools and definitely meeting customer expectations in terms of our technology in particular our yield and also our quality of interconnect. So I think that gives probably confidence that we will continue to garner more shares of the HBM market going forward. Now, in particular, we are confident of our next generation TCB2 technology for HBM, in particular for HBM4 and beyond. As we have been saying for many quarters right now, we have this AOR technology, Active Oxide Removal technology that can truly differentiate us from our competitors. And I think with the advent of the new AI chip coming into the market, I think HBM4 will be put to use, the devices HBM4 will be put to use. And since we are the first mobile for HBM4, we remain confident that going forward, we will continue to win orders for HBM4 using of tools and in the future also using our AR technology. So this is how we see the HBM market going forward.

So a quick follow-up is earlier you mentioned QCTCB booking should continue to grow. Is that driven by HBM or Logic?

Both, I would say both. In fact, on that note, probably worth to mention that we are also in the meantime expanding our customer base as well, globally, worldwide. So we have TCB tools for both logic and HVM shipped on a global basis. So we're spreading out, diversifying our customer base.

Got it. Thank you very much. My second question is on your TCB engagement with the leading foundry. So could you share a bit more color on the progress that you have moved to volume production, and would you be able to get the boat orders maybe through second half of the year or early 2026, since they should be ramping up the next generation air accelerators into second half of 2026? And are you the sole supplier for these maybe AOR type of PCBs for chip-on-wafer in 2026? Yes.

Again, yes, you are right. We have moved from... I would say it progressed from pilot production to volume production and the leading company for this very advanced ultra-fine pitch AOR-TCB for chip to wafer. There is competition. We have a competitor over there. But as far as we are concerned, I think our technology has put us on an advantage, we believe. And in terms of order, we mentioned many times, Sunny, that Though even we win this particular battle, you know, we were competitive over there, the volume this year for chip-to-wafer would not be material. Probably a couple of tools for 2025, but we believe that 2026, that's where the volume production will start for chip-to-wafer tools at the leading foundries. Back to you, Samir.

Got it. Also, for them to maybe start using your tools for chip to wafer, it could support the key accelerator platform upgrade in second half of next year. Should we expect, if that happens, the order should cancel maybe by early 2026, the latest?

Possible, yes, certainly possible. Definitely, yeah.

Okay, got it. Thank you very much.

Okay, thank you, Shadi. May I ask Goku to unmute himself, please?

Hi. Good morning, Robin and Kay. Thank you. Thanks, Lynette, for taking my question. My first question is on HPM for TCP. Could you talk a little bit about HPM4? What are you hearing from your customers? Are they basically going to fluxless TCP for HPM4 across the board? using your AOR or other kind of flexless technologies, or can they still reuse the existing flex-based older TCP tools of which they have a pretty large install base of? And HBM3e, especially at your lead customer, looks like very competitive right now. There are three vendors. Do you think that HBM4 also is going to be like that, or you think the vendor list will narrow when it comes to HBM4?

Thanks, Gokul. I'll take your question now. I would say not across the board. It depends on the customer. One customer is still doing a lot of a lot of experiment using and not using AOR and other customers have definitely started using our tool for sampling build for using AOR. So as far as we are concerned, as I said earlier, we are the first mover in terms of HBM4, so we are well positioned to capture the market for HBM4 when that takes off and we believe HBM4 will probably come around sometime in the second half of 2025, in line with the launch of the new AI chip architecture. Now, I think your second question is about HBM3e with a late customer versus competitors. HBM4 will be a big vendor, this one will be narrow. I think all major players, let me try to answer you from the customer perspective, the way we see it, I think all major HPM players will have to move to HPM4 to support the new AI architecture. One of them, probably a little bit ahead in terms of HPM4 deployment. As I said, we are doing something built for all customers actually. Two of them we have a midshipman. I think the other The other leading HPM players, we are also engaging them in terms of sample build, using their vehicles for our site. So we have a lot of engagement with all these HPM players.

Just to follow up, Robin, just to clarify this. So when it comes to HBM 412i, which is probably the one that's going into production early next year, you think most of your customers, especially the lead customer, will have to use AOR or fluxless PCB, or they can still stick with older PCB machines?

They will try to use the, you know, they will not try to transition to a new, if they can. But the way we see, once they start to move into HBM 4E, where the chip architecture gets a little bit more challenging, we strongly believe that the ALR will be the technology that they have to employ.

Okay, understood. So that's more 40 than 4. Okay, that's clear. Just on the logic side for PCB, could you talk a little bit about your chip-to-substrate shipment run rate? It's been very strong since second half last year and early part of this year. Do you think it grows into next year, or given the COVOS capacity expansion is kind of decelerating, we should see a little bit of slowdown in the chip-to-substrate growth next year? And is C2W chip-to-wafer big enough to kind of offset that next year so that your logic PCB can still continue at a pretty good clip, even if chip-to-substrate kind of slows down?

Yeah, on the chip-to-substrate, the way we look at it is the dye and the compound dye are getting bigger and bigger. In fact, there's really no way except using TCP. you know, to package the compound dye onto the substrate. I think as we speak, you know, our customer base are already thinking of, you know, new tools, new tool architecture, you know, to handle larger and larger dye. So we are already in a lot of engagement with all these customer base, you know, to come up with new tools that can handle a larger compound die. So I think this trend, if this is a multi-year trend, it will continue as long as data centers continue to have this massive build-up into the future. Now, in terms of C2W, your question is whether it's big enough to offset logic TCP growth, even as cheap to substrate slows. I'm not sure whether, you know, cheap to subscribe was slow. As I said, you know, you're continuing. In fact, you're trying to continue. So let me answer the other part, whether cheap to waive or is big enough. I think, as I said earlier, I answered the question posed by Sunny. uh tip to wafer tcb tools i think the demand will pick up in 2026 because right now the por is still mastery flow but we believe increasingly even at the wafer level tcb is needed to package all the various type including hbm as far as logic die uh passive die onto the uh wafer going forward sorry go back to

Yeah, so then we're expecting that logic TCP C2W combined everything is still going to be growing next year based on the order flow that you've got. Because the order flow seems to be a little bit up and down, right? You had a decline in TCB last quarter. This quarter is kind of growing. So I just wanted to understand how logic TCB next year.

Yes, yes. I think Goku, remember, we came up with a kind of a time picture for TCB. So the time will continue to grow from now to 2027. And we're still confident. that by the time we hit 2027, that time will be $1 billion for TCP.

Got it. Understood. Maybe one last follow-up for KD. Could we talk a little bit about the margin leverage from increase in advanced packaging? Because when I look at semis margins, margins are kind of largely still in the same mid-40s kind of warts. once packaging mix has risen almost to 40% in first half of the year, even though you don't break it out. But we don't seem to be getting that margin uptake on gross margin or operating margin level in a meaningful sense when I look at the semiconductor solution business. Is there anything that we should look at it or is there any turning point where we could kind of see that we hit that meaningful gross margin leverage or operating margin leverage happening from this mix improvement?

Yeah, so Goko, I think there are a couple of layers to the question. First, maybe just a little bit of color on Q2. So for semi-Q2 gross margin, the mix actually was relatively more favorable towards the mainstream products. You know, comparing to Q1, which you probably remember where we delivered a bulk to the order, right? was very much supported by the bulk order for AP. Now, also kind of going forward, like what we've mentioned before, TCB margins are creative. And we do expect that in looking out to future quarters that the margin expansion for semi will be there. But, again, like we always say, right, each quarter really depends on the product mix volume, et cetera. But in the long run for semi, yes, definitely we do expect that the margin will expand gradually because of the TCB slash AP content.

Okay. All right. Yeah, thank you very much. I'll go back to the queue.

Thank you, Goku. May I now ask Daisy to unmute yourself, please?

Thank you, Leona. And hello, Robin and Katie. I've got one question regarding the new power management capability. As you mentioned, that AI data center demand has begun to benefit the group's mainstream business driven by the increased need for new power management capabilities. So could you elaborate more what other semiconductor components in the is power management? Is it power discrete or it's a component like PMIC?

Yeah, I think that is a varied range of devices or power components that are needed, you know, in the AI data center, right? That's why we see that it's driving a demand for mainstream tools like, for example, Y-Bone, Y-Bone, molding equipment, for example, sintering equipment. So we see this trend happening in the last two or three days.

Okay. So do you think that this trend will benefit all the vendors or because of ASMPT's leadership technology, it will benefit ASMPT more versus your competitors? Yes.

I think these are pretty standard tools. I think these are more, you know, pretty more capacity-buy than technology-buy. Typically more towards capacity-buy than capability-buy, in our opinion.

Okay, thank you. Robin, it's clear. And another question is for Katie. And you mentioned that Q2, you benefit from their tax credit. Do you think that this is sustainable or it's just a one-off in Q2?

Yeah, Daisy, so this short answer is this is a one-off tax credit. We have R&D centers in certain jurisdictions where we have tax credits, and usually, actually, tax credits will flow through the P&L. You probably won't even notice it every quarter, but for these two specific locations, due to the local practices, the tax credits could only benefit P&L after they filed the audited statutory reports. So for these two locations, the 2024 audited statutory reports were filed in Q2 this year, and that's why it came through into the Q2 financials. I do not expect them to repeat next quarter for sure. Thank you, Katie.

It's clear. Thank you.

Thank you, Daisy. And I may now ask Le Ping to unmute yourself and ask your question.

Yeah, thank you for taking my question. So here are two questions. My first question is about the SMT business. So you mentioned that the order went in the AI server market. So recently we also see some PCB vendors are ramping up their capacity for the server PCB. Are these two things related? And can you comment on where are the cutback cycle of the PCB's capacity expansion for the server market? And what are the market position of your SMT equipment in this market? Thank you.

Let me take a question. I think they should be related. Of course, without understanding more where they're coming from, but I think it sounds like they're related. And in terms of Quebec cycle, as far as the AI center continue to build up, I think this could be the trend going forward. So in terms of SMT positioning. I know we are pretty strong in this particular area because of our technology. So we are quite pleased that we are capturing this part of the market share.

Okay. The second question is, I noticed that the computer this quarter, quite impressive, account for more than 30% of the total sales. Can you share some color about the breakdown of this computer? How much is currently roughly coming from the SMT? It's quite big. And how sustainable do you expect this? computer will continue to account for this such high or the 30% market of your total sales looking forward in the second half.

Thank you. Thanks for noticing that the 30%, it is very high for the first half of revenue, but I just want to call it out that I just mentioned a few minutes ago that in Q1, we had a bulk order for TCV, right? So actually that's actually has supported the computer percentage going up. But your question, I think, specifically is on SMT. We do not break down the end market by the specific segments of ours. But overall, like Robin mentioned, the SMT side is benefiting from the overall AI server trend, and we have very strong position in that specific sector.

Okay, thank you.

Thank you. May I now ask Alex to unmute yourself and ask your questions, please.

Thank you, Benjamin, for taking my questions. So I'm just a little confused about your TCB order trend. So in your TPT, you mentioned in the second quarter, TCB declined. I mean, the bookings, TCB declined quarter to quarter and YY. However, you also mentioned that the HBM TCP order has been solid and you received various HBM players and we have seen your revenue contribution from Korea jump to 14% in the second quarter. Does this mean that the logic application, in the logic application, TCP order has been rather weak? We need to wait for the C2W application to introduce. So before that, we won't see any meaningful rebound for TCP for logic applications.

Yeah, Alex, let me try to make it clear for you now. I think first and foremost, for a TCV market driven by AI, I think it's the nature of the business, right? So the customer base in the first place are not very big. Although we have been expanding our customer base, we are talking about 20 to 30 customers, right? Compared to, say, on the Semi side or SMT side, we're talking about hundreds of customers. So you can imagine, you know, with that cost, skill at the mainstream side versus the AP side, the order flow for AP and TCB will be uneven quarter to quarter. So, of course, it takes a longer period, you know, six months, one year, you can see a better trend. So don't read into too much. My suggestion, don't read too much into quarter to quarter variation in terms of order flow for AP. Now, you're talking about whether logic is we need to wait for chip to wafer to come back. Again, it's also related to the fact that, you know, the nature of the business. You can't expect the same customer to continue to place all the quarter after quarter, quarter after quarter. That's not possible. So I think in terms of the logic side, chip to substrate, I mentioned, I think one of the questions earlier, either by Coco or somebody else, that we're in a very strong position, you know, we continue to engage our customers in the future generation of the compound line for chip-to-substrate. And you're right, the chip-to-wafer will come in in a bigger way in 2026. So that will help to add to the order flow level comes 2026.

And also, Katie, I just want to clarify one thing. You were asking about booking, which Robin answered, but then also you were kind of observing from the geographic location of the career, right, that that's actually on revenue. So I just want to make sure that this, again, the career jumped really in first half is because of that bulk order that we were referring to earlier. So just be mindful when you look at the order and the revenue. For AP orders, CCB orders, there is a longer lead time, typical six to nine months. So if you're trying to figure out the timing of those, the quarterly or yearly timing from booking to revenue, please keep that in mind as well.

Yeah, I understood. I'm just wondering, you know, probably the HBM order, you know, you obtained like one or two quarters ago, you know, deliver this quarter. So, I mean, for the past two quarters, logic applications, TCB order was quite weak compared with the HBM application. Okay, so my follow-up question is about your market share for the, you know, HBM customers. Can you give some, you know, probably some color on like your market share with your largest HBM customers or your token market share in the next one to two years? And also, have you shifted the HBM application TCP for Chinese customers?

Alex, in terms of market share for HPM, if you have been following us for a while, we acknowledge that we are not the first mover in terms of HPM. Considering where we come from, from zero base to where we are today, I think we have made a huge improvement in terms of market share for the HPM market. And we are confident that with the advent of HPM4 and going forward, we certainly have a differentiating technology compared to our competitors. I think that will put us in a good state to continue to demand in that particular space. Now, I can't comment in any particular region for TCP customers, but certainly we have been stressing our AP solution, our TCP solution, we have a worldwide customer base. Thank you.

Thank you. Thank you, Alex. May I now ask Catherine to unmute yourself and ask your question, please?

Hi, thank you for taking my question. My first question is regarding the hybrid bonding. How do you see yourself that the key differentiator between your hybrid bonding tools and the leading and the other peers hybrid bonding tools? And how do you think about your market share in the hybrid bonding market? Thanks.

Yeah, for hybrid bonding, We have been saying our LBG2, our generation 2 hybrid boundary, which is going to be shipped sometime, the first two is going to be shipped sometime in Q3, will be very competitive. We have taken note of all the pain points for hybrid bonding over the years, do a lot of channel chatting with our customers, what are the pain points. So we are addressing those pain points one by one. And we believe our Gen2 machine or tool for hybrid bonding will be a very competitive one. So we are confident that going forward, if hybrid bonding picks up in terms of demand, we are ready to play in a more active way in the particular market for hybrid bonding. Now, so as I said, in terms of features, you know, we're very competitive, whether is it in terms of bonding accuracy, in terms of UPH, in terms of footprint, and more importantly, in terms of total cost of ownership, I think Gen 2 is a very competitive tool.

Got it. Thank you. And my second question is regarding your China revenue. So congratulations on your China revenue pickup. Just with the China localization accelerating, are you seeing any shift in the customer preference? And how is your pricing and margin profile in China evolving versus other regions?

Your first question is for localization, acceleration. Yes, certainly. I think that's a trend. A lot of localization happening in China as well. So the volume in China, we have seen, as I said, the momentum in China is pretty good. In the recent quarters, we see the demand coming more and more from the consumer market. It seems that the consumer market is starting to pick up. the EV market in China as well. That's also benefiting both our Semi as well as SMT mainstream. The Chinese market is obviously very, very competitive. But what is really interesting for the Chinese market is the volume, right? So we all know the semiconductor market is also highly dependent on China for the volume. So the Chinese market would give us a volume more than anything else. Back to you, Patrick.

Yeah, so a little bit of follow-up on that. So how do you think about the pricing and margin of your China revenue compared with other regions? And do you think that this pickup in China market is a one-off event, or will it continue probably to the next few quarters? Thank you.

Sometimes it's not exactly comparable. It depends on the configuration that you want, right? So depending on the application as well, you know, say, for example, smartphone requirement is totally different from automotive. So the pricing does vary from more on the application and market rather than the regional pricing from one region to another region. So it all depends also on the application and market. Now, whether is it sustainable? Is this speaker sustainable? We certainly hope so. We did a lot of channel checking with our customer base. They seem to be more optimistic than before. And also looking at the utilization rate across the factories, the momentum seems to be there, inching up, picking up in terms of utilization rate. So I think that's a good sign. Consumption of lead frames seems to be on the rise as well. That's also a good sign that the factories are working well. they're really working the utilization itself in China. So I think that helps to give a little bit more confidence that this time around, the momentum in China may be sustainable.

David, thank you. Thank you, Catherine. May I now ask Simon to unmute yourself?

Okay, thanks a lot. Thank you. Robin, yeah, today, again, the overall guidance for the AP, broadly optimistic, but I want to double-check why you think the NEOTA model is still volatile for the, I mean, the TCV, because foundry, based on the long-term agreement with the U.S. feminist customers, hdm memory also is more annual contract basis and why you think the um your customers uh the tcb order some quarter very strong the other quarter some sequential decline do you see the any signs for the little bit you know downward trend of the ai theme or what do you think the background of the order you know, uneven trend, yeah.

Yeah, Simon, I tried to explain, I think it was someone asked a question, so I tried to explain, when we mention an even order flow, it's a matter of timing, right? So if we break up, you know, by quarter, bound to have volatility, you use the word volatility, so I'm using the same word, uh we use uneven order flow bound to have uh from quarter to quarter but if you take a longer one year versus another year if you look at first half of 2035 we mentioned our ap our tcb have grown half on your idea so so on the half yearly based on a longer term basis you can see a better trend but on a quarterly basis because of the nature of the business as i mentioned earlier if you talk about AI customers, we're talking about 20 or 30 in the world right now, right? You compare to the SMG mainstream and semi-advocate, hundreds of customers. So you see the scale is totally different. So by nature of the business, there will be an even with all the people from quarter to quarter. So don't look at quarter to quarter to discern the trend. Look at a longer period to discern the trend.

Of course. Yeah, yeah. That's why you are TAM guidance not really change the best. You know, maybe opposite way, for example, if I make HVM, if I run the HVM memory business or even co-op packaging business, and then the equipment vendor's leading time, six to nine months, so I think your customers should worry about the six to nine months or sometimes one year, you know, the equipment delivery time. But how the ASMPT can respond to the unexpected wash order if your manufacturing cycle time becomes more like three quarters?

Yeah. I think first of all, once we are trying to cut down internally, make it more efficient in terms of cutting down our lead time, for sure we are working hard on that so that we can respond faster to our customer requirement. And secondly, the other way to mitigate some of this issue is really to have a buffer stop. So because of our deep engagement with all these customers, we roughly know where the demand roadmap is so we can build buffer cells just in case they want the machine very fast you know we are able to supply in a very quick time but limited quantities right so there's always this balance here you know so this is how we can mitigate that situation yeah very clear

And then very quickly, we are very impressed on the AP, the Advanced Taxing Revenue Contribution. So first half, 39%. Last year, 30%. That automatically implies more than 30% growth in AP. But the question is, such a strong growth is more HBM memory-driven or logic or half and half? Any rough idea, though? contribution mix?

Yeah. So Simon, I think first of all, when we talk about AP, it's actually more than TCP because we have a whole range of AP tools, including TCP, although TCP is the major revenue contributor. Okay. Now, within TCP itself, I think going in the longer term, I mentioned before as well, if you look at our TAM, how we arrived at our TAM, of 1B in the longer term, HBM in terms of ranking, in terms of the demand for TCB tools by application, HBM will be the largest, followed by cheaper wafer tools, and then followed by cheaper substrate tools. This trend will play out in the long run, not now. But in the meantime, in the meantime, HPM is still the number one, you know, premium followed by cheaper substrate for our case.

Yeah, yeah. Oh, one thing maybe some investors emailed me. The 500 units so far, would you specify the term since when? Is it until the second quarter of this year on page six of your slide, the 500 units in months?

Yes, since we have been making this information since 2012, I think.

Yeah, 2012. 2012.

2012.

2012.

Okay.

2012 up to second quarter this year?

That's correct. That's correct.

Okay. Okay. All right. One, sorry, last thing. TCV declined quote-unquote. Sorry for the very short-term question, but it's more HVM memory related near-term trend, right? TCV declined quote-unquote. I mean, the bookings declined quote-unquote. That's the more HVM memory related rather than a logic, right?

Luckily, yes, yes.

Yeah. Okay. All clear, sir. Thank you very much, Robin. Thank you.

Thank you, Simon. With that, we now conclude our Q&A session. Before we close the call, I'd like to ask Robin to say a few words. Yeah.

Thank you, Leonard. Okay, thank you all for your questions and really allow me to quickly highlight some of the key takeaways from today's call. So the first, I think first, strong demand continues to be driven by AI tailwinds across AP and increasingly, you know, mainstream solution as well. For the first half of 2025, we had better than expected bookings, something that's probably worth taking away as well. AP continued to grow, with AP revenue contributing significantly to growth revenue in the first half of 2025. This growth was primarily driven by ongoing demand for TCP tools. In the mainstream business, the growth is beginning to benefit from AI tailwinds. AI data center demand has driven Booking's growth for new power management capabilities, and we also experienced Booking growth in China. Finally, we maintain group grants managing about 40% despite foreign exchange trade wins in the first half of 2020. With that, it concludes our call. I will see you all in the next quarter. Thank you very much and take care.