Good morning and good evening, ladies and gentlemen. This is Rommel, the Head of Investor Relations for ASMPT, and I will be the moderator for today's call. On behalf of the group, let me welcome all of you to our third quarter investor conference call, and thank you for your interest and your continued support in the company. Please note that all participants will be on listen-only mode when the management is presenting. We will start the Q&A only after the management has gone through the entire presentation. During the Q&A session, priority will be given to the covering analysts. Let me quickly highlight the disclaimer. Please do note that during this conference call, there may be forward-looking statements with respect to the company's business and financial conditions.
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. For your reference, the investor relations presentation related to our recent results can be downloaded from our website. On today's call, we have our Group CEO, Robin, and our Group CFO, Katie. Robin will begin with a brief discussion and highlight on some of the developments of the group, and then Katie will provide details on the financial performance. This will be followed by an update on the revenue guidance and outlook, and then we will open the floor for Q&A. With that, let me hand the time over to Robin now.
Thank you, Ro. Good morning. Good evening, everyone. It's good to see all of you on our earnings conference call for Q3 2023. Let me start the presentation by providing some of our observations on recent developments in the semiconductor industry and the overall macroeconomic environment before we provide an update about our third quarter performance. For Q3 2023, the group continued to be impacted by an ongoing challenging macroeconomic environment, in particular, tepid consumer spending. Consequently, weak conditions were prevalent in the semiconductor industry, with low electronics demand and conservative CapEx investment. Due to this industry weakness, demand-supply dynamics did not turn favorable during this quarter. As the group navigated these challenging industry conditions, our unique and broad-based portfolio helped to partially mitigate the adverse impact.
This can be seen from our SMT business, which continued to deliver higher revenue than our semi business for a fifth consecutive quarter, even as SMT goes through a normalization phase. Our SMT business has delivered strong performance for over two years and is quite expected that it will undergo an adjustment period as its overall addressable market normalizes. For our semi business, we are witnessing gradual improvement in factory utilizations by the semi customer base, but this has not yet reached optimum levels for strong order flows. As a result, while semi witnessed some sporadic demand, we believe that this was not reflective of a broad-based recovery in the segment. It is interesting to note that on a year-on-year basis, semi bookings have been declining at a gradually slower pace in the recent quarters.
This indicates that bookings are trending in the right direction, and we believe this is a signal that the market could be stabilizing. Looking at our unique and broad-based portfolio across both our Semi and SMT segments, and our deep partnership with major customers on their technology roadmaps, we are well-positioned to capitalize on growth opportunities. In particular, our comprehensive suite of Advanced Packaging solutions continues to witness a growing global demand, largely driven by Generative AI and high-performance computing, or HPC, applications. I will share more about the progress of our Advanced Packaging, or AP, solutions in the next two slides. The increasing demand for such applications give us confidence in the long-term potential of our AP solution suite. We are deeply engaged with key customers to enable a strong growth in Generative AI and HPC by meeting high-precision bonding requirements and stringent total cost of ownership criteria.
The group saw continued demand for AP solutions, with our thermal compression bonding, or TCB, solutions continue to contribute the most to both the group's AP bookings and revenue for this quarter. Let me provide more detail about our TCB solutions. I have highlighted this part in the previous call, but let me repeat that for logic applications, our TCB solutions are enabling both chip-to-wafer and chip-to-substrate processes for major customers. These are critical for the heterogeneous integration and assembly of increasingly sophisticated advanced packages. Thus, logic-related packaging demand continued to drive momentum for our TCB solutions. For the third quarter, orders for the group's TCB solutions or generative AI applications came from a leading foundry and also from OSAT customers. Orders from this leading foundry player were the first batch of TCB orders secured from them, and we expect more order flow as they expand their AP capacity.
For HPC applications, we saw order flow from a leading logic IDM customer. We expect continued TCB order momentum for logic applications as we continue to work closely with major Generative AI and HPC players. Let me also update you about the progress of our TCB solutions for the High Bandwidth Memory or HBM market. We continue to engage multiple memory players to cater to the more demanding packaging requirements for customers' next-generation HBM needs. For these engagements, with complex requirements, we are using our next-generation TCB tools, featuring green ultra-fine pitch capabilities. Next, let me highlight our mass reflow high-precision Die Bonding solutions. Even as the TCB gains momentum, our mass reflow solutions remain relevant for such applications, and we saw demand from AI and HPC-related customers.
We also deepened our engagements with leading foundry, memory, and OSAT customers, and we expect more order flows for mass reflow tools going forward. Let me also give a quick update about Hybrid Bonding solution. After winning our maiden order in the second quarter of this year, we secured another Hybrid Bonding tool order from another customer. This tool will be used for 3D Integration, with delivery expected in the second half of this year. We also continue to engage players in the various end market applications for our Hybrid Bonding solutions. Besides these high-precision bonding solutions, other AP solutions in a comprehensive AP portfolio are obvious beneficiaries from increasing demand for Generative AI products. Let me highlight two of these. First, our high-end SMT placement tool, placement tools continue to have robust demand, mainly from AI-powered server customers.
I'm also pleased to note that our advanced placement tools are gaining traction for chip packaging at the substrate level, and won orders from the leading foundry customer. Lastly, an update on the Photonics and Silicon Photonics solutions. These solutions are able to meet the significantly high bandwidth transfer requirements of AI-powered data centers. We repeat orders for solutions from leading Generative AI customers for the transceiver expansion, and we expect this order momentum to persist as Generative AI needs are fueling the expansion on transceivers for data centers. From this slide, we are trying to highlight some of the key AP solutions benefiting from the growing demand in Generative AI and HPC applications. On the left of the slide is an example of a high-end Generative AI or HPC device that goes into a data center.
As these devices get more complex, such packages will require an increasing number of logic chiplets and new-generation high-bandwidth memory stacking. This will lead to a significant increase in the number of interconnects required to be bonded by high-precision bonding tools. These also come with more stringent technology and cost criteria. With that background, I'm pleased to highlight that we have high-precision bonding solutions that are relevant for the different interconnect requirements for such complex devices. In analyzing high-precision bonding requirements, along with the total cost of ownership criteria, we strongly believe that the majority of these interconnects can be handled by TCB tools. This give us confidence in the long-term potential of our TCB solutions as we continue to work with different Generative AI and HPC-related players, including IDM, foundry, OSAT, and memory players.
You can also see in the slide that Mass Reflow tools remain relevant, with our solutions experiencing continued demand. Our SMT solutions are also able to meet the requirements for such devices in data centers. For these complex packages, our advanced SMT tools help to place integrated passive devices or IPDs on the substrate, both between the passive interposer and the substrate, and between the substrate and the circuit board. On the right side of this slide, we mention our Photonics solutions. Our solutions are required for die placement and lens attach, and can even handle 400G transceivers and beyond. These transceivers are experiencing demand growth with increasingly high-bandwidth transfer requirements to help efficiently process high-speed optical data in data centers...
I hope that with these highlights, you get a clearer picture on where exactly many of our AP solutions are relevant and obvious beneficiaries with the growing demand from generative AI and HPC applications. With those highlights, let me now pass the time over to Katie, our Group CFO, who will talk about our group and segment performance.
Thank you, Robin. Good morning, and good evening, everyone. Let me take you through the financials. This slide covers the group's key financial metrics for the third quarter of 2023. As Robin has highlighted in his opening, the group was impacted by weak industry conditions and a challenging macroeconomic environment. Our unique broad-based portfolio partially shielded us as SMT continued to deliver higher revenue than Semi, even though SMT has entered a normalization phase. Semi remained at a relatively low revenue base in absence of a broad-based recovery. Looking at a quote-unquote comparisons, group revenue decreased by 10.9%, while bookings came down by just 1.8% for the third quarter of 2023. Our Q3 ending backlog decreased by 7.2% sequentially, to about $922 million.
Our gross margin of 34.2% declined quote-unquote by 594 basis points. I will provide a detailed explanation on this decline in the next few slides. Also, let me highlight that while navigating the challenging environment, the group continued to control cost and drive efficiency. This included a targeted headcount reduction in Q3 of a low single-digit percentage of the group's workforce. Since there were restructuring costs incurred, we now use adjusted net profit and adjusted earnings per share metrics. For more information about these non-HKFRS measures, you may refer to the reconciliation section on page 10 of the group's third quarter 2023 results announcement. Group's adjusted net profit for Q3 was HKD 45 million, down 85.3% quote-unquote, and was adversely impacted by lower sales volume and gross margin.
Adjusted earnings per share for the third quarter was HKD 0.11, a decrease of 85.3% quote unquote. The group maintained a strong balance sheet with healthy liquidity position. At the end of Q3, cash and bank deposits were at HKD 4.17 billion, up from HKD 3.77 billion in Q2, while bank borrowing remained stable at HKD 2 billion. Let me now go through more detailed financials and segment performance in the next few slides. For third quarter of 2023, group revenue of $444 million was almost at the midpoint of our previous issued guidance. This was a decline of 10.9% quote unquote, and a 23.8% year-over-year.
Group revenue declined due to prevalent industry weakness, with both Semi and SMT generating lower revenues. For end market applications, revenue contributions from both automotive and industrial continued to contribute to the highest proportion to group revenue. Group bookings were at $378.5 million for third quarter, a decline of 1.8% quarter-on-quarter, and of 18.3% year-on-year. Let me now provide some further insight into our third quarter bookings. The group had an isolated order cancellation in the third quarter for its panel deposition tools under the Semi segment. This cancellation came from a leading high-density substrate manufacturer in response to its slower-than-expected digestion of existing capacity. If we exclude this cancellation, the group's third quarter bookings would have been about 6% higher quarter-on-quarter. For our year-on-year bookings, the decline was mainly due to relatively weaker industry conditions.
Please take note that this isolated cancellation was only for one particular customer, and we do not foresee any material cancellation risk for our panel deposition tools for other customers, as we continue to deliver tools to them. Group gross margin of 34.2% declined by 594 basis points quarter-on-quarter, and by 670 basis points year-on-year. Gross margin declined mainly due to unfavorable product mix, volume effect, and provision for agent inventories. A majority of the gross margin decline came from Semi, and I will go to it in the next slide. The group's operating margin and net profit for third quarter decreased quarter-on-quarter and year-on-year, mainly due to lower sales volume and gross margin. For Q3 2023, the Semi segment revenue was down moderately, quarter-on-quarter by 4.9% to about $201 million.
The IC discrete business unit had stable revenue quarter-on-quarter, with the highest revenue contribution from TCB, followed by molding and the sintering solutions combined for automotive applications. The optoelectronics business unit recorded higher revenue quarter-on-quarter. Revenue growth was mainly driven by higher revenue from wire bonders for conventional displays. The CIS business unit revenue was adversely impacted by continued weakness in the global smartphone market. Third quarter bookings for Semi increased to approximately $169 million, with contributions mainly from AP and automotive applications. This was an increase of 4.4% quarter-on-quarter. Semi bookings would have increased by approximately 22% quarter-on-quarter if we exclude the order cancellation for the panel deposition tools I highlighted in the previous slide. Semi gross margin was 31.9%, mainly due to unfavorable product mix and provision for agent inventories.
The impact on product mix was mainly due to three factors. First, it included the phasing out of an older generation, lower margin product as part of a planned product upgrades. Second, it included relatively higher wire bond sales under the opto business unit. And then lastly, there was a lower volume of high-end smartphone-related products. Our provision for aging inventories, this was down mostly due to the accounting policy on aging, but these inventories are largely related to active products and can be utilized in the future. The SMT segment contributed about $243 million, or about 55% of group revenue for the third quarter of 2023. This was a decline of 15.4% quarter-on-quarter, mainly due to lower revenue from high-end placement and printing tools, but partially offset by revenue growth from SMT advanced packaging tools.
Automotive and industrial applications combined contributed to about half of SMT's revenue. Our SMT business has enjoyed a high level of bookings for over two years, and its addressable market is now in a normalization phase. SMT bookings declined to about $209 million for the third quarter, a decline of 6.4% quarter-on-quarter. Similar to revenue, SMT bookings were also driven mostly by industrial and automotive end markets. Moreover, there was an uptick in demand from AI-related server applications. SMT gross margin decreased to 36% in the third quarter, down by 218 basis points quarter-on-quarter, mainly due to volume effect. Let me now pass the time back to Robin for fourth quarter revenue guidance.
Thank you, Katie. In the short term, weak economy and end market electronics demand will continue to prolong the industry inventory adjustment and constrain the capital spending of our customer base. With these considerations in mind and the seasonality effects, the group expects revenue for the fourth quarter of 2023 to be between $390 million-$460 million. At the midpoint of $425 million, this represents a decline of 23.2% year-on-year and 4.2% quarter-on-quarter. The decline will be mainly due to SMT normalizing, while we expect semi revenue to remain stable. Looking ahead, we continue to remain optimistic about the long-term prospects and growth potential for our group.
This optimism is supported by the long-term structural trends of automotive electrification, smart factories, green infrastructure, 5G, IoT, and high-performance computing, fueled by generative AI growth. To support this increasingly digitally connected world, more organizations are also preparing for a future with increasingly dynamic global supply chains. We firmly believe that these factors will lead to an increase in overall CapEx spend. This concludes our third quarter prepared presentation. Thank you. Now we are ready for Q&A. Let me pass the time to Rommel to facilitate.
Thank you, Robin. Just want to highlight one correction. For our Hybrid Bonding made an order that was received in Q1, first quarter of 2023, not second quarter. For the Q&A session, for asking questions, please either use the raise hand function or type your questions in the chat to ASMPT Q&A. Please ask your questions one by one and limit them to maximum two questions at each time. With that, can I request Donnie to unmute yourself and ask your questions?
Hi. Thank you. Well, can you hear me?
Yes, Donnie, go ahead.
Oh, thank you, sir. Thank you very much for taking my question. I have two questions. The first one is that, may I ask your view, current view on the booking trend by different product lines or applications into fourth quarter 2023? And would like to know a little bit more on the main drivers of TCB in third quarter and fourth quarter, whether it's mainly driven by foundry, memory, or CPU companies, or more evenly distributed in between these main customers. And this is the first, first one. Thank you.
Okay, Donnie, thanks. I think Robin will answer both your questions. The first part of your first question is on the booking trend, and you would want to know how was the booking trend if broken down by application or product side. So let me request Robin on that.
Hi, Donnie. Thanks for the question. Now, in our opening remark, I mentioned that for third quarters, the orders for group TCB solutions for generative AI applications came from a leading foundry, and these are for chip-to-substrate applications. And also, we receive a TCB orders also for generative AI from OSAT customers as well. I wish to highlight that the orders from this leading foundry player, we believe is the first batch of our TCB orders to come from them, and we expect more of this order flow as they continue to expand their AP capacity. Also, for HPC applications, we continue to see order flow from a leading logic IDM customer. And we also expect continue to see order momentum for this logic application, you know, going forward.
Now, as to some Q4 booking color, we continue to see AP being our AP demand coming from Generative AI and HPC applications. For Q4, we believe that, you know, the transceivers demand for data centers will pick up, and we are well positioned for that application because we have a suite of very high precision tools for placing laser diodes and lenses attached for the transceiver application. So part of the AI demand are also coming from the transceivers application area in Q4. TCB will also continue for Generative AI, also for the substrate level.
As I said, we received the first batch from the leading foundry, and we believe that, you know, we should continue to receive more orders in the coming quarters. So I hope I've given you some color for the AP bookings in Q3 and Q4.
Thank you, Robin. And my second question is regarding to... maybe it's more for Katie. So for Semi Solutions gross margin, wondering if you or Katie can quantify the percentage point of the impact from unfavorable product mix and provision for aging inventories. And for the provision of aging inventories, is that more like a one-time issue, or we should expect it will continue to be the case in coming quarters? And what is your estimate on the reasonable gross margin of Semi Solution if the overall semi cycle is back to normal? Thank you.
I think you have three parts of your second question. So, first part, I think I would request Katie to maybe highlight. I don't think she can quantify exactly in terms of percentage breakdown by different, different sizes of the factors, what is impacting the gross margin, but maybe I will let Katie give some color, comment on it, and also, you know, to comment whether it will continue in Q4.
Thanks, Donnie, for the question. So let me spend a few minutes on margin for Semi. As we have always communicated, the margin can be influenced by volume, product mix, or combination of the two, and in this quarter, it, it did play out this way. So, the largest negative impact came from product mix. For example, we had a roadmap to phase out an older generation of wire bonders, and in Q3, we had a significant opportunistic sale for that. For general lighting and other consumer products, there were some sporadic sales. And as you know, Donnie, the mainstream products have relatively lower margins. Lastly, we had relatively higher volume of high-end products serving the structural markets in the previous quarters.
So that's kind of the combination of those is the product mix, which is the largest driver for the margin decline. Now, as to the inventory aging, since we have been in the down cycle for more than a year, as per our accounting policy, we need to provide for some slow-moving inventories due to aging. However, most of these inventories are still active. So to answer your question, whether, you know, it will be repeated or not, since we've been in this downturn, we have subsequently controlled inventory intake, so we do not expect the need for another significant wave of inventory provision. So I hope that answers your question.
So last part of the question, Katie, is, you know, can you give a little bit color on, you know, if we minus off some of these one-offs, what can be like, you know-
Yeah.
Normalized margin for the Semi side?
All right. So, we do expect, like we always say, the volume and the specific product mix will continue to influence the margin level for Semi. But based on our visibility, our, for Q4, our Semi gross margin should be around 40%, mainly due to better product mix of AP and automotives.
Thanks, Katie. Next, can I request Gokul to unmute yourself and ask your questions?
Yeah, hi. Thanks, Rommel, and thanks, management team. So my first question is on the TCB growth outlook. I think, last call, I think, management mentioned that memory seemed like the bigger opportunity within the TCB for generative AI and HP applications.
... is that still the same view, that memory is likely to represent a bigger market and more market opportunity? Or are you seeing more progress on the foundry and the OSAT side? And secondly, could you talk a little bit more about the wins in the HBM area? I think you talk about two customers. Are you already shipping to these customers, or are you just starting this engagement and booking process? I didn't understand where is the progress on the memory memory customers for HBM. That's my first question.
Hi, Gokul. This is Robin. Concerning our view, where the demand for TC is coming, whether it's HBM or logic, if you look at a typical AI package, it be said in 2, 2 year, until 2 year, right? That the potentially going forward, the HBM market could be exciting, you know, because of memory stacking using TCB. For ASMPT at this moment, we see order flow more on the logic side at this point rather than on the HBM side. We did receive some a few tools order, and we have released them to the memory customer. And we are engaging not just one, but more memory customers for our TCB solutions.
We are targeting to use our next generation of TCB tools with better capabilities, you know, for memory, HBM or memory stacking, going forward. Now, what's the second part of Gokul's question?
No, I think he just wants to know whether we are shipping out right now, or is this still in designing and qualification scale?
Yeah. So as I said, we are engaging. I think I already answered part of the question. We are engaging multiple memory players to use our TCB solutions going forward for HBM. Gokul. Back to you, Gokul. Do you have follow-up questions?
Yeah, I have a follow-up question. Thank you very much. So on the memory side for HBM, could you talk a little bit, or maybe also on overall TCB itself, could you talk a little bit about ASMPT's competitive positioning? You have been pretty dominant in the logic IDM space, but there are a lot of new companies that have come up in the memory HBM space as well. So could you talk a little bit about where is your competitive positioning right now, as you see it? And secondly, within TCB, there seems to be a migration to fluxless TCBs starting to happen for higher precision.
Could you talk a little bit about ASMPT's roadmap on fluxless TCBs as well, and how low can you get down in terms of performance, down to maybe 10-micron pitch density levels or even lower than that? What is your roadmap there? And also, just wanted to understand competitive positioning, and if you could also talk a little bit about what is your market share that you expect in TCBs currently, given a lot of these smaller competition emerging in niche applications. Thank you.
Gokul, I think your question is broken down into three. Robin will answer all of them, but the first one will start with memory related, and you want to know our competitive positioning in the market, especially for memory related to our TCB.
Okay. So I think for HBM, the existing POR is mainly on mass reflow. That's what we understand. But increasingly, we see our customer base moving to TCB. So, the potential for a TCB application for HBM is definitely there. As I said earlier, the TCB market for HBM potentially can be significant going forward. But still, at this stage, we see our TCB being adopted more for the logic side for, currently, chip-to-substrate applications. Now, for chip-to-chip to wafer, I think I can follow on with the... What's the second question?
So second, second part of the question is more on fluxless TCB.
Yeah.
Whether we are moving into the same direction, and whether we are working on something similar. In terms of the precision, how low can our next generation tools go, the 10 micron or, you know, the indication?
Yeah, sure. I think we have a solution for, you call it fluxless, but you know the different terminology, but the other terminologies is etch oxide removal, EOR. Now, our solutions are green solutions. They are ultra-fine pitch. You're asking our capabilities for the EOR, we believe we can even for pitch, we can even go down to below 10 micron, right? And for FVC, definitely around 1 micron or more or less. Yeah. Now, for EOR, we believe probably going forward will be needed for the chip-to-wafer applications, because at the chip-to-wafer level, you need higher precision, you need a smaller bond pitch. So, AOR, going forward, the application we see will be more for the chip-to-wafer.
... So last part of the question is, can you comment on the market share of our TCB solution?
I believe we have a very good market share because we have been in, you know, in the TCB market for more than 10 years now. So, coming from that perspective, I think, we have, as I said many times, we have a very good experience handling TCB for leading logic in a player, and now also going into the the leading foundry space as well.
Next, can I request, Sunny? Sunny, you can unmute yourself and ask your questions.
Yeah. Good morning. Could you hear me okay?
Yes, morning. We can hear you.
Thank you very much. So my first question is also to follow up on your progress, same TCB. And so on the logic side, any expectation in terms of when this leading foundry client will, will start to use your TCB for chip-to-wafer? And then for HBM, I guess from technology side, different memory makers seem to be saying different thoughts about using mass reflow or TCB for next generation HBM technology. And so could you share with us what's giving you confidence that TCB should be gaining more opportunities over mass reflow in the coming 2-3 years?
Okay. Thanks, Sunny. Okay, I'll break down your question into two parts. First is more on the progress of our TCB, and what are we expecting in terms of, you know, potential for foundry orders and whether it's going to be for Chip-to-Wafer. I will request Robin to comment on it.
Yeah. I think generally, we can't comment on any specific customer application. But generally, I think we all understand that for mass reflow versus TCB, mass reflow, in terms of pitch, in terms of accuracy, is not as fine as TCB. So as packages increasingly become more advanced, you know, they need to pack more chips closer together to enhance the performance of the AI or the HPC packages. I think increasingly, TCB will be, you know, the solution rather than mass reflow. But currently, you're right. I think we still see mass reflow tools being the POR, you know, for chip-to-wafer. However, as I mentioned earlier, we have already received a first batch of a TCB tools for chip-to-substrate to handle a large compound die.
So I think going forward, we believe the industry will move towards adopting a TCB, you know, for chip-to-wafer application, for the logic application at the AI chip substrate level. Now, for HBM, you are right as well. I think the existing POR is using mass reflow, but as the requirement for higher performance is expected from the HBM, from 8-high to 12-high to 16-high, we see that would demand more precision, and that's why a TCB becomes more relevant as we move into the technology space, you know, as our memory players continue to stack more and more memory die for HBM. Okay.
Sunny, I think, Robin answered both part of your questions. Do you have a follow-up question?
Yeah. So, my first follow-up on HBM side. So when, like, would you have any initial expectations, when would some of your engagement with the key memory makers be converted into the real order win? And then secondly, on hybrid bonder, how's the engagement with your first key customer? And would you be able to provide us a bit more color on your second customer? Thank you.
Okay. So first question is on HBM side. When are we expecting all these engagements, which are ongoing, to convert into order wins?
Yeah. We believe sometime in 2024, but it's exactly difficult to pin down exactly when in 2024, you know, we should see some orders coming in for HBM TCB tools. Second question?
So, second question is on Hybrid Bonding side. We won our maiden order in Q1 recently. In Q3, also announced a second customer-
Yeah.
-for our second tool. So can you provide some more details on these customers or, you know, where is it going?
Yeah. We can't... As I said, we have never provided specific details about customer base, but we are certainly pleased to receive a second order for Hybrid Bonding. And these two are coming from—these two demands coming, again, also from Asia-based customer.
Next, can I request Wan to unmute yourself and ask your questions?
Thanks, Robin, for taking my question, and thank you, everyone, for giving me the time. So essentially, my first question is regarding the SMT business. My first question is, have you seen any shift in discussions, whether it's from the auto or industrial side, just in terms of a bit more cautiousness? And if you could provide, if you have seen any, just in terms of the geographical market, that would be helpful.
Yeah. Okay. I think for auto, we have mentioned also in Q2 that we see overall, I think SMT is normalizing. I think that's quite kind of expected, because if you look at the whole process chain, right? We are in back-end, and then after the packaging, you know, the chips need to be put onto the PCB board. So it's a natural progression from back-end to SMT. Because back-end has been experiencing a prolonged downturn for close to a year, so we expect SMT to feel, you know, that the impact sooner or later. So SMT is definitely normalizing. And for the fact that our SMT, we are also very exposed also to automotive and industrial.
So we see automotive also as a whole normalizing, whereas industrial are still, you know, the momentum is still strong. So as a result, our SMT overall, in terms of bookings and revenues, are starting to normalize since Q2. Okay?
Okay. Then in terms of my second question, just going back over to your HBM side. I mean, there have been different players saying with a high degree of confidence that they'll begin to win share in HBM memory. I'm just curious, as we move from Mass Reflow to TCB, do you see space for, aside from the dominant player, as you migrate from Mass Reflow to TCB, do you see space for two, two or three players in these customers? Or do you think the bulk of the orders would be primarily concentrated in two?
Generally speaking, our customer base tend to dual source, you know? So we see space for, you know, a multiple source of a supplier for even the HBM space. Yeah.
Okay. And lastly, could you just talk about the relative cost of your TCB systems compared to what you currently have for hybrid? And then also just in terms of the time to bond of your TCB systems versus hybrid. Any percentage difference in numbers would be helpful. Thanks.
Yeah. So we can't—we, I mean, this is a competitive reason we can't reveal the pricing. But we mentioned a few times already before, don't just look at the absolute price of a hybrid bond versus a TCB. You also have to take into account the whole ecosystem, you know, for TCB solution versus hybrid bonding. So we mentioned many times TCB at this moment in time, in consideration of the technology requirement, TCB is still coming up tops in terms of total cost of ownership versus hybrid bonding.
But, we also said the hybrid bonding will gain importance as we, as industry move towards more and more finer pitch and requiring much more capability in terms of accuracy. So but at this moment, TCB is still tops in terms of total cost of ownership. Thanks.
Got it. Thank you so much.
Dylan, can I request you next to unmute yourself and ask your questions?
Yes. Good morning, everyone. Thanks for taking my question. So, yeah, my first question is also on TCB, maybe two parts. One is that, I wanna follow up on the comment on 12 high migrating to 16 high, because my understanding is that some of the in the industry people, they're discussing about using hybrid bonding for 16 high, potentially. So what is our view on that? Do we think 12 high is a bigger opportunity for our TCB and HBM or 16 high, and why is that? That's the first part.
And the second part should be, I recall that last quarter, we showed an interesting chart suggesting that next year could be an inflection point for TCB and for us, ASMPT, do we think it's mainly driven by logic customers or memory customers?
You're right. I think for HBM, the industry moved from 8-high, which is, we believe is using Mass Reflow, to 12-high using TCB. Possibly, you know, higher than 12, hybrid bonding could be the solution. Because the higher you stack the memory die, you want the memory die to be as thin as possible, because TCBs in between the die, TCB still need copper pillar, no matter how small is it, right? Whereas for a hybrid bonding, you are actually stacking die to die without copper pillar. So if you want the die to be as thin as possible, when you stack more and more memory die going forward, hybrid bonding may be required. Now, your second question is on the TCB inflection point?
Yep.
Yes, we came out with that, kind of, based on industry research last quarter. We came up with a chart that, potentially TCB could see an inflection point. The TCB addressable market could see an inflection point in 2024, driven by both, HBM as well as, logic applications. For ASMPT, we see at this moment, a logic applications will be, our main driver. For HBM, as I said earlier, we could see some order flow, in 2024. But we already seeing order flow for TCB, for AI generative package, already in Q3.
Yeah, thanks. I have a second question, actually, also a follow-up on TCB. Yeah, like you mentioned, 12-high should be the major targeted market within HBM, and I believe there will be more 12-high variant starting from 2024 for HBM market. So in our view, back to the market share question, but within HBM, in our view, going forward, what will be the market share dynamic? Because you also mentioned that our customer base tend to dual source. So I'm curious about ASMPT's view in terms of market share within HBM.
Yeah. As I mentioned in my opening remark, you know, we are certainly targeting the HBM application for TCB, right? Using our next generation of TCB tool, which are more precise to handle a finer pitch. And, as I said earlier, also that, there's always room for more than one supplier in any application. Okay, thanks.
Next, can I request Nicholas to unmute yourself and ask your questions?
Yes, hello. My apologies, I don't have any question on TCB and HBM. To go back to semi margins, I understand what you said about the wire bonding one-off orders. But when I read the, you know, since you say that the weak margin was driven by the mix, when I read the earnings release, I... It looks like TCB revenues increased, optoelectronics revenue increased, CIS declined. So does that suggest that TCB and optoelectronics margins are lower than average?
Let me request Katie to comment on this.
Nicholas, the TCB gross margin is accretive to the overall company's margin.
Optoelectronics, maybe?
Opto depends on what in the opto, like what we said, mentioned earlier for general lighting, et cetera, right? The opto wire-
Yes. Okay, okay.
That was the issue. Yeah.
Understood. Last quarter, maybe I'm wrong, but I remember that you were expecting SMT to weaken and automobile demand to weaken, and now it seems that, you know, as Robin was saying, SMT normalizing, auto normalizing. So is there no more automobile weakness? Is this what this means?
Let me request Robin to comment on this. So basically, you want to know under the SMT some comments on the automotive side, because, you know, we did talk about it normalizing. So what do you see?
Last quarter, weakening, right?
Yeah, it's normalizing. That's why you see our Q, SMT, both revenue and, bookings, came down, quarter-on-quarter. And, going forward to Q4, we believe SMT, you know, revenue will also continue to, to decline, while, semi revenue kind of stable for Q4. Yeah.
And industrially strong, Robin, right? I'm still, you know, you said that last quarter, I think. I'm still a bit puzzled by this, because from the semiconductor vendor, for example, Texas Instruments last night, they pretty much consistently mentioned weak industrial demand.
Yeah, but we don't serve just one customer. We serve a variety.
Neither do they.
For SMT, for SMT, they're more on the SMT side, right? So we see industrial demand. So the customer base for industrial is quite different from SMT.
Understood. Thank you.
Next, can I request, Laping to unmute yourself and ask your questions?
Yeah. Okay. Thank you. Thanks. My question is, I also have a question of your margin. So you keep— So you mentioned that TCB is above corporate average margin. So, is it possible to separate the impact first from the aging inventory and the change of product mix? And actually, which products are facing margin pressure, as in... This is the first question. Thank you.
Okay. Uh-
Yeah
... I think your question is also on the mark-
Yeah
for different products, right, Laping?
Yes, yes.
So, in the opening remarks, I mentioned for the products, the more the mainstream products, wire bonders, for example, the opto wire bonders for general display, as an example, right? That, that has relatively lower margin. TCB, I just want to confirm, the TCB is above the corporate margin.
So, for where you run business, so for the Semi biz, what should be the normalized margin you are expecting, looking for? Since actually 31%, 32% is quite low if you look at your long history of the operations.
That's correct, and that's why we spent a few minutes to explain the, maybe I'll call it the extraordinary or one-time items that happened in Q3. Usually, we don't guide our margin, as you guys all know, but based on our visibility, as I mentioned, in Q4, we think the semi gross margin should be around 40%.
Okay. Okay. So, the second question is about, I mean, the cycle. So I think your leading foundry customer just last week said that they see some early sign of the recovery. So, we see our Q4 is still down, Q4 is still slightly declined. Do you, when you talk to your customer, do you see any sign of recovery on the consumer side? Since I was quite surprised that auto and industrial already account for 50% of SMT, so it means consumer is really weak these days. Yeah, thank you.
Yes, I'll answer the question. I think you are right. I think on the consumer side, or what we call the high volume demand area, like consumer, which goes in, and also smartphones, for example. So consumer will be things like TVs, for example, you know. So all these have yet to really recover. That's why we said that we don't see any broad-based recovery, because the consumer side is still weak, whereas the automotive, in general, for both Semi and SMT, are still at a pretty good level. Industrial is strong by SMT. So as a whole, automotive and industrial are the two segments that continue to do well, you know, for ASMPT, in addition to advanced packaging. Does that answer your question?
Yeah. So it's related. So every people, since every people we are asking teams, so do you have any vision or view that when you go to the revenue in 2024, so how, how big the TCB or, or advanced packaging should be account for the whole group sales? Yeah.
Yeah.
Especially, yeah.
We look too far. We don't have a crystal ball, but all we can say is that the AP will continue to be a strong driver for ASMPT as a whole. We took some pain in to describe, you know, our optimism about AP solution because we have a comprehensive suite of solutions, you know, not just TCB alone. As I said earlier, our solution for photonics and silicon photonics are used for transceivers packaging. We see healthy demand picking up in that area as well. You need both to work, right? If you look at what goes in your data center, you need an AI chip, right? We have a page just now on our deck. On the left-hand side, we have the AI packaging, right?
Also on the right-hand side, we need transceivers as well. So, and then you also need server boards that go into the data center. So if you picture the data center, we are playing in multiple areas in the data center. First, at the chip, at the substrate level, packaging the... Using TCB for a memory die, for logic die, chip to wafer, chip to substrate, SMT coming in at, you know, placing IPD or integrated passive devices onto the substrate. And then on the right-hand side, we have transceivers packaging using our Photonics and Silicon Photonics tools. And then, not shown in the diagram is our TCB, I mean, our SMT placement tools are placing all these AI chips and all onto a server board, right?
So all this would go into a data center. So this is really driving a demand for AP solutions in a very comprehensive way for ASMPT going forward.
Okay. Thank you very much.
Mm.
We'll take one last round of questions. Simon, can you unmute yourself and ask your questions?
Okay, great. Can you hear me well?
Yes, Simon, go ahead.
Yes. Great. Yeah, great discussion so far, focusing on the AP or TCB. But, let's do simple math. If we use the industry average of the TCB, you know, on the equipment price, say $1 million-$2 million, I wonder how the, you know, the ASMPT can enjoy meaningful revenue growth with TCB. The realistic is, issue is that the equipment price is just so low. So do you agree our view, for example, based on your previous comment, the, your ASMPT's historical shipment unit since, 2012 up to maybe early this year, about 300, unit, right? That means annual shipment, maybe recently, 40-50 unit. And then the seven-digit price means maybe $1 million, et cetera. But, yes, we appreciate the management efforts, the TCB, you know, equipment shipment increase.
But the problem is the price per unit is so low, so how are you gonna make meaningful revenue or profit increase with the TCB bundle then? Thank you.
So Simon, before I pass it to, Robin, a couple of comments. First thing, we don't really disclose our ASP for TCB, for that matter, majority of our, products. Second thing is, we did highlight about, TCB's potential, but we are talking about TCB's potential, you know, as part of AP, not really, not really putting as a percentage of the entire group's one, because you know, our group's portfolio is, pretty comprehensive and it's, quite vast. But having said that, let me request Robin to maybe give you some color in terms of, you know, how our TCB has progressed and, you know, how it can be one of the, you know, one of the pillars of growth going forward.
... Honestly, nothing much to add to Rommel's commentary. I think for TCB, the potential is really on the generative AI packages, right? So if that is expected to take off in a big way, you know, from 2024, 2025 onwards, you can imagine the number of tools that TCB tools that require for both logic, for both HBM can be significant going forward.
Yeah. Yeah, but okay. Sure. But it is true that the, your historical shipment unit volume all together around the 300 unit since, 2012? Make sense.
Roughly there. For ten, for about 10 years, we ship, roughly the count amount. We don't discuss the exact number, but thereabout, yes. Last quarter, we also mentioned that, you know, for the first 10 years, the number of tools that we ship, and in the next 3 years, you know, from 2022 to 2024, we believe, we probably, will see the same demand for TC, for TCB tools in the next 3 years versus the last 10 years.
Yep, that's a great color. But, for now, your TCB equipment is mainly for the logic and the foundry, customers. Just the engagement with the memory makers, right? For now.
Yeah. For now, yes. The logic for AI, generative AI, for HPC, yes. For memory, as I mentioned earlier, we ship a few tools already to a memory maker. Yeah. So we are already in memory. We are already in memory for TCB, but we are looking for more potential going forward.
Great. Yeah. Okay, great. Yeah, you covered all the questions. I really appreciate it. Thank you so much, Robin. Yeah.
Thanks, Simon. With that, I think we'll end the earnings call. Let me thank all of you again for attending today's call, and we hope to see you during the next quarters call. Thank you. Take care.
Thank you.