ASM Pacific Technology Ltd

HKEX:522

Earnings Call Transcript 2023-Q2

[Romil Singh, executive]

Good morning and good evening, ladies and gentlemen. My name is Romil, the group's Head of Investor Relations, and I will be the moderator for today's call. On behalf of ASMPT Limited, let me welcome all of you to the Group's Second Quarter and First Half 2023 Investor Conference Call. I would like to thank you all for your continued support and interest in the company. [Operator Instructions] Let me quickly go through 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 Mr. Robin Ng, the Group Chief Executive Officer, and Ms. Katie Xu, the Group Chief Financial Officer. Robin will begin with a brief discussion and Group's key highlights, and then Katie will provide details on the financials and segmental performance. This will be followed by an update on the guidance and outlook, and then we will open the floor for Q&A.

Without further ado, let me hand the time over to Robin now.

[Cher Ng, executive]

Thank you, Rom. Good morning and good evening, everyone. It is a pleasure to have you all on the earnings conference call for the second quarter and the first half of 2023. First, let me give some highlights on recent developments in the semiconductor industry and the overall macro environment before we provide an update about our second quarter and first half performance.

For the first half of 2023, the group was impacted by the continued weak conditions prevailing in the semiconductor industry. This was due to conservative consumer spending and CapEx investment and ongoing industry supply chain inventory digestion. Looking at the performance of our 2 segments. They offer a sense of how we are positioned as a business and for the future.

For our SEMI business, recovery is taking longer than anticipated, expect utilization of the SEMI customer base has been gradually improving, but has not yet reached optimal levels. In contrast, our SMT business continued to perform resiliently, and the second quarter revenue exceeded that of SEMI for the fourth consecutive quarter. Taken as a whole, this demonstrates how our unique and broad-based portfolio is helping to partially mitigate adverse impact from the current semiconductor down cycle. The combination of this unique and broad-based portfolio and a deep partnership with major customers on their technology roadmap positions us well to capitalize on growth opportunities.

In particular, our Advanced Packaging Solution stands to benefit significantly from the strong growth that has been seen in generating AI and High-Performance Computing, which I will share more about next.

Basically, for Advanced Packaging or AP, we are confident that our solutions are well positioned to meet the growing needs of generative AI and HPC applications. We are engaged in deep collaborations with key customers to enable a strong growth in generating AI to meet high precision volume requirements and stringent total cost of ownership criteria.

In terms of performance, our AP solutions generated above USD 195 million in revenue, representing 19% of group revenue for the first half of 2023, with Thermo-Compression Bonding or TCB making up the highest proportion of our AP revenue. Let me elaborate a little bit more on the TCB solutions, which are in a commanding position to address crucial logic and memory packaging bottlenecks in generative AI.

For logic, our TCB solutions are enabling Chip-to-Wafer and Chip-to-Substrate processes for major customers that are critical for heterogenous integration and assembly of increasingly sophisticated AI computing packages. We see promising TCB order flow for logic from our foundry and OSAT customers. In particular, demand from our foundry customer base is growing due to the urgent need for expansion of the AP capacity.

A popular question we have been getting from analysts and investors is our presence in High Bandwidth Memory or HBM. Simply put, our TCB solutions are able to fulfill the demand in packaging deployments of next-generation HBM. We strongly believe that as generative AI proliferates, customers will increasingly migrate to this advanced HBM packages to meet ever increasing storage and processing needs, and we are well positioned to capitalize on the trend.

We have won repeat orders for HBM, and we continue our deep engagement with multiple memory players. Based on these opportunities, we are confident on our TCB order flow for both logic and memory in the second half of 2023. Besides TCB, our mass reflow high precision die-bonding solutions are also benefiting from generative AI, with continued flow from top-tier global customers. For hybrid bonding, after securing our maiden order, as we mentioned in the last earnings call, we continue our engagement with key customers for qualification in various end market applications, including memory.

Generative AI's increasing demand is not just relevant for high precision bonding solution, but also benefits other tools in the group's portfolio. Let me quickly highlight this. In Silicon Photonics, we have market-leading solutions with high placement accuracy that are relevant for devices such as optical transceivers and photonic engines. We received repeat orders for Silicon Photonics tool to support a key customer's transceiver expansion plans to meet high bandwidth transfer requirements, and we expect more such orders in the second half of 2023.

As generative AI and HPC evolve, they will require increasingly complex chip architectures, which benefits our laser single-agent solutions and panel electrochemical deposition tools. In particular, our laser singulation solutions have seen preliminary engagement with global IDMs to have built the next generations of tools.

On the SMT side of our business, we are seeing traction in the server business that is being driven by AI applications. Here, our SMT basement tools provide flexibility for customers in terms of handling larger servable rates and sizes.

This slide helps [indiscernible] on how TCB technology as a whole and ASMPT, in particular stand to benefit from the growth of generating AI and HPC applications. In terms of drivers, generative AI packages will require an increasing number of logic chiplets and new generation high-bandwidth memory. This leads to a significant increase in the number of interconnects required to be bonded especially for memory. This will also come with more stringent technology and cost criteria.

If you recall, during the last quarter's earnings call, I highlighted an example of high-end HPC device that showed number of interconnects handled by TCB versus hybrid bonding. We estimate that around 90% of interconnects in complex generative AI and HPC packages will require TCB and our TCB solutions are capable of handling a variety of interconnectors.

Based on high precision requirements, balanced with the total cost of ownership, we strongly believe that TCB is a key enabler for generative AI and HPC, and there will be an accelerated adoption of TCB. The graph in the middle depicts how we see the TCB equipment market developing with an inflection point coming in the next year or so. We are in a commanding position to benefit from this accelerated TCB adoption based on our unique capabilities, as you can see on the right-hand side of the slide.

Our TCB solution has the best-in-class die placement accuracy of below 1 micron. They are also capable of handling tin die, ultra-fine precision bonding requirements of below 50 microns and multi-die format of sizes up to 100 x 100 mm.

With this anticipated market potential and TCP capabilities, we are excited about the growth prospects in this particular area. Our Automotive end market applications continue their robust contribution to the group. The Automotive end market contributed the highest proportion of group revenue at approximately USD 230 million or 23% of revenue for the first half of 2023. This contribution spans across the group's mainstream solutions, particularly SMT placement tools, molding tools and die bonders.

Our Automotive solutions have contributed strongly to our overall performance in the last 2 years and this sector has begun to normalize. Notwithstanding demand from EV players and for Silicon Carbide related applications remains robust. Moreover, our Automotive solutions continue to bag design wins that will eventually translate into high-volume manufacturing demand. We believe we have a strong foundation for future growth in the Automotive market due to our role as a key partner in the technology road maps on major customers.

With those highlights, let me now hand over the time to Katie, our Group CFO, who will talk about the group and segment performance.

[Yifan Xu, executive]

Thank you, Robin. Good morning and good evening, everyone. This slide covers the group's key financial metrics for the first half of 2023.

As highlighted earlier, we continue to be impacted by weak industry conditions. As our SEMI business experienced a brunt of impact from the ongoing semiconductor down cycle, our SMT business stayed relatively resilient and was able to partially offset the impact for the group. Looking at half-on-half comparisons. Group revenue decreased by 12.1%, while bookings came down slightly by 2.7% for the first half of 2023. Our backlog decreased by 13.4% in the last 6 months to about USD 993 million. This backlog helps to support group's performance during this ongoing industry down cycle.

Gross margin was 40.3%, a decline of 86 basis points half-on-half. Group net profit for the first half was HKD 623 million, down 29.5% half-on-half and was adversely impacted by lower sales volume. Earnings per share for the first half came in at HKD 1.52, a decrease of 29.3% half-on-half.

Let me take you through more detailed financials and segment performance in the next few slides. For first half of 2023, group revenue was close to USD 1 billion. This was a decline of 25.3% year-on-year and 12.1% half-on-half, mostly due to a decrease in SEMI revenue, while SMT revenue remained stable. Group bookings were at USD 838 million, declining 43.9% year-on-year due to a high base effect. Bookings decreased slightly by 2.7% half-on-half due to a drop in SMT bookings, while SEMI bookings grew from a low base.

Contributions from Advanced Packaging, Automotive and Industrial end markets combined accounted for approximately 57% of group bookings in the first half. Group gross margin of 40.3% declined slightly by 89 basis points year-on-year and by 86 basis points half-on-half. This was partly due to a segment mix, as SMT contributed about 59% to group revenue in the first half. Group operating margin of 10.9% declined by 804 basis points year-on-year and by 318 basis points half-on-half. It was adversely impacted by lower sales volume.

For the second quarter of 2023, Group revenue of USD 497 million was higher than the midpoint of guidance. This was a decline of 25.0% year-on-year due to high base effect and roughly flat quarter-on-quarter. Group bookings were at about USD 384 million, a decline of 35.1% year-on-year due to a high base effect and a decline of 14.9% quarter-on-quarter due to the ongoing industry down cycle.

Group gross margin of 40.1% was a decrease of 158 basis points year-on-year and 33 basis points quarter-on-quarter. Margin decline is mainly due to SEMI, partially offset by S&T. Group operating margin was 9.9%, a decrease of 892 basis points year-on-year and 197 basis points quarter-on-quarter. Year-on-year decline was due to lower sales volume.

For the second quarter of 2023, the SEMI segment revenue grew 7.4% quarter-on-quarter to about USD 211 million from a low base. The IC/Discrete business unit had stable revenue quarter-on-quarter with the highest revenue contribution from TCB. The business unit also had some increase in contribution from its mainstream tools. The Optoelectronics business unit recorded higher revenue quarter-on-quarter. Revenue growth was mainly driven by wire bonders for conventional displays and high-end Silicon Photonics applications also grew. The CIS business unit revenue continued to remain relatively low due to ongoing weakness in global smartphone market.

With the ongoing semiconductor down cycle, second quarter bookings for SEMI declined to approximately USD 162 million. This was a decline of 40.5% year-on-year and 15.5% quarter-on-quarter. SEMI gross margin was 42.7%, a decline of 201 basis points year-on-year due to volume effect. Gross margin declined 243 basis points quarter-on-quarter, partially due to a higher mix of wire bonders.

The SMT segment continued to deliver relatively stable revenue for the second quarter of 2023, contributing about USD 286 million or about 58% of total revenue. This was an increase of 5.3% year-on-year with a decline of 5.5% quarter-on-quarter. SMT's revenue is mainly from Industrial and Automotive applications. These 2 combined made up almost half of SMT's revenue with demand mostly from Europe. Our SMT business has enjoyed a high level of bookings for more than 2 years and has entered a normalization phase.

SMT bookings declined to about USD 223 million for the second quarter, a drop of 30.6% year-on-year and 14.5% quarter-on-quarter. Similar to revenue, SMT bookings were also driven mostly by Industrial and Automotive end markets. SMT gross margin increased to 38.2% in the second quarter. Gross margin increased 85 basis points year-on-year and 81 basis points quarter-on-quarter, mainly due to a favorable product mix.

This slide highlights the management's best estimates of revenue breakdown by end market applications for first half of 2023, highlighting the extent of our broad-based portfolio. The Automotive market continued to have the highest contribution to group revenue at about 23%, mostly from mainstream solutions across both SMT and SEMI. The Industrial market continues robust contribution to group revenue at about 18%, mostly coming from SMT. Consumer, communication and computers or the CCC markets remain weak due to market sentiment.

We serve a diverse global customer base, which includes IDMs, OSATs, fabless, foundries, high-density substrate manufacturers, EMS players and more. Having this wide range of customers has helped us maintain a low level of customer concentration risk. For first half of 2023, our top 5 customers accounted for approximately 20% of revenue. This slide shows the half yearly revenue contribution by different geographies over 3 consecutive periods.

Looking at the graph, contribution from China declined year-on-year for first half of 2023, dropping from 44% to 30%. Revenue from Europe grew from 15% to 30% year-on-year and Americas grew from 9% to 19% year-on-year.

The group remains committed to enhancing shareholder value and returning to shareholders. The existing dividend policy is to maintain payouts of about 50% of group's profit on an annual basis. For first half of 2023, the Board of ASMPT has declared an interim dividend of HKD 0.61 per share, down 53.1% year-on-year and that is in line with 64.1% year-on-year decline in net profit.

This now concludes the financial section. Let me pass the time back to Robin for Q3 revenue guidance.

[Cher Ng, executive]

Thank you, Katie. Near-term visibility continues to be limited due to uncertainty in the macroeconomic environment, marked by persistent inflation, tepid consumer sentiment and ongoing inventory digestion. This makes it more difficult to estimate the timing and the pace of industry recovery.

With this consideration and due to SMT normalizing, the group expects revenue for the third quarter of 2023 to be between USD 410 million to USD 480 million. At midpoint of USD 445 million, this is a decline of 23.4% year-on-year and 10.5% quarter-on-quarter. However, we remain bullish on the longer-term outlook for the group. We have a key competitive advantage as a major supplier and technology partner across many applications and solutions area, while our continued investment in R&D across industry cycles enables us to remain at the forefront of technology development.

Our optimism also stands from the enduring long-term structural trends of Automotive electrification, smart factories, green infrastructure, 5G, IoT, and HPC fueled by AI generative growth.

This slide summarizes our unique broad-based portfolio across Advanced Packaging tools to mainstream and applicative tools that have certain tailwinds for longer-term growth. Before I conclude, I want to highlight a key point that we generally update addressable market details for Advanced Packaging and Automotive markets only once a year and disclose this during our full year results announcement. For the Advanced Packaging addressable market announced during our full year 2022 results, we did not factor in the full potential for the group coming from a strong growth in generative AI and HPC.

This concludes the presentation. Thank you, and we are now ready for Q&A. Let me pass it to Romil to facilitate.

[Romil Singh, executive]

Thanks, Robin and Katie for the presentation. [Operator Instructions] With that, can I request Gokul to unmute yourself and ask your questions.

[Gokul Hariharan, analyst]

My first question is on TCB. Could you give us a little bit more detail on ASMPT's position -- market position in TCB? Our understanding is that ASMPT probably has the highest market share in TCB. Also, historically, ASMPT has been very IDM focused in the TCB presence in -- within the logic space. Could you talk a little bit about the progress that you're having in penetrating the foundry fabless and OSAT ecosystem so far? Is the foundry fabless ecosystem going to be bigger than the IDM within your TCB mix? And lastly, maybe also within the memory TCB market, where this ASMPT's positioning currently stand, given historically, what we understand, there's a lot of the smaller other Asian players actually have a meaningful presence in the memory -- the HBM TCB market? That's my first question.

[Romil Singh, executive]

Thanks, Gokul. So okay, I will break it down into 2 segments. First, I will request Robin to talk more from TCB perspective in terms of our market positioning for both logic and high bandwidth memory or the memory side. And noted on your point that we have been -- for the Logic side, especially, we have been quite focused on IDMs. And then how has the scenario evolved in terms of handling more of, say, foundries, OSAT, fabless. So let Robin answer this. After that, we will come back to your second question.

[Cher Ng, executive]

Thank you, Gokul. Indeed, I think for ASMPT, we have been in the TCB business for more than 10 years now. And I think over the last 10 years or so, the concentration of the TCB business has always been on the IDM. So you also asked about our market share. I think last year, we made the remark that we have installed close to 300 sets of TCB. So we believe we have the largest stock base of TCB in the market. I think your second question is about whether on the foundry...

[Romil Singh, executive]

How have we evolved from serving IDMs towards more OSAT, foundry, fabless.

[Cher Ng, executive]

I think certainly, with the event of the generative AI and HPC, we see a lot of potential coming from the foundry side of our customer base. The -- on the -- the other question about memory, right?

[Romil Singh, executive]

Yes.

[Cher Ng, executive]

Yes. I think besides engaging the foundry customer base in terms of Chip-to-Substrate, Chip-to-Wafer kind of applications, we are also involved with a key -- with a leading HBM memory device maker for packaging the advance 24-gigabyte memory. So TCB is also involved in the memory space from that perspective as well. What else is your question, can you tell? Gokul, did I answer all your questions already, Gokul? Or you're getting...

[Gokul Hariharan, analyst]

Yes. Maybe if you could give us a little bit more detail on the foundry OSAT side. Are you now kind of becoming a major player there as well? Or you're kind of still -- kind of like a #2 player, #3 player in that market for TCB, given -- and I think maybe also historically, foundry OSAT were not very keen to adopt TCB because of cost considerations. Could you talk a little bit about -- I think you showed this chart of inflection of demand, could you talk a little bit more in detail about what are you seeing for TCB adoption? Is it mainly only for the gen AI applications? Or are you also seeing a more wider TCB adoption for some of the other HPC applications in the foundry OSAT space as well?

[Cher Ng, executive]

Thanks, Gokul. I think we see using both. Actually, HPC, we had one slide in the last quarter already, even before this generative AI frenzy started after our conference call last quarter. So we already had one slide saying that our TCB is well positioned for a lot of bonding requirements for HPC packages Chip-to-Wafer and subsequently Chip-to-Substrate. And with the advent of this generative AI, we see potentially more demand for TCB technology use in generative AI packages for sure.

[Gokul Hariharan, analyst]

Got it. Romil, can I move to the next question?

[Romil Singh, executive]

Please go ahead.

[Gokul Hariharan, analyst]

So next question more on the 3Q outlook. I think clearly, the industry seems to be still in the downturn. Could you talk a little bit more on what is the bookings outlook into Q3? And how do you see that trending for Advanced Packaging, the conventional wire bonding and maybe for SMT, for each of these segments, how do you see the bookings outlook?

[Cher Ng, executive]

Sure. I think for Q3, look, we don't really give a solid guidance as usual, we only give guidance for revenue, but I can certainly give you some color or some perspective. We expect Q3 bookings will probably decline about 10% Q-on-Q compared to Q2 this year. Most of the decline we expect are coming from SMT because as I said in our opening remarks, we see SMT starting to go into a normalization phase after 2 -- after more than 2 years of strong growth in terms of bookings.

So I think it's healthy. I would say it's mainly driven by Automotive. Automotive has also starting to stabilize and going to the normalization phase. I believe this is a healthy development because Automotive has been on a high for SMT for 2 years. So it's a healthy development that we start to sort of normalize going forward. Now so taking into consideration the overall climate in terms of semiconductor downturn and SMT normalizing, we are expecting our booking to be around about 10% decline Q-on-Q.

Now we believe all those started to normalize. So we still remain probably highest contribution segment overall in terms of end market applications for our group into Q3. For AP, we expect it to be kind of steady as we mentioned many times before. AP bookings and [indiscernible] tend to be lumpy because of the concentration of few customer base compared to mainstream has much larger customer base. So AP for Q3 booking, we expect it to be kind of stable compared to Q2.

[Romil Singh, executive]

Thanks, Gokul. Next, can I request Wern to unmute yourself and ask your questions.

[Wern Juan Chng, analyst]

This is Wern Chng from HSBC. My first question is on TCB as well. Just regarding your Chip-to-Wafer to your foundry customers, it seems like you're getting a bit more traction there. My question is, are you selling that Chip-to-Wafer -- these Chip-to-Wafer tools as a package in combination with tools for a larger within the same application?

[Cher Ng, executive]

I do not really understand what we you mean by larger bond feature. But let me answer your first question first. Yes, certainly, I think heterogeneous integration, one of the key step for heterogeneous integration is really Chip-to-Wafer application. And we believe that with the current configuration in the marketplace, 90% of those interconnects for Chip-to-Wafer and Chip-to-Substrate will require TCB. So that's why we are excited about our TCB being in a good position to benefit from this trend. Maybe you would like to repeat your second part of the question?

[Romil Singh, executive]

Yes. Wern Juan, maybe you want to rephrase your question about packaging in the larger bump pitches?

[Wern Juan Chng, analyst]

Yes. Sure. Are you selling your Chip-to-Substrate tools in combination with your Chip-to-Wafer to the foundry customer?

[Cher Ng, executive]

Certainly, certainly, yes. Actually, our initial demand for many years has been for cheaper substrate. But with the advent of more complicated packaging requirement like heterogeneous integration. So TCBs are also being used for Chip-to-Wafer application.

[Wern Juan Chng, analyst]

Okay. And the Chip-to-Substrate tools, these are based on past programs, say, about 1 to 1.5 years back, right? Just to clarify.

[Cher Ng, executive]

Yes, last year, we won a big order.

[Wern Juan Chng, analyst]

For foundry, sorry, for foundry.

[Cher Ng, executive]

Yes, yes, yes. So I think that we started off with Chip-to-Substrate. And last year, when we won the big order is really going for Chip-to-Wafer.

[Wern Juan Chng, analyst]

Got it. And then my follow-up question, my second question is regarding just -- I just wanted to hear your thoughts given that 2020 and 2021, you obviously had a lot of orders from the Chinese market. It seems to me that even if there's a pickup in recovery from that market, just in terms of the sheer amount of orders, that has to be digested. I was wondering, it seems to be there will be a lag. So I just wanted to hear your thoughts, even if the Chinese market recovers, how long will it take for that huge orders in 2020 and 2021 to be digested?

[Romil Singh, executive]

So I think Juan, you are talking more from a perspective of inventory digestion mostly from the China side. So even if the China side shows a certain level of recovery, what is Robin's view in terms of potential order flow to come towards our side.

[Cher Ng, executive]

Juan, you're right. I think 2021, 2022 was a -- both years were very high year for midstream product. When I talk about midstream products mainly the die bonders and wire bonders. Now because of the very high demand during that point in time, I think the industry will take time to digest the capacity. However, having said that, we see in Q2, we see some sporadic interest in our wire bonds -- wire bonders basically. And we also highlighted an MD&A that in the announcement that we actually had some demand coming from the conventional display for LED, for example, they're using wire bonders for that purpose.

So yes, I think is encouraging that there's some sporadic interest from this customer base despite the prolonged downturn situation. It shows that this customer base has some confidence to start investing in CapEx. But having said that, don't take this as a signal that is going to be a broad-based recovery, especially from China, because most of these LED tools are coming from China, but don't take this as a signal. Basically the basis for sporadic interest driven by specific situation by our customer base. We believe that some of them have a subsidy backing for some of these investments. That's why they went and hit with these purchases for the wire bonders.

[Romil Singh, executive]

Next, can I request Dylan to unmute yourself and ask your questions.

[Dylan Liu, analyst]

Yes, sure. I also would like to follow up on our TCB kind of strength. So as mentioned, we are now penetrating our foundry customer and now we are also expanding our customer base and the OSAT. I'm wondering what would be our strength for TCB for us to penetrate new customers? Maybe it would be nice if we can compare the throughput position and price with our peers. And also speaking of price, I'm wondering what will be the margin profile for our TCB solutions, because I also noticed that our gross margin is contracting. I'm wondering if our TCB solution is margin accretive or dilutive to our SEMI solution kind of gross margin profile.

[Romil Singh, executive]

Dylan, maybe I'll repeat the first part of the question. So you want to talk about our TCB strength, especially how we have penetrated into more, say, customer base, including foundry OSAT. And you'd want to know more like a competitive analysis for price than throughput and some of these other measures. I will request here that we don't really want to comment on our competitors or what's there in the market, but we can definitely give an overview and talk more from our own strengths or our uniqueness in the TCB solutions. So for that part, I will request Robin to give you an update. For the second part on the GM or some color on the TCB numbers, for that, I will request Katie.

[Cher Ng, executive]

Thanks, Romil. Now Dylan, these are just to correct you. So our TCB penetration is not just into the foundry customer base or set but also very importantly, memory as well, right, high bandwidth memory. So just like in the announcement, we said that we are also expecting the order flow for TCB coming in the second half of 2023 for both logic as well as memory. Now you talk about our strength. I think I mentioned earlier when Gokul asked. We have been in TCB for more than a decade now, right? We have accumulated a lot of learning experience in terms of processes and help to improve the performance of the TCB tools.

Having the largest in-stock base would provide a very useful and very valuable platform for continued growth in this area. When customers need to ramp up to the next level or when they need to upgrade to the next generation of tools, I think they will go for supplier like ASM, which has been providing TCB solution for a long, long time. Now in one of the slides, we also gave some -- an idea of our capabilities into TCB technology, right? Our TCB solution can achieve probably the best-in-class accuracy of below 1 micron. We are capable of handling tin die, a very fine pitch, we call ultrafine pitch bonding of less than 15 microns. And also our ability to handle multi-type format of sizes up to 100 by 100 mm.

Now that's important for Chip-to-Substrate application. As you can imagine, as we put more logic chiplets and memory die. Let me give you an example, I think increasingly, the market is moving to -- because of AI requirement, they could be up to 6 to 8 logic chiplets. And this can be surrounded by 6 to 8 high-bandwidth memory. So you can imagine at that level, at the first level of packaging, the die, the compound die in a pretty large stacks, right? So our Chip-to-Substrate are able to bond the compound die onto a substrate.

So I think these are some of the capabilities, which can distinguish us from our competition.

[Romil Singh, executive]

So Katie, the next part of the question is whether the TCB side is margin accretive or dilutive towards the SEMI side? So maybe you can take that question?

[Yifan Xu, executive]

Yes. So Dylan, thanks for the question. I think your question has 2 layers. One is the contracting of margin in the quarter and then the second is the TCB margin. So I just want to clarify the contraction of our margin is due to low volume, which has been impacting utilization absorption and also due to product mix, as Robin mentioned, that we had taken in some wide bonders for consumer end markets. So it is not due to TCB in the second quarter. Now I want to comment on pricing or specific margin for a very specific transaction with our customers on TCB. But what I can say is overall TCB as an AP solution is accretive to our margin.

[Romil Singh, executive]

Thanks, Katie. Dylan, do you have a follow-up question?

[Dylan Liu, analyst]

Yes. My follow-up question will also be on the margin trend because when we discuss the third quarter bookings kind of outlook, I know that's not a guidance, that's more of a color that the company provides. But based on what we can see right now, most of the weakness of next quarter's bookings coming from SMT. So I wonder, in the very near term, if the third quarter gross margin will expand because of the product mix as well. And also, I noticed our OpEx has been quite stable despite the top line has been declining. So I would love to get a sense on how we think about the OpEx trend either into the second half or longer term or R&D kind of -- longer term kind of a plan for OpEx.

[Yifan Xu, executive]

Yes, Dylan, so let me take this. So I think there are few questions in your 1 question here. Let me talk about the margins first. You're correct. We don't quantitatively guide margin, but I do have some commentaries. The group has been very consistent in saying that our margin depends on volume, product mix and segment mix. For Q3, as we guided, volume will be down, so that pressure continues. But product mix, as we look into Q3, if we continue to experience more demand from mainstream products, gross margin may be negatively impacted due to product mix.

On the segment mix, you mentioned about booking for SMT, but there is a time delay between booking and billing. So as we go into Q3, SMT most likely will continue to be a higher contributor to the group revenue, which means it will -- the segment mix will be under pressure as well.

Now coming on to the next part of the question, which is OpEx. You're correct that the OpEx has been stable. There are 2 layers of OpEx I want to address. One is overall, the group is very committed to long-term growth and profitability, which means we'll continue to invest in the right technology for the future, especially on the right R&D projects, like -- in technologies like Advanced Packaging. Second, the group has been investing in process improvement tools, for example, our ERP system and our global people system. So that investment will continue.

Having said that, just like previous quarters, we have been consistently assessing and rolling out a range of cost measures. Very similar to previous quarters, we have been reducing the flex workers, as an example, very sensible hiring, tightening our T&E travel, et cetera, right? So all these cost measures will continue on.

[Romil Singh, executive]

Next, can I request Sunny to ask your questions?

[Sunny Lin, analyst]

So my first question is also to follow up on the TCB ramp. And so you mentioned that you expect TCB to reach about 90% of market share at some point. So I wonder if there is a time line that you expect TCB to account for 90% of the new tools for HBM and 2.5D package? And the second part of my first question is, if you look into the ramp by HBM and by Logic, when would you expect -- I mean, which kind of customers or would you expect to ramp a bit more meaningfully into 2024 and 2025?

[Romil Singh, executive]

Okay. Sunny, I think I have to correct you at 1 stage. We did not say that TCB will reach 90% of market share. We were just giving an example of HPC device where it's showing different interconnects and how TCB and hybrid bonding are playing into that. And based on that particular example, we have highlighted that we estimate and when you look at the number of interconnects and all these generative AI or HPC packages, we estimate that roughly 90% of those interconnects are handled by TCB compared to hybrid bonding. But that is not reflective of the market demand or the market share.

So maybe for the first part of your question, I will request Robin to just talk about the ramp we are expecting in the overall market -- equipment market of TCB. And the second part of the question is that looking at HBM and Logic, what kind of customers are we seeing, which are ramping up and will require more of this TCB.

[Cher Ng, executive]

Now Sunny, the customer base, as we always see for AP is kind of small compared to the mainstream product. So the users suspect is leading foundry all sets, right? We see all sets are also moving into using more TCB solutions also for -- we believe also for generative AI or HPC requirements. And of course, IDMs, right? And increasingly, because of the storage and the processing needs for memory -- memory makers are also very much part of the supply chain also for generative AI. I mentioned earlier, right, for a technical HPC or generative AI packaging, you have chiplet, you have logic chiplets surrounded by HBM, right? So memory increasingly is also an important enabler for generative AI. I hope I answered your question, Sunny?

[Sunny Lin, analyst]

Well, so thank you for the color. I just wonder if directionally management will be able to guide us through the pace of the ramp by logic and by memory. Both markets could be substantial, but I think, investors will be interested in knowing whether logic will ramp up faster or HBM will ramp up faster for the next 2, 3 years?

[Cher Ng, executive]

No. I think both are required for HPC and generative AI. But if you are looking from the interconnect, basically the number of interconnects required from memory and logic, I think if you look at the advance memory space, the interconnect may be higher than the logic space. So from that perspective, the requirement for a TCB for memory could be coming at a faster rate compared to the logic.

[Sunny Lin, analyst]

Got it. My second question is also on TCB. And so number one, how should we think about the combined TCB orders for 2024 and 2025? I understand memory and foundry are showing upside, but your traditional large customer has placed a big order last year. So will there be some slowdown by your traditional large customer? And so combined, how should we think about the TCB revenue into 2024 and 2025? And also wonder, how should we think about the mix for TCB? For foundry and memory, will they be using mostly the next-generation TCB tools or the current generation?

[Romil Singh, executive]

Okay. There are a couple of parts of the question, Sunny. Maybe on the first part, I will request Robin. But do take note of that we don't really give that kind of longer-term guidance. And for our overall AP, we don't really break it down in terms of numbers when it comes to individual solutions like TCB. But maybe I will request Robin to give more color what he looked -- how is he looking at the demand for TCB and from the market side for '24-'25, especially looking at the mix of the customer base.

[Cher Ng, executive]

Now Sunny, before we came out with the chart that we believe or we forecast, the market will go for TCB in an accelerated fashion in the years to come. We also base -- our internal model assumption is also based on indication of the growth prospect of generative AI. I think lately, there have been a lot of market data on how strong this generative AI will grow in the years to come. So our modeling is also based on that, and also based on the fact that we have engagement with leading customers in this space. So these are all -- our assumptions are based on these kind of indicators.

Of course, we do have to make some assumption, the number of interconnects that will be required by TCB going forward. Now that's why we came up with the term chart for TCB. We are excited about it that the term for TCB will grow in an accelerated fashion, maybe the inflection point could be 2024 or 2025. The exact timing is hard to predict, but the growth prospect for TCB is there because of this generative AI growth.

Now basing on our own experience alone, as I say we are building TCB for more than a decade, right, it took us 10 years to reach a certain level of TCB demand. But just for -- since starting 2022, we believe that -- starting from 2022 to about 2024, we reached the same level of demand for TCB that we -- it took us 10 -- in the initial 10 years to reach from 2012 to 2011 -- 2012 to 2021. So you can imagine even from a own perspective, we are seeing [ several adoption on ] TCB in the last couple of years.

[Romil Singh, executive]

Okay. Thanks, Robin. I think the next part of the question is more from the demand mix between foundry and same memory. This is more from a perspective of the current generation of tools and the requirement for the next generation of tools. So your views on that, Robin.

[Cher Ng, executive]

Now we believe the customer base will continue to push the boundary of TCB towards more and more ultra fine pitch bonding. Now I mentioned earlier that the TCB is capable of less than 15-micron bond pitch -- fine pitch bonding. That is very tight in terms of requirement already, right? So it's really moving towards the hybrid bonding space. Hybrid bonding we're talking about pitch of 10 micron and below. But TCB, we are working with a customer base that are pushing the boundary of TCB to go down to less than 50 micron pitch. So I think in short, there's a lot of leg room for TCB to grow.

[Romil Singh, executive]

Next, can I request Nicolas to unmute yourself and ask your questions?

[Nicolas Guy Gabriel Baratte, analyst]

Looking at your revenues, the SEMI solutions revenue has been declining, let's say, 40%, 45% year-on-year for 4 quarters. Probably given your 3Q guidance, 1 more quarter of minus 40%, right? And the opposite SMT was, let's call it, stable, right? Do you think it's the opposite for the next coming quarters that SMT decline will improve while -- sorry, SEMI solution revenue decline will improve, will be less negative, while SMT will start shrinking more. Is that a reasonable conclusion of what you're saying?

[Cher Ng, executive]

Negative, from our perspective. For SEMI bookings, you're right. It has been -- we have been experiencing a lower level of SEMI bookings for a number of quarters already. So historically, that has been a low number, primarily also because China -- China and the Chinese market has to recover in terms of demand. So -- and also because ASMPT's, our business -- traditionally, our SEMI business is weighted towards small Chinese market. So the Chinese markets are not recovering, the SEMI -- our SEMI business will also be impacted. But as I said, having said that, we have been floating around this low level SEMI booking for a number of quarters already. So another question is about...

[Romil Singh, executive]

SMT as well. But Robin, I think Nicolas wants a view more from a revenue perspective as well, and how we're looking at Q3.

[Cher Ng, executive]

I think for -- for second half. For revenue typically it follows from the bookings, right? So typically it takes about a quarter kind of timeframe before we deliver the booking demand. Now for second half, I think a similar picture, right? So -- when we look at the second half, as we said, SMT will start to normalize. I think that's healthy. And for SEMI, it's difficult to really predict the bottom. But so the negative is that the SMT is normalizing. However, the positive is that we see the business -- the booking coming from generative AI, HPC flow in the second half of 2023.

[Nicolas Guy Gabriel Baratte, analyst]

By normalize, Robin, so whether you're talking about SMT or automobile when you're saying after 2 years of strong growth, you are expecting revenue to normalize, I suppose that means decline, right? So you think Automobile SMT spending was very high for the past 2 years as the result of, for example, European customers' investment in EV or another reason. And why would that decline?

[Cher Ng, executive]

Automotive has been high for 8 quarters maybe or 2 years already. It's a market dynamic. So you don't expect any particular segment will continue to grow and grow without taking a cohort. So as automotive will have to normalize [ somewhere ], I think we see that for SMT.

[Nicolas Guy Gabriel Baratte, analyst]

Understood. Given your percentage of revenue from advanced packaging, if I'm calculating right, which is not guaranteed, but if I get it right, it looks like in the first half of '23, your AP revenue actually declined year-on-year and half-on-half.

[Cher Ng, executive]

Yes. Indeed, as I said...

[Nicolas Guy Gabriel Baratte, analyst]

And declined by quite a lot.

[Cher Ng, executive]

I think overall, the market has turned down. The semiconductor market has turned down. So AP is part of the -- [ luckily ] coming from the semiconductor market. So we don't -- I think it's kind of normal to expect AP also.

[Nicolas Guy Gabriel Baratte, analyst]

I don't know. If you are listening to the largest foundry out there, everything is declining except precisely HPC demand and especially GPU and AI, right? So which component of AP is declining in that case?

[Yifan Xu, executive]

Nicolas, I think you're talking about the first half and -- within AP rather than the TCB that we talked a whole about today on the call, there's also other AP applications, for example, panel -- on the panel side in terms of revenue and the booking. We do experience some weakness due to the substrate inventory digestion.

[Nicolas Guy Gabriel Baratte, analyst]

So inventory digestion, I understand. I'm just trying to reconcile how TCB has become the largest revenue contributor within AP, right? While at the same time, there's something big enough in AP that is declining so much that the total AP revenue are going down.

[Cher Ng, executive]

So as Katie mentioned, last year, we have pretty good demand for our panel ECD tools for high-density ABF substrate. So after 2 years of strong growth in that area, we see that area -- we see that particular segment is going through inventory taxation. So I think the part of the decline in the AP demand is coming from that area.

[Nicolas Guy Gabriel Baratte, analyst]

Understood. Do you have a CapEx number for this year and the tax number?

[Cher Ng, executive]

We don't have a CapEx number on hand, but for tax...

[Yifan Xu, executive]

If you're talking about ETR, we've been looking at 20 -- about the high 20s, I think, that's it. You probably noticed similar to previous quarters because of the high contribution from SMT whose entities are in relatively higher tax jurisdictions.

[Nicolas Guy Gabriel Baratte, analyst]

Understood. If I may ask last question, I promise it's the last one. You have a very sharp increase in revenue from Europe and Americas in the first half of the year, continuing since the first half of '22. Anything to highlight there specially to explain that very large increase in revenue?

[Cher Ng, executive]

They are coming mainly from the Automotive and Industrial market for our SMB business. As you are probably aware, our SMT -- our position in that space is very significant.

[Nicolas Guy Gabriel Baratte, analyst]

Or is their Advanced Packaging in America is also going up?

[Cher Ng, executive]

Yes, certainly. I think for Advanced Packaging, it's coming more on the SEMI side. So America will be also Advanced Packaging. But for Europe is typically coming from the SMT for Automotive industry.

Just on the TCB, just what you mentioned, right, you're kind of surprised why our TCB revenue was so high. Remember, we had a big order last year. So the shipment has taken place this year, and we are anticipating to ship out most of the requirement or the demand for that order by the end of this year.

[Nicolas Guy Gabriel Baratte, analyst]

So in the second half of the year, basically.

[Cher Ng, executive]

No, we have shipped already in the first half, but the shipment will continue into the second half of 2023.

[Nicolas Guy Gabriel Baratte, analyst]

So in that case, is it possible that we've seen SEMI revenue, everything declines, but AP increase in the second half?

[Cher Ng, executive]

We give you a little bit more color in the next quarter. Okay?

[Nicolas Guy Gabriel Baratte, analyst]

Yes. I will be waiting for that.

[Romil Singh, executive]

With that, I think we'll just take 1 last round of questions. Can Randy please unmute yourself and ask your questions?

[Unknown Analyst]

First question following up from Nicolas on the regional split. Curious as China it's been quite depressed. As that comes back, do you -- or should we factor in some dilution? Historically, it was lower margin on the volume orders. So just the margin impact, where do you see the U.S., Europe sustaining at a higher level?

[Yifan Xu, executive]

I guess, Randy, I'll take that on the margin. I touched on it a little bit earlier. Actually, when China recovers, especially consumer end markets covered, it will have adverse impact to our gross margin due to the mainstream product mix.

[Unknown Analyst]

Okay. Are you -- actually, it's maybe a question as you look initial view, a lot of writing off second half. As you take an initial view or talk to customers on the outlook into next year, I think in semis, do you see potential we go back to finally digesting a lot of the capacity and getting to capacity by -- in semis? And then correspondingly SMT, with this high base, does this look like a short-term digestion or the same factor has been strong in the last couple of years may continue to weigh. So I guess just an initial view when you talk to customers on the 2 segments.

[Cher Ng, executive]

Randy, I don't think we can really predict when the [ update ] will come. I think everybody is looking for something answer in the space. So we also take reference from both market research, right, for -- in particular for the SEMI side, our equipment space, market resource is predicting a full in terms of market size this year compared to last year. However, they are predicting a growth in 2024 onwards. So we also have to take reference from the kind of market data.

[Unknown Analyst]

Okay. SMT, do you see an overhang now? Or do you think it looks like a short-term digestion?

[Cher Ng, executive]

I think for SMT, we're still -- because Automotive or [indiscernible], but I don't think it will come down a lot because Automotive overall is still a growth segment for the semiconductor industry. And SMT or SMT is really strong in a particular area.

[Unknown Analyst]

Okay. Great. The last question I wanted to ask for hybrid bonding. I think a lot of that focused TCB. Could you talk a little more of the feedback on hybrid bonding with the first order and then the time line when you see the cut in or potential ramp that tools rather to the 3D IC stacks or how far away for high-bandwidth memory to actually use hybrid bonding?

[Cher Ng, executive]

I think we said many times that the industry always favor a solution that is most effective, right? So we strongly believe that at this point, at this juncture of the development in terms of HPC -- TCB probably has no play in those packages. That's why in the diagram that we have shown last quarter, we believe 90% of those interconnect can be [ appended ] by TCB in a very cost-effective manner, yes.

[Unknown Analyst]

Okay. And [ superior ] tool probably don't expect much ramp than next couple of years. Like it's -- I mean primary focus TCB, but for the early 3D engagements, do you see much? Or is it still going to be a couple of years to where I think about the hybrid binding ramp?

[Cher Ng, executive]

Well for us, it is probably a couple of years down because the development broadband has to factor in those really evolving dynamics in terms of -- for example, in terms of the accuracy, in terms of higher yield into the throughput. And we believe that our next-generation -- [ I think, has to be ] or should be able to intercept. We have a key customer base, a road map for HBM.

[Unknown Analyst]

Okay. And the timing for that next generation. What is the timing for the next generation?

[Cher Ng, executive]

Probably a couple of years from now.

[Romil Singh, executive]

And with that, we will 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 quarter's call. Thank you. Take care.

[Cher Ng, executive]

Thank you.