Okay, I want to thank everyone for joining this next session at the UBS Tech Conference. I'm Randy Abrams, Head of the Taiwan Research, also now covering hardware, and have Sunny Lin covering semiconductors out of Taiwan. We're pleased to have Amkor with us. Kevin Engel, we can congratulate him. He'll be the incoming CEO after a long history at Amkor moving up through the organization, so good appointment. We also have Megan Faust, the CFO. Feel free to add your questions, and we'll try to weave it into the conversation. I'll kick off, I think, for Kevin just coming in to Amkor. Give an overview. How do you see the business, what you're coming into with Amkor? If you take an initial cut and get appointed, any changes?
If you think the business is already running or we could see some fine-tuning around the strategy?
Yeah, thanks. Good morning, everybody. For me, a couple of things. First, really honored, obviously, that the board and the Kim family has the confidence in me to take Amkor on the next phase of our journey. Really exciting opportunity there. When I think about the strategic pillars, which hopefully some of you have heard these before, we kind of have three pillars that we continue to adapt. I'd say these are pillars that we've been modifying over time and will continue to modify moving forward. Kind of step through them a little bit. We think about our first pillar, which is our technology leadership, and that's really to work with our customers to drive the advanced packaging leadership into the market and make sure that we're giving the customers what they need in collaboration with them.
We think about our second pillar, which is our geographical footprint and really the diversity that Amkor has in multiple regions, multiple countries. You can think about that also as customers try to build in resiliency into their supply chain. That enables us to really help them when we think about advanced packaging, specifically where most OSATs today in advanced packaging are in the Taiwan region, maybe a few in China, whereas Amkor has optionalities in Korea where we're continuing to expand and obviously in the future in the U.S. That's a good footprint for our customers. Our last pillar there is focusing on mega trends and markets that we want to grow in and with close collaboration with our customers. Obviously, AI plays into that, and we have a lot of strong engagements there.
I'll try to kind of talk about some of those pillars as we go through the rest of the questions because I think that's important to think about how this transition looks for me. I think those pillars align with our future direction, especially where we are in the AI scale-out. All three of those components play into that market dynamic. I wouldn't expect any significant changes there. I think potentially some fine-tuning on some things, and we'll talk more about that at our investor day, which will be coming up around the middle of next year.
Okay. Yeah, look forward to that. Actually, a follow-up question on AI. That's seeing a lot of interest and activity. Sunny's been raising the CoWoS forecasts, so it feels like a good outlook. Could you talk about how Amkor is positioned, the solutions you have for high-performance computing, whether across GPU, ASIC, some of the CPUs?
Yeah, sure. I think there's a few things that also, again, kind of tie it back to the pillars a little bit. From the geographical footprint perspective, again, customers really want diversity in the supply chain and make sure they have resiliency built in. That's really creating some opportunities in our Korea location where we're expanding in that footprint from a couple of different dimensions. Number one, from a capacity perspective, but also from a building perspective. Obviously in the U.S. in the future. When we think about our platforms, 2.5D, which is something we've had in volume production for quite some time now, that platform we've continued to see increased interest from multiple customers. We've gone from a couple of customers that have been in production in the GPU and other space.
What we've seen over time is that that's grown to over five customers, again, with new products being introduced. I highlight that more from the perspective of we see [legs] in that platform. I think if we go back a year ago, a lot of people thought that or commented that all customers were going to migrate from 2.5D into organic interposer, and they were asking us about what we would do with those assets. Even though the assets are fungible back and forth, what we're seeing is we're still seeing customer interest in that 2.5D platform. I think that's very positive. We look at the HDFO platform. Now we're talking about our, again, the organic RDL type platforms. We've talked about how we've ramped our first volume product this year. We have our second product qualified and beginning ramp.
Over the next several months, we have another several products that are continuing to migrate from qualification into production. We are really excited about that. That gives us a good tailwind for that to continue to grow. When we think about the next generation products, more of the bridge type applications, we are still in development phase with some customers in that area. Again, broad platform. I think when you tie it back into our strategic initiatives and where we are investing, that is definitely a high level of our investment percentages.
Okay. Maybe a quick follow-up to that. How are you seeing 2026 scale-up? When you talk about backfilling the existing 2.5D and then also ramping up high-density fan-out or some of the bridge solutions, how should we look at the AI business into 2026?
Yeah. If I think about the products that are coming in the market, some of our first HDFO products were PC-related products. Again, kind of think of AI in the PC market. The next generation products are getting more in the data center now. I think that's obviously an exciting trend. Going into 2026, we've invested pretty heavily this year to prepare for 2026. We've talked about how that was a little bit of a margin challenge for us this year as we invested heavily. Depreciation increased a little bit. We invested on the people side to make sure we had the right amount of labor resources available that were trained and skilled up.
There is room to grow going into next year just with that type of space that we have already invested for, and then we are already investing for incremental capacity going into 2026 or later in 2026. I feel very confident. If you look at our compute segment, over the past five years, we have been growing at around a 12% CAGR. That is almost double kind of what Amkor total has been growing. I think that is obviously a positive direction. We continue to see that level of growth moving forward.
Yeah. Maybe a few more questions on the compute segment overall for both of you. Compute at this point is roughly 20% of your total sales. Could you remind us what are the key product drivers at this point, and how is that evolving going to 2026?
Yeah. Let's talk about our compute segment in general. Within that, we have the PC space, and then we have data center, a couple of other small things. In the PC space, as I talked about, we're still seeing momentum on some of these AI applications. Typically, maybe in the past, these were a lot of standard flip chip or multi-chip module type applications. We're starting to see some of these products getting launched in more of an HDFO platform. When you think about the data center, like we said, that's continuing to grow. We see a lot of different application spaces there from ASIC players to direct engagements with foundries and others.
You think about the application space, that's a mix of things like you can kind of across the board, we have switches, NPUs or NICs, different CPUs, lots of different application spaces there. I think all those are areas where we're going to continue to see that growth.
Right. We are seeing the 20% of sales from computing overall. Would you be able to quantify how much is driven by advanced packaging and testing?
I don't have that number off the top of my head. I'd say if we look at the compute segment in general, it's a significant portion, and again, that's what's driving the growth. That CAGR is driven by the advanced packaging space.
Got it. With the stronger demand opportunities across maybe also PC and also server, how should we think about the CapEx going to 2026?
The what?
The CapEx. How do you expenditure?
We'll talk about CapEx for 2026 on our earnings call coming up in a few months. I think that'll be the key. As I mentioned, if you go back to this year, we invested heavily to be prepared for early 2026, and we'll continue to invest in that area going through the year.
Okay. If I could follow up about the profitability. You talked about this year you put in place investments, that there was a bit of OpEx. How do you see the fall through? As you grow this high-performance compute, I think you've talked about a 30% fall through of margin. For advanced packaging, is the margin structure different, and is it reaching that potential, or do you still have teething issues to ramp up and it'll take that time to get that leverage?
Sure. Sure. Generally, our financial model would suggest a 30% flow through, as you mentioned, Randy. What we're seeing with the very advanced high-performance computing is this does demand a premium, and there is a higher margin profile for that high-performance computing. We see that today in our Korea facility. As we are bringing on incremental headcount and incremental CapEx, there is some what I would call short-term operational efficiency that we need to gain. As Kevin mentioned, we see that those products are coming online. We have line of sight, and that growth and that scale is going to come in 2026, which will definitely help that short-term margin perspective. That is really what we're planning to do here in Arizona. The margin profile for that very high-performance computing space is accretive, will be accretive to our overall gross margin.
Okay. So it sounds like to clarify next year, by 2026, you're kind of through that learning curve. So you could be accretive to the 30% fall through as we see advanced packaging growth.
Yeah.
Okay. Good. Actually, I think when you were talking about the different expansion, Korea test, that's been an area that you've added capacity. When I look long-term at Amkor, I feel like there's been a strategy to grow test, but then you look at the revenue mix and it's 10-15% of revenue. Do you see with the test investment inflection where it could be growing at a premium to advanced packaging, or does it scale up with advanced packaging? Maybe talk about the test opportunities you see.
Yeah. If I think back to, again, those strategic pillars and how the resilience in the supply chain is bringing opportunities to the Korea location, customers ultimately want test and packaging to be done in the same location. They do not want it to shift back and forth between different regions. As we grow our advanced packaging within the Korea site, customers also want to tie on the test. What we've—and then I'd say there's another opportunity there. Let me not pass up on that opportunity. As we think about some of these custom ASIC type players, they do not really have a significant test infrastructure. Those types of customers are also looking for what is their test strategy, how do they build out that infrastructure. I think that's a good opportunity.
In Korea, as you mentioned, we've been working on expanding the footprint within the existing building, and that space is becoming available this year. We look at another building, which we just broke ground a couple of weeks ago on that facility. That will be more focused on test at scale, and that will be ready going into 2027.
Is there a way to think about the capacity addition? Because you have two buildings, quite a bit of test capacity. Will it be a pretty significant scale-up in your testing?
It'll be significant. I think the other thing you got to think about is we have a significant amount of test in the K5 building today. That would migrate into this new test building so that we can expand the assembly footprint in the existing building. In general, the new building has the ability to also continue to scale and go even bigger. As the test area grows, that'll definitely be an area that we have a lot of room to expand.
On your expansions here in Arizona, what drove the expansion to second phase before you complete the first phase? How should you think about the timeline for your $7 billion CapEx? I think for this year, CapEx is only approaching $1 billion.
Megan, you want to
Sure. Sure. Obviously, there's been a lot of announcements, a lot of expanded investment, especially here in Arizona with respect to the semiconductor supply chain. Our announcement of expanding our investment is following the demand of our customers and our partners. To think about that $7 billion investment, that is going to be a multi-facility, multi-phased approach. We've broken ground on our first phase. We expect that to be completed mid-2027 with production beginning in 2028. The timing for phase two is really going to be a function of demand. We can accelerate that. We can time that as far as when that demand is ready to come online to make sure that we're managing that CapEx.
If we look at TSMC, they have tried to build the fab shell for the third phase, while they are only moving tools for the second phase going to 2026. I wonder for Amkor, would you also consider maybe setting up the fab shell before the demand starts to emerge so that could shrink the lead time for your expansions?
Right. We're evaluating all scenarios. We're looking at parallel pathing. We're looking at serial. Those are all in discussion with our customers today.
Got it. TSMC has also announced two advanced packaging fabs in Arizona. How should you think about the dynamics and your collaborations with TSMC? Should we assume a good part of your capacity will be related to the collaboration with TSMC maybe on substrate or fan-out?
Yeah. The partnership with TSMC is very strong. We're talking to them a lot about, again, how we both support the mutual end customers. I think that's the overall goal. When we think about their packaging facilities coming online, I think that creates more opportunities for Amkor, especially when you think about, again, like you mentioned, on-substrate opportunities, things like that. I do want to make sure that it's clear that the Arizona facility is not dedicated to TSMC. We'll have several different business models. In some cases, TSMC would be the direct customer for Amkor. In other cases, the end customer would be the direct customer directly to Amkor. I think there'll be lots of different dimensions and dynamics related to what kind of customer agreements, what kind of capacity planning we're working through.
Ultimately, them continuing to scale in the U.S., it adds additional value for Amkor.
Got it.
Okay. I think if we look at timing for Arizona, it comes up. I think you talked about middle of 2027, you'll have tools moved in, 2028 it starts up. How should we think about the ramp-up of profitability with the higher cost in the U.S., and it takes time to get scale? Could 2028 we see the type of pricing where that could be a corporate average, accretive, or is there a period we have to ramp it up to scale to get to a profitable model?
Sure. I would say from an order of magnitude, we're anticipating our first phase will be around 10% of our portfolio on a revenue basis. As we move in tools and we ramp up in 2028, we would expect that to probably take two years where we would get to comfortable scale where we would see profitability that would be accretive to margins. When we think about the broader aspect of high-performance computing, as I mentioned before, with that being on the very advanced side, that general portfolio there is also what's going to enable that profitability. We've also are in discussions about how to manage the higher labor costs. This factory is going to be more of a level loading. We're going to ensure that we have level loading agreements.
Given that there's a portion of capacity happening in the U.S., we want to fill that capacity first. We also see this factory being more low mix, high volume in order to manage utilization. Turnkey, as Kevin mentioned, we're going to offer that broad portfolio of services, which also supports profitability. It'll be our highest automated factory worldwide. All of those are going to contribute to the profitability and some of those labor challenges.
Okay. When you talked about 10% of the portfolio, how quick do you see the demand to scale it up? I think you talked two years to break even or be accretive. Could we see the revenue scale up to that 10% of the footprint pretty quickly?
Yeah. The exciting thing about Arizona is the demand. The U.S., the demand is here. It is really a function of how fast can we scale up. It is a bit of a different environment than some of our other, the Vietnam greenfield. That is progressing and even better than expected. As far as the demand, that is not the issue. It is how fast can we bring that scale up?
Okay. You brought up Vietnam. One part of the business, you have quite a bit of system-in-package. Vietnam's been a good site for that business. How do we see SiP, one as a growth area, if you can build that business further? The other part is it's more material content, so a bit of a drag on margin. Is that just the structural margin for the business, or are there ways to improve the profitability of the SiP business?
Can we say again?
Yeah.
I think there's a few things there. If we think about Vietnam specifically, that region we went to because it has some benefits from a cost structure perspective. That ramp that we've been going through, there's a couple of different portfolios. We have our memory products that are ramping, and I think that's been very successful with our end customer. We're in the mode now with them of talking about as the memory market recovers, how do we split between the two different manufacturing sites we have for memory? On the SiP side specifically, again, we've been ramping this year. Those ramps have gone very good. We've now kind of proven to the end customers that we have the quality, the reliability, the yields that you would expect from an advanced facility.
That's creating more momentum for that customer to start introducing additional products going into 2026. In addition to that, we have another customer qualified already ramping in production, and then a couple more customers in the qualification phase just for SiP. That is an interesting dynamic in that if we go back to this year, again, we have a very large factory, not as much scale to really get the profitability we would have liked. If we project that forward, now we have the labor force beginning to be more trained and more skilled. We'll get more operational efficiency from that. We have much higher volumes from customer dynamics. Overall, that profitability perspective, we think we'll see some benefits as we continue to scale in a lower-cost region over time.
As these products migrate from some products migrating from our Korea facility into Vietnam, that frees up additional space in Korea to put in more of our advanced package portfolio.
It sounds like we're back to growth. I mean, you recovered that, I think, a key socket with a key customer this year. I mean, you see continued opportunity to grow SiP with additional products, additional customers?
Yeah. I mean, I don't know that I would kind of envision that as really accelerated growth. Again, if we think about where are we really focused and what are our fastest-growing markets, the computing segment is definitely where we see the most growth, and that's where we're very highly focused.
Okay. Great.
Maybe a question for short term. You have guided down quarter for Q4. Any upside, downside that you're seeing across smartphone, PC, industrial, and auto? With a down quarter in Q4, should we expect a better start going to Q1?
Okay. Let's talk about Q4 first. Q4, we're guiding about an 8% down from Q3. That's going back to more kind of average levels of seasonality. We saw similar levels in 2022, which was kind of before we started seeing the downturn in the market. If we look at last year as an example, that reduction from Q3 to Q4 is around 12%. Definitely a better seasonality for Q4 this year. If we look at Q1, obviously, we're not ready to guide there. I would say from a forecast perspective, we feel optimistic about how Q1 is performing. Obviously, we'll provide that details in our earnings call coming up.
Sure. Any view on the strengths or weakness across applications?
Yeah. If I kind of think about Q4, for communications, from a very strong Q3, we saw a little bit deeper declines in Q4 than normal, but really off of a high. For automotive and industrial, I'd say pretty much as expected. We continue to see a little more strength in the mainstream automotive products. Very slow, but at least in a recovery mode. Advanced is, again, doing pretty well on the automotive side. When we think about computing, we did see a little bit of a decline in Q4 as we went through some product mix changes. I think the demand is there going into next year. We feel very positive about that. The consumer space, kind of similar from some of the product launches, a little bit of declines as expected.
In the standard communication or consumer market, we're seeing a little more strength than typical.
Got it. So now looking at 2026, maybe full year, how should we think about some of the tailwinds, headwinds for the sector, and whether we should think about maybe Amkor outperforming or growing in line with the sector?
Yeah. So how do I think about '26? Obviously, we're not going to give a lot of color on that yet. I'd say, in general, what we see on the communication side, I think market projections are to grow low single digits from a unit perspective. We feel our position within Android as well as iOS is pretty solid. So we feel comfortable with those market dynamics. When we look in automotive and industrial, or focus on automotive first. In automotive, on the mainstream side, again, seeing that different customer dynamics, but overall, kind of a slow, slow growth out of that trough. I think positive there. Then on the advanced side, so think ADAS, in-car infotainment, things like that. Definitely seeing continued momentum there. Some of that is in IDM space, but also in fabless companies. So positive momentum in that space.
When we think about compute, obviously, we've talked about that a lot already. We expect strong compute market, mostly driven by the AI data center space, but also some opportunities in PCs related to transitions to ARM, things like that. In consumer, that'll be strongly tied to product launches. In general, 2026 is feeling positive.
Yeah. Maybe if I could squeeze in one more question. Have you started to sense any concerns from especially consumer clients on the rising bound cost because of DRAM and NAND, and therefore potential impact on [NDMA] if they have to raise the pricing?
I would say from our customers, we're not hearing too much about that yet. Maybe just a little bit too early for that question.
Okay. I want to ask how it translates to profitability. Margins, when I look back in the late before 2020, you could get closer to 20% gross margin. With the rise of SiP, it's been mid-teens. How do you think about that margin outlook? It sounds like compute will be the strongest growth driver, and you've gotten through some of the learning curve. Do you see margins starting to creep back up? How do you see the leverage in the model?
Sure. Our margins did come down, I would say, starting in 2023, which was at the beginning of the semiconductor cycle. I would say Amkor has demonstrated tremendous resilience during that cycle. In fact, a lot of our investment decisions were strategic investment decisions that do have some short-term impact on profitability, but are intentional in order to set ourselves up for future growth, which is what we're experiencing now. It is really going to be three factors that we see that are going to drive that profitability back up and beyond where we've been. We've talked a lot about Vietnam. As we now build scale, we have what I would call operational excellence has been demonstrated, so we can pivot to operational efficiency. We've got our top 10 customers all engaged and looking at what products that we're going to bring up in Vietnam.
We see that that margin drag that has been throughout 2025 is really going to lift as we move forward. The second area is the very advanced high-performance computing. We've made some very significant investments, not only in R&D, CapEx, but our people in order to ensure that we have successful ramps and as that scales going forward at a higher growth rate. The last is the recovery of mainstream. While SiP has become a higher proportion of our revenue, that's really a function of how deep the cycle has been for our more mainstream business. As we see that start to gradually recover, we'll definitely see an impact on our profitability moving forward. The timing of how all those kind of comes into play, and then adding that Arizona facility on top, we definitely see expansion in margins.
Okay. How should we think about CapEx? Does it start to move to a much higher level where you're bringing on Arizona? If you could talk about the implication for cash flow and ability to increase payout while you go through this expansion phase.
Sure. We have not yet guided 2026 CapEx. We will do that in February. We need a little more line of sight in how this Arizona build is going to progress. You may expect to see some temporary increase in CapEx as we put in this very strategic investment here in Arizona. There may be fluctuations in free cash flow. Amkor has exhibited very disciplined and prudent capital allocation and will continue to do so. We would expect to continue to grow our dividend modestly as well.
Okay. Great. I think with that, we have to wrap up. I want to thank everyone for coming out. Also, thanks to Kevin. Congrats on the new role. Also to Megan. Yeah. Thank you.
All right. Thank you, everybody.
Thanks.