Okay, I want to thank everyone for coming out for this session. Pleased to have Amkor. And with us, we have Megan Faust, the CFO, and also Kevin Engel, EVP of Business Units. Randy Abrams heads up Taiwan Research. Also, Sunny Lin heads up our semiconductors out of Asia. The format will be mostly a fireside chat. I think a lot of things we can talk about with Amkor really seeing a big pickup in activity and interest in the company. It has always been a strong player in back-end packaging and test, but also now really at the center of some of the industry trends.
It caught the momentum on 5G, automotive, and now with AI. So I think just to kick off, get everyone on the same page of just how the outlook is going now. If you could recap how third quarter went, which was pretty good results. Fourth quarter, I think you did guide a bit more than seasonal for decline. So maybe talk about just on the near term some of the dynamics you're seeing and, as a setup, how it leads you into 2025.
Okay, thanks, Randy. Yeah, Q3, we were in line with guidance. We had a very strong quarter, increase sequentially of 27%. And that is traditionally our strongest quarter, really around the communication cycle and the significant launches of premium tier smartphones. We also had a very high scale launch of a consumer wearable IoT product. So both of those are what contributed to that strength, which required our advanced SiP technology. So we had record advanced SiP in Q3. So with that, we also had an 80% bottom line EPS expansion in Q3, given that flow through on that significant increase. Setting up into Q4, our guide is down 11%.
That is more than seasonal. Typically, we would see more of a flat Q4. That was driven off of the communication cycle being softer than seasonal for Q4. As we look out into 2025, structurally, we see everything intact. Things are progressing slowly but surely, so we are anticipating, given how we're diversified off all those markets, growth in each of those, albeit the pace and the speed of those will moderate and we can get into more of the different end market dynamics through the conversation.
Yeah, so maybe if I could follow up on 2025. So if you look at the major applications, what kind of dynamics and setups are you expecting?
Sure. Why don't we kick off with computing, Kevin, you want to talk about?
Computing for us obviously has been a very focused segment. If you think about obviously 2.5D, which is similar to TSMC's CoWoS-S, that's an area where we've been ramping production significantly. From 2023 to 2024, we tripled our capacity and we expect to quadruple our revenue this year compared to last year. That's an area we're going to continue to invest going into next year at a similar level and for two reasons. Number one, to increase the capacity for the 2.5D, but also to prepare ourselves for the next generation technologies like SWIFT, Amkor SWIFT technology, which is similar to CoWoS-R, that's our RDL interposer, as well as our S-Connect technology, which is similar to TSMC's CoWoS-L, which is the silicon bridge technology. Those are areas that we're really focused on our spending and making sure that we're preparing ourselves for those ramps.
So that's obviously thinking a little bit about the supply chain in general for that segment. People want diversity from a supply chain perspective. Today, it's very concentrated in Taiwan. Our offering is in Korea. So that's a little bit different and that offers some customers some advantages. And then we think even beyond 2025, the U.S. is also going to play a segment in that advanced packaging.
Then maybe moving to some of the other markets, we would expect to see growth in communications as well, albeit it was a bit softer coming out of 2024. Some headwinds possibly in the first half. We're really looking for that AI boost where we might start to see refresh cycles coming in. Depending upon how that plays out, that could be a larger contributor in 2025, if not for sure in 2026. That would be unit growth. We also see potential content growth with complexities required to support those applications, both in the apps processor as well as memory that would require advanced packaging from Amkor. Just rounding out consumer, we've got our high volume IoT product. That'll continue into 2025, taper down. Those life cycles aren't as predictable as communications.
And then automotive is definitely slower to recover, but we believe for Amkor we had that trough in Q2 and we're optimistic to see a slow gradual recovery in automotive during 2025.
Okay. Oh, yeah.
Well, so maybe just one question on tariff. I know maybe still a bit early, but have you started to see supply chain acting differently in terms of some preview or even looking into production shift?
Yeah, great question. So we are not yet seeing any sort of pre-build or unique customer behavior. Given that Amkor's broad geographic footprint is not concentrated in Taiwan, we do see customers taking advantage of that. But that has been happening over time and isn't necessarily accelerating or picking up in this time. And just to add with the recent announcement yesterday on some of the export controls, we don't see any material impact to Amkor's business. And given our communications with our customers, we don't yet see any material impact with respect to the new restrictions announced yesterday.
Okay, good. I'll ask, just because you've gone through AI and there's a lot of attention on it. If you could think about 2025, you had the quadruple in revenue this year. How do you see it broadening out in 2025? Just for GPU, it is moving to Blackwell and now CoWoS-L, so you have kind of a shift in that architecture, so how do you see addressing that opportunity? And then even the long tail for existing CoWoS, like the packaging you're doing.
Yeah, so the way I view it is, sure, if you look at NVIDIA specifically, obviously their newer products are not on the 2.5D or CoWoS-S type structure. What we're seeing in that space though is there's still legs for some of the existing products, and then we're going to be adding incremental capacity along with their customers that are also looking for ways to increase in the 2.5D space, so I think that segment in general has legs. If we go back over the course of this year and last year, there were several different supply chain disruptions. In some cases, silicon interposers were constrained, and that was having customers start looking at RDL structures more rapidly because they were worried about silicon interposer supply. If we look at where the market is today, there's additional supply chains for silicon interposers.
So some customers are even pivoting back to silicon interposer while others are going full steam with RDL interposers. So we see legs in both areas. And for us, we think they're both going to be growth segments. And the other nice piece about it is the assets we're purchasing are basically fungible back and forth between these two technologies. So if one accelerates, we can just pivot those assets over to the other area.
Okay. And a follow-up on that. How do you see the broadening out? You talked about the core NVIDIA, but just the diversity of your customer base, ability to address ASICs. And if we can think mid to long term, there's talk more about panel level packaging if you're starting to put effort into that area.
So a few things. So I think within the data center space, obviously you have the GPUs and then you have other spaces. You have routers, you have switches, others that are started going in this more advanced packaging. So that's definitely happening. We also see it starting to migrate into some of the PC space with the CPUs, GPUs, and how that's going to proliferate as we continue to see more adoption of AI and the need of increasing compute power. So that's an area where we see advanced packaging and then even into automotive and some of the ADAS applications. So we definitely see that long term, this is almost this whole kind of chiplet architecture is kind of an extension of Moore's Law to some extent.
Yeah, well, so if we look into the coming few years, there's more discussions on the package size becoming even larger, and so maybe it would be great if you could walk us through on your technology migrations for panel and also the silicon bridge solutions.
Okay. Yeah, so we come back to panel first, and then we'll kind of go there. So panel, obviously most of the advanced packaging we do is on a wafer format today. So it's sort of 300 millimeter. Like you said, Sunny, as the packages get bigger, you get less panels or less packages per wafer. So to get economies of scale, the industry wants to migrate to panels. So now you're going to 600 by 600 or larger panel structures so you can get basically efficiency of that process. So we have a panel line in Portugal today, which is focused. It's an R&D NPI line, working with several customers on bringing those products into market. Today, they're more focused on lower density applications. And then our R&D team is working on how to bridge that gap into these higher density applications, finer line of spacing, more metal layers.
I think that transition will happen over time and offer that scale and flexibility. As far as where we move forward with that panel structure in Portugal or start looking at Asia supply chains or potentially even the U.S., we need to see how that develops over time. Coming back to Sunny's technology question, again, we see 2.5D is going to continue and packages will get a little more complex. CoWoS-R or Amkor's SWIFT technology, that type of structure, basically we're going through qualifications now and they'll be ramping early next year. As we go into the bridge type technologies, those qualifications will be going through the course of next year. We would expect meaningful revenue in 2026 timeframe.
Sounds good. Well, so I mean, addressable market is going pretty rapidly, but competitions could also be picking up. And if you look at your key competitor, ASE in Taiwan, they have been very aggressively raising CapEx for a few times through 2024. And seems like recently they also made some breakthrough in CoWoS-L with a key customer. So just from your point of view, what are some of your key advantages here and how do you think about the competitive landscape into 2025?
Yeah, so yeah, I won't talk specifically about our primary competitor. I would say in general, obviously both of us, all the tier one suppliers in the SoC space are very focused on that market segment, making sure that we're all moving aggressively from a capital spending perspective, from an R&D perspective, and then just getting the collaboration with the customers. So I think that's all there. When we think about our competitive position, I'd say a couple of things. Number one, our leadership on 2.5D has enabled us to really form these tight collaborations with the customers. And we're working very closely with them on what's their roadmap, what generation is coming next. So I think that's all in place. And then again, our geographical diversity is very important.
While advanced packaging is only offered really by three players, you can argue maybe a couple more, but the majority is with three suppliers, and the concentration in Taiwan is very high, and while all of us love Taiwan, people do want diversity and optionality just for resiliency, so we do offer that option in Korea and we're one if not the only OSAT that offers something outside of Taiwan and China.
Okay. Actually, I'll just position into it. There's some nervousness. We've been trying to catch up to demand for so long. Sunny's model has CoWoS rising, just TSMC alone from 40,000 to 80,000 by the end of 2025. How do you see? Because I mean, you ran through some of the qualification to move into CoWoS-R, CoWoS-L. But do you see a risk over capacity that everyone investing to catch up? And so as you get qualified, the demand doesn't necessarily there or it becomes more captive than expected at TSMC?
Yeah, so I don't see that short term. The way I would view it is, yes, TSMC is increasing quickly. Amkor, ASE, we're all moving forward to increasing capacity. NVIDIA demand will not take all that capacity. But I think as we talked about, if you view it more from the perspective of this is an extension of Moore's Law and more and more applications are coming in, more customers are getting engaged, whether it's in the HPC space, data center in general, computing. It's kind of proliferating down and we expect that to continue.
And then again, if we got in that scenario, which we don't see as a likely scenario today or in the short term, we still can pivot those assets to other technologies and other packages. As an example, a lot of the 2.5D assets can be pivoted over Flip Chip BGA, which again, that's an automotive, other computing segments, a lot of areas, and then all the wafer-based assets, again, can be pivoted over to a lot of different package technologies as well, so we don't see a major risk there today.
Okay. Just maybe the two follow-ups. One, the fungibility. So when you invest in that advanced capacity, if there's a way to think about how much of the CapEx is actually fungible and can transfer over to what you mentioned, Flip Chip. And then I think earlier you mentioned PC and GPU. If you could talk where the opportunity was, because traditionally you have captive, companies like Intel did a lot of the packaging. What areas do you see on both of those to capture opportunity?
Yeah, so from a percentage perspective, I don't want to give a number because I know you guys will hold me to it, but it's a very high percentage. I mean, over 75%. So I think that's a reasonable number. When we think about the PC space, a couple of things. Number one, the migration to Arm is an important component. So you talked about a lot of it. Integrated supply chain is more on the x86 platform. So Arm definitely, you start seeing a lot more fabless companies involved, which is outsourcing the assembly.
And that's where we're participating, whether on the Arm architecture or in the Microsoft ecosystem on Windows platform. So I think that's all positive. And then again, so there's the Arm piece where we participate heavily. And then what is AI going to do to that? Is AI going to continue a refresh cycle? That's going to create a dynamic to where you need more computing power, more memory in the laptops. That's an area where we see advanced packaging playing a key role.
Just one quick follow-up. I mean, is that still traditional Flip Chip? Because you've mentioned advanced packaging. Should we think like PC, some of those would start to adopt CoWoS or some other type of packaging?
We're seeing that, yes.
Okay. What timeframe? I mean, how quickly do you see the evolution?
I'd say probably 2026, maybe small in 2025, late 2025.
Yeah.
Sure. Maybe switching gears to regionalization, which is also an important topic for the industry nowadays, and so maybe starting from here in Arizona, and so with the recent MOU announcement with TSMC, what are the types of collaborations that we should be expecting? What are the degrees of the technology collaborations that are involved in this MOU? And maybe a more derivative question is with this kind of technology collaborations, would you be able to provide this kind of service to the other customers going forward?
Okay, so maybe let me start and then Megan can jump in. So first of all, let me kind of give a little bit of a landscape of the Arizona facility. It's in Peoria. It's about 15-20 minutes from the TSMC facility. Roughly 50 acres. It will be built in two phases. The first phase is what we're focused on today, and it will be focused on advanced packaging. So that's wafer bump, probe, assembly, things like CoWoS type structures, and then final test. So we've had press releases with Apple and TSMC that they will be important customers. And then we've had this MOU, like you mentioned, with TSMC talking about bringing advanced packaging, TSMC type structures into Arizona.
So within that collaboration, the mindset is for mutual customers where TSMC enables us to do the packaging in the Arizona facility. It would be complementary to the same type of structure that they're building in Asia. So a customer can get a package from Taiwan that looks the same as a package in the U.S. with the same silicon structures. So I think that's the intent. Again, it will not be captive to those two customers.
So we would envision that we have this TSMC structure and MOU and technology to where we're working with TSMC and the customers in a collaborative type structure. And then we'll have other customers, maybe TSMC silicon, maybe not TSMC silicon, where that package may be more of an Amkor technology package. So again, not captive. I think that's important. We see a lot of interest related to customers around, again, some level of diversification and having a U.S. supply chain.
So I think we're really excited about being the most scaled OSAT to come to the U.S. and partnering with TSMC and also with Apple. We do have a preliminary memorandum of terms with the CHIPS Office for a $400 million direct grant to support the build-out of this facility. And then we'll also take advantage of the investment tax credits, which are sizable. It's up to 25% of that investment. So we'll break ground mid-year next year. That will be about an 18- to 24-month construction. So we would anticipate bringing in equipment and starting to qualify at the end of 2027 with meaningful revenue into 2028.
Got it. Well, so you mentioned the revenue contribution is starting to come through maybe from late 2027. And so maybe from 2028, what type of the revenue contribution should we expect from this expansion?
Yeah, it's a little early at this point to give some size of revenue. And as Kevin mentioned, this is going to be very targeted to the most advanced products. We have not yet confirmed exactly what those products are, but we are excited to see our partnership and growth here in the U.S.
Got it. And so you do have other projects, one in Vietnam. And so how is that FAB important for your SiP expansion? How does that differentiate from the other FABs that you currently have for SiP?
Sure. So currently, our advanced SiP is concentrated in Korea. And we actually had seen that we were starting to get space constrained, and that was part of our strategic decision for Vietnam. That is also playing out to be quite convenient for those that are looking to de-risk their supply chain. So we did begin volume production in Q3 for advanced SiP, but also memory. So we have two products that will start to build scale into Q4 and over 2025. So we see that as an opportunity for incremental advanced SiP business, but also an opportunity to relieve some of that space constraints in Korea in order to expand some of those other very advanced technologies such as 2.5D and the roadmap for that technology.
Got it. Maybe one last question on regionalization for Portugal. You have also earlier, Kevin touched on Panel for that FAB. And so what's your current strategy for the expansion for that FAB? And should we also expand that FAB to be able to support other packaging solutions to address the demand coming from Europe?
Yeah, so good lead-in. So that facility, we acquired Nanium, and they were predominantly known for wafer-level fan-out. Since then, a lot of things have happened. Between the regionalization discussions, a lot of the European IDMs were looking for ways to strengthen the European supply chain. So we've added MEMS and sensors. We've added Flip Chip BGA. We already talked about the panel line, and we're working through collaborations with several customers on power modules as well. So that facility is definitely growing. We're actually in the middle of constructing another building on that site to continue that expansion. We have a couple of press releases, one with Infineon about a collaboration there, as well as an announcement with GlobalFoundries related to doing some bumping in Portugal to support their Dresden facility. So in general, a lot of regionalization efforts related to the European supply chain.
Okay. If I could ask actually just a follow-up to Arizona, because you mentioned on the tax credits and some of the grants, how do you think about the profitability as you ramp up that facility? And a second one, it's kind of a unique arrangement where TSMC may transfer or subcontract some of the advanced technology. Have you had those relationships in the past? And just curious, does that give you capability you can bring back to your Asia facilities?
Okay, so I'll start out on the profitability side. So of course, with the labor differential, that's the biggest concern. And so strategically for that facility to be profitable and not diluted, we've really targeted for the very advanced technologies. We are looking for it to be a targeted, small, mixed, high volume. We also need to ensure that we're maintaining high utilization throughout the year, which isn't necessarily how our other factories have. So those loading agreements with our customers will require smoother loading so that we're loading our U.S. facility first, and then anything necessary beyond that could be in Asia. There will also be a high level of automation, which is enabled by having the most advanced technologies here in the U.S. So those are the factors to ensure that we're able to operate profitably.
Yeah, and on the technology side, again, I think you have the baseline that we have complementary technologies already. So I think in our Asia supply chain, we would envision using our existing technologies and continue to expand those. And then this collaboration with TSMC would be very focused in the US.
Okay, good. I'll ask just a couple of financial questions since we have you. The CapEx, because you did talk continuing to invest in AI, but then I think when you talked about the other applications, still somewhat mixed. What's the implication for CapEx? You give maybe a breakdown how you're spending now and how you see that moving into 2025.
Okay, so for 2024, our guide is $750 million. About 75% of that is focused on machinery and equipment, both expansion and capability. That is primarily for the advanced packaging, specifically 2.5D. We had a significant expansion as well as advanced SiP where we had hit record levels. We also significantly invested in test in 2024. Absolute dollars, we really grew our test investment significantly and doubled from 2023's test investment. And then the other 25% is what we would characterize as common. So that's some of the build-outs, continued build-out for Vietnam, the building in Portugal, as well as automation, quality, R&D.
As we look at how CapEx spend will be in 2025, we would anticipate having similar capital intensity in the low teens. That would enable us to also continue to grow our 2.5D expansion and support other technology, R&D activities. We will begin spend in 2025 in U.S., albeit it won't be significant at that time, and it's still looking out to 2026 how those credits, etc., will offset the CapEx, so I would say for 2025, nothing unexpected and being able to continue to expand.
Okay, good. Maybe just a quick follow-up to that on capital allocation. Like I've seen you shift from net debt, but you've become more free cash flow positive. So now as you have been generating cash flow, and I think on that capital intensity continue, how do you see shareholder return, just dividend buybacks, or if there's other investment priorities?
Sure. Yeah, so we definitely have shifted away from one of the major priorities of debt reduction. We've met our leverage target. We'll continue to look at that and optimize that, and that has enabled us to have a shareholder framework that we announced in 2022, so the framework is to return 40%-50% of our free cash flow to shareholders over time using dividends. We did announce just this month or last month an increase in our regular quarterly dividend of 5%, but we also announced a special dividend that amounted to a total of $100 million, and that is our journey of meeting that framework of returning 40%-50%.
As far as buybacks, we don't currently have a program authorized, but that is something that the board and management will continue to look at as a potential other mechanism to return capital to shareholders. Yeah, we have one last minute. And so if I could squeeze in one last question on demand profile, especially within smartphones. Just also not just looking at the current quarter, but looking out for the coming three or four quarters, how would you characterize the dynamic within iOS versus Android?
iOS versus Android. So I guess what we see is if we look again at the cycle this year, a little bit different profile than we would typically expect to see. For next year, you may have the lower-end phone released in the first half, which typically gives us a little bit of a boost in Q1 and going into Q2. The second half, again, it's a content story. We continue to strive for increasing content and very focused on the premium tier segment. For Android, we do expect to see some recovery there. Yeah, so I think in general, we feel pretty good about it. It's not going to be a super strong supercycle unless there's a refresh related to AI content. And whether that happens this year, 2025 or 2026 is TBD. Anything you want to add?
No, that sounds good.
I think the risk of going slightly over. I'm just curious, auto industrial, if you're seeing, I mean, everyone's kind of asking if there's green shoots. It's been a tough segment. How do you see that outlook? Challenged. The inventory is still there. Again, I think you parse two different segments. You have the MCU area where there still seems to be a significant amount of inventory. I think you probably potentially heard from some of the IDMs that are here today that that's still an area that needs to be worked through. On more of the advanced side, we start talking about Flip Chip BGA and the ADAS infotainment. Those areas seem to look a little bit better. And so I think in general, it just comes down to how is that market going to pull through the inventory.
We still feel long-term that the trend towards EV is going to continue. Obviously, today, maybe a little bit of a pause there. People are more focused on hybrids and things, but I think in general, that trend will continue. And you see that every day in Arizona if you're here and the amount of Waymos that are around. They're everywhere. So I mean, the amount of electrical content with semiconductor content within that autonomous driving segment is significant.
Okay, great. On behalf of Sunny as well, I want to thank Megan and Kevin from Amkor for coming out and also for your joining. Thank you.
Thank you.
Thank you.
Thank you very much.
Okay.