Welcome. I'm Joe Moore from Morgan Stanley. Very happy to have with us today the CEO of Amkor, Giel Rutten. Thanks for being here today.
Thanks, Joe.
So maybe we just jump right into questions. If you could talk about your Arizona facility? In early October, you broke ground on the Arizona facility, increased the investment to $7 billion across two phases. You had comments from Howard Lutnick, Jensen Huang. This is clearly a really important project. Can you talk about why this is the right time for sort of packaging to be in America?
Yeah. I think this is indeed a very strategic and important project for Amkor. You know, we see that the supply chain for semiconductors is changing, and reshoring manufacturing back into the U.S. is very critical. You know, government is pushing, but also customers are pushing to reshore manufacturing capabilities back into the U.S. to ensure that when AI is getting deployed deeper into the economy, that governments keep control on their core technology, and that's the strategic rationale behind it. It's an important investment. It's a two-phase approach. Of course, we're not going to invest the full, let's say, $7 billion at day one. It's two phases. It's a step-by-step investment, and we will put only technology in place that is differentiating, and that's needed to be in the U.S.
Great. So you did mention Apple and NVIDIA as supporting key customers. Can you talk about what those conversations are like and to the extent that they're looking to reshore their supply chain? You know, how is that a motivation for you to do it?
We have conversations with these customers and also with TSMC on multiple fronts. It's about capacity that we're going to put in place, the need for the volume over the next couple of years, the technology requirements, what product families are customers planning to manufacture in the U.S., and what assembly and test capabilities do they need in the U.S. And with TSMC, it's on technology alignment between their silicon nodes and our packaging capability that we're going to put in place, also to make a seamless technology transition from Asia into the U.S. possible. I think these are the conversations, and they're actually pretty much in detail, and they're ongoing in a three-party mode as we speak.
Great. And can you talk about the profitability generally of doing this in Arizona to the extent that the costs are higher, but there's important strategic priorities, you know, where your customers would be willing to pay more to have the product made here?
Yeah. I mean, the technology, and that also holds for our customers, the product portfolio that will be manufactured in the U.S. clearly has value added by the location in which they are manufactured. So customers are seeing the strategic advantage to have that capability in the U.S. And, you know, we're going to put our most advanced technology in the U.S., mostly catering for the AI proliferation into data centers as well as edge devices. And that, of course, in the end, these devices will constitute the, let's say, a cornerstone of the transition into a digital economy, which also for governments gets very important to control that supply chain.
Great. Thank you. Your revenue and gross margin have a fairly seasonal component with a lot of fluctuation from half to half. You know, are you able to use this Arizona, the scale of this opportunity to kind of level that out? Just how does that affect your overall?
Yeah. Our current view is that the manufacturing services that we offer in Arizona will be accreted to our corporate gross margins. That also relates to the financial support that we're getting, the pricing power that we believe that we have in the US. We will implement the factory with a high degree of automation to mitigate the higher labor cost. So the overall business, as we see it, a combination of support from government, customers, high level of automation, and premium value-added technologies will make it accreted for our business.
Okay. Thank you. The $7 billion investment over several years is a lot of money for a company of your size and scale, but I know there's also a sharing of those costs and investments. Can you talk about that a little bit?
Yeah. Let me talk a bit on the investment level in the U.S. $7 billion is indeed a significant investment. You know, we increased our investment from $2 billion, that was our initial plan, to $7 billion because we see over the last one and a half to two years that the demand for local manufacturing in the U.S. is significantly increasing. Was it limited and very focused to one or two customers two years ago? The appetite is really increasing now, and we see that also reflected in TSMC increasing their investment level from $80 billion initially to $163 billion currently. So that scales up, and if we want to keep pace with that scale, that scale requirement, if we want to take most of the volume that's being generated by TSMC and assemble and package that, I think we see an opportunity to increase our scale also.
Now, if you take a little bit the support structure for this financing, I mean, public knowledge, we received $400 million from CHIPS funding. Of course, that's milestone related. In the U.S., there's strong support, even by the latest administration, for 35% investment tax credit. If you take that combined, that adds up to close to $3 billion of support in the U.S. And there are other ingredients that we work with customers currently to participate in different models that could be prepayment, co-investments, and we see that customers understand that that's needed in order to start building an ecosystem in the U.S. So we're pretty comfortable, and of course, we highly appreciate the support that we get from the government in the U.S.
Great. Well, congratulations. It's an impressive project that you guys are doing there. Maybe talk about some of your end markets, starting with compute. Your compute revenue was up mid-teens in 2024 and again in 2025. Can you talk about your role in the compute markets?
Yeah. I mean, proliferation of AI into data centers as well as into edge devices going forward is accelerating the compute market, not only with respect to growth, also with respect to innovation. Innovation is accelerating. And we, you know, we started off with what we call two and a half D with our lead customers, and I think that was two years ago. In the meantime, we achieved a couple of important milestones. We broadened our customer base from our lead customer initially to four customers currently, basically engaging with all the key leaders in the compute market. We also broadened and deepened our product portfolio, our technology portfolio, started with two and a half D or CoWoS-S.
We're currently, this year, we ramped up already the first product for what we call high-density fan-out, which is equivalent to TSMC CoWoS-R, and we have multiple products in the pipeline for further ramp. So we see strength in the markets, initially, of course, in data centers, but more and more interest in edge devices, whether it comes AI-enabled PCs, smartphones where a new architecture will be launched in one or two years from now, automotive connected cars or self-driving cars. So all of these AI features, once they proliferate, will drive more semiconductors, and all these semiconductors need advanced packaging. So we're currently built a project pipeline that is pretty strong, diversified with a deepening technology base.
Thank you. I spent time with TSMC a couple of weeks ago, and we just talked about the CoWoS market and your role, and they seem, you know, quite enthused to work with you guys. Just how should we think about this? This has been a market that they served initially broadening out with more suppliers. You know, how do you see your role with them? How do you see that cooperation going forward?
I mean, we have a longstanding, I would almost say multi-decade engagement with TSMC, where we engage on technology sharing, but also making sure that there's sufficient capacity in the industry. It started with bump. It started after that with test, with flip chip technology, and we're now moving more into the CoWoS domain. We have a strong cooperation in the sense that our strategy is to be complementary to what TSMC is doing. We don't see TSMC as a competitor. Our goal is to be complementary to what they're doing, and that can get more shapes and forms. I think if they choose certain technology areas that we could be complementary on other technology domains, like, for example, high-density fan-out currently, that seems to be more attractive for OSATs than for foundries.
In the U.S., we're definitely complementary from their U.S. investment, and we have a clear alignment in the U.S. on scale technologies. We even signed MOUs based on which technologies to ramp in the U.S. by which timeline. So close cooperation, it will continue, and I think we have a common goal is to build a complementary ecosystem in the U.S., and there's one thing we want to avoid is to overcapacitize or to undercapacitize. So that's very important.
And on that note, has CoWoS utilization put pressure on your gross margins, and are there dependencies, whether it's, you know, regional dependencies or, you know, customer dependencies there?
I mean, the AI market has been very dynamic over the last one to two years. I mean, if we take elements like trade restrictions that restricted exports of certain products from U.S. companies into China, all of that impacted the supply chain, and some of these also impacted our business. We were able to maintain a high level of utilization of the assets by shifting products around and working with customers. The way that we put our manufacturing lines together is that the individual equipment is highly fungible. So when we switch from, for example, CoWoS-S or two and a half D into CoWoS-R, we can utilize the equipment and make it fungible to shift from one technology or from one product portfolio to another. So in that sense, we're slightly differently, let's say, positioned than narrow-focused companies.
So we have a broad portfolio of technology, and we can shift capacities around pretty easily.
Excuse me, and then my customer. NVIDIA is about 85% revenue share of AI accelerators, but much lower unit share given the high prices, so there's a lot of other opportunities, a lot of investor enthusiasm as well for ASICs and other programs. Can you talk about your prospects to better broaden that out?
Yeah. We see the core of, well, first, let's step back. I mean, the AI market and the growth of AI, as is reflected in semiconductors, is still in the early stage of growth. Two years ago, AI was very low volume, hardly there, and you see that reflected in how the share price of some of these companies developed over the last two years. So we are in an early stage, and we also see that, you know, that newcomers with new solutions like ASICs take certain positions, very pointed point solutions, and they are creating alternatives for general-purpose GPU devices in data centers and in other products. So we believe, you know, we have a broad engagement. We work with both ASICs companies as well as with the GPU companies.
There's a third category of customers that we believe are important, and these are the what is labeled the hyperscaler companies. So these are Microsoft, Google, Facebook, Amazon Web Services. All of these companies have their own internal semiconductor activity, and they develop their own products. They're all architected around an ASIC model, and they are customers for Amkor also. So we're engaging with the major players there. One thing to see, Joe, here is from an opportunity perspective. If you go back five years ago in the first rollout of cloud compute and cloud storage, it was a vertically integrated market that was owned by Intel.
Now, the new market is basically all an outsourced manufacturing market, and that's a huge opportunity for a company like TSMC and a company like Amkor, that the future of cloud AI or AI on the edge is supported by an outsourced semiconductor manufacturing base, and that's a huge change and in that sense, a huge opportunity for us.
Yeah. That's great. Maybe we could pivot and talk about smartphone. Your business there, sorry, I'm losing my voice a little. Long day. The business there has been kind of flat. You've had some ups and downs with your major customer. Can you just talk about the trajectories there and your prospects for next year?
Yeah. I think on the smartphone market, we see two elements. First of all, a few comments on the iOS ecosystem, and I think that it's publicly known that there was an architectural transition where, you know, one critical slot we temporarily lost while the customer changed their architecture, and we worked with them to co-develop the solution. That's back in the current generation phones. The iPhone 17 is fully covered by our solution, so that's good. First half of this year, we still were on the old lower run rate. Second half, we're back. If you look to, you know, the communication market in general, third quarter, we had an all-time record quarter, and the fourth quarter, we guided the 20% year on year up. So we definitely see that we are, you know, that the market is good and our share in the market is completely recovered.
On the Android side, Android went through a couple of very difficult years. There was inventory in the China market. Some of the Samsung projects were not that successful, but we're currently seeing Android is coming back strongly, and we hold a very strong position in the high end of Android. So the combination actually is a good starting point for this year, for 2026.
Okay. And as you think about next year, one of people's concerns is memory pricing going higher. Do you see that causing any hesitation in builds? Do you think there could be any demand-side impact from that? And just it's not just that the price is going higher and there's an elastic impact. I'm actually hearing about situations where people are, you know, worried about even being able to build stuff, you know, given those issues.
Yeah. Yeah, the higher memory prices, in our view, could and maybe will impact some segments of the smartphone market, but certainly the consumer market. We expect an impact mostly on the low end and mid-range of the phones. If you take the memory part of the bill of material in the low end phones, it's about 20% of the bill of material is memory. In the high end phones, it's less than 10%. Also, on the high end market, I think we see that some of these customers have long-term contracts with memory companies. They have more buying power on the memory side, and they can demand a higher price for increased memory in their market. So we see an impact mostly on the low and mid-range part of the mobile phones. For consumers, same thing holds.
I think if you're operating at the low end of the consumer market, it's tough if you need memory. We also expect that memory prices stay up there for a while, and we also see that, for example, on the NAND market, that that will be a tight supply because currently the preference for memory makers is to invest in DRAM, specifically because AI is using a lot of DRAM and in high bandwidth memory, same thing. So the preferred solution is not NAND flash, and that creates a shortage there, and that could impact the phone market also. So multiple dynamics. In the end, I think for us, for Amkor, if it doesn't result in a market decline, it's good.
I think we're mostly focused on NAND flash, and we see that price sensitivity is improving in a sense that customers are not that keen to get a lower price, but they more want to secure supply.
Okay. I want to go into financials a little bit. Let me see first if we have questions from the audience. Sorry. No, I'll just keep going. So from a financial standpoint, you increased your 2025 CapEx from $850 million to $950 million last quarter. How do you think about capital intensity going forward?
I mean, so far over the last, let's say, years, we were operating, or we operated the business at a capital intensity of in the low teens, and we see that that may increase slightly basically because of the investments we're making in the US. Although, you know, Amkor has a track record of being a very prudent company in spending cash, so we're, you know, we try to manage it in a very prudent way, and I mentioned some of the cash contributions earlier on, and we're happy to be able to do that, so for, you know, with respect to slightly higher capital intensity, you know, we prepared our balance sheet in order to cater for that, so we have a very strong balance sheet with over $2 billion of cash on our balance sheet, so overall, I think we can carry this.
Okay. Great. And then the revenue performance has been strong, but gross margins have been generally under some pressure. You were at 20% three years ago. You're at 15% now, and you've talked about some improvement, but kind of gradual. Can you talk about, you know, long term, is there a path back to 20% gross margins, or is that sort of an anomalous point in time?
I think our ambition definitely is to go back to that level, Joe. I mean, there are a couple of important ingredients currently that have a downward pressure on margin. The first one is the underutilization of our mainstream manufacturing lines. Mainstream basically is catering for industrial and automotive business products like, you know, microcontrollers, analog components that go into industrial and automotive applications. After COVID, there was a significant surge in inventory, and that took a while to build that down, and we see that coming back the last, let's say, three quarters, we see year on year growth. For a company like Amkor, an OSAT company, manufacturing or factory utilization is the key element of profitability, so with heavily underutilized lines, you know, below 50% utilization in our mainstream business, that puts pressure on our margin, and we believe that going forward that will improve.
On top of that, we make some specific organizational adaptations, specifically in Japan, to right-size our Japan cost base. So that combination is an important thing. There are two other elements that will contribute to a positive or will have a positive effect on our margin. One is our Vietnam factory. I think we put that into production last year. We're ramping up Vietnam currently. We expect Vietnam as an organization to be break-even early part of 2026, and from there, we can start building up to have a positive gross margin contribution from that. We have a lot of interest of that factory that we can offer customers a second source to their China manufacturing, and five of our top 10 customers are currently qualifying our Vietnam factory. So we believe that once we're going through that initial curve, that Vietnam will be accretive to our corporate gross margin.
fourth element is on, or the third element is on, advanced packaging. I think we invested a lot, not only in capital investments, but also in people to staff for advanced packaging ramps. And while, you know, while we're still building the scale, we believe that once the scale is there, that will become significantly accretive. So overall, we believe that with these three activities and actions in place, that we will get back to historic levels for our gross margin, Joe.
Your gross margin guidance for the next few quarters, you know, what's kind of contemplated for those utilization variables? Do you, and you're sort of trying to be conservative about where that may go. Is there a possibility of upside if those utilizations end up higher?
Currently we see performance, and you know, we stick to our guidance. We guided the, let's say, the $1.8 billion-$5 billion for the fourth quarter, so I don't want to see that. But if we take a step back, then we see a more positive trend than we expected when it comes to mainstream utilization of these lines, and that gives us good hope. Keep in mind that our mainstream factories are generally highly depreciated already. So once we fill these factories, I think the gross margin is definitely very rapidly goes above corporate gross margin.
Okay. Very helpful. So, final question: You're retiring at the end of the year, and I know, you know, you've achieved a lot when you're here. If you think about the next five years, you know, can you just talk about generally where you see Amkor?
I mean, I can say whatever I want now.
That's right.
It's not up to me. You know, I believe that, you know, Amkor is part of a semiconductor supply chain, and there are a couple of macro trends that are dictating the industry forward. They're irreversible. I think people can say, well, maybe this happened, but there are certain known elements in the industry, and they are in line with Amkor's strategy. I think the technology innovation is accelerating. You know, if we see where AI is going, what happens in data centers where the compute power that's required is needed, I think you need significant acceleration and innovation both on the compute side, but also on the power management side. I think this is ongoing. We invested a lot in that, and we believe that that will move forward.
A second core pillar besides technology for Amkor is regionalization of manufacturing where governments realize now that you cannot be dependent on one or two Asian countries when it comes to core technologies that are the foundation of a digital economy in the future. So you need to get control. It was a little bit of a far-fetched idea two years ago, but it's sinking in currently, and we are investing in that. I think we started to invest in that four years ago. You know, we acquired a factory in Portugal, for example, in Europe to support the European automotive industry. We're building in the U.S. We're upgrading our Japan manufacturing base, and we built in Vietnam as an alternative for China manufacturing. So that regionalization and the criticality of regional supply chains will continue. It will not be reversed.
I don't expect tomorrow that everybody says, "Well, okay, let's go back to and manufacture everything in China." I don't think that that's smart, but also I don't think that politically that's an acceptable solution. And then the third element, which is very important, is relationships with customers and relationships in the supply chain with partners like foundries. More and more the business develops in a virtually integrated way that there is a very deep cross-linked cooperation between customers, suppliers, and the supply chain just to avoid technology gaps, oversupply, overinvestment while building these new supply chains regionally. That will continue. So where will Amkor be in five or ten years from now? We believe we, you know, we have the ingredients to support that growth in the technologies, the geographic presence, and we are 55 years in the industry with deep relations with customers.
So, we're well positioned, and I believe that, you know, with the industry growing along the two main factors being AI on one hand and AI proliferation and edge devices like smartphone, automotive, et cetera, you know, we will be able to cater for that, and that will bring us on an above-average growth path going forward, Joe. I'm confident that that will happen.
We appreciate you leaving the company in a strong position, and good luck in your career.
Thanks, Joe. It's appreciated.
So we'll wrap it up there. The conference organizer did ask me to remind people if you aren't coming tomorrow, if you could leave your lanyard. I think there's a bin outside behind because they're running short. All right. Thank you so much.
Thank you. Thank you so much.