My name is Peter Peng, small and mid-cap semiconductor analyst for the firm. I'm pleased to have Kevin Engel, CEO at Amkor, with us here today. I asked Kevin to start off with an overview of Amkor and a summary of the March quarter and June quarter outlook, and then we can kick off the Q&A.
Okay, thanks, Peter. And just for the record, COO, not CEO, if anybody thought that was—couldn't hear Peter on that. Let me just talk a little bit about Amkor. Amkor is an OSAT, or Outsourced Assembly and Test Company. We provide services to semiconductor companies that assemble and test that support the automotive, communications, computing, and consumer markets. If you look at our value proposition, there are three strategic pillars that we typically talk about. One is our advanced packaging technologies, where we work closely with our customers to provide packaging solutions for their next-generation products. Two, we look at our geographical diversity. With all the dynamics going on with tariffs and trade and other just kind of natural disasters, things like that, we are a very diverse footprint in manufacturing. That is an advantage.
Third, we try to partner very closely with our customers, again, to work with them on their product adoptions. Those are customers that lead in the markets in areas where we feel are going to be growth markets in the future. Another unique attribute of Amkor is we're the only U.S.-headquartered OSAT at scale. Again, with our very strong, diverse geographical footprint, we're in eight countries and soon to be in nine with our discussion around U.S. manufacturing. We'll be breaking ground the second half of this year. If we look at our Q1 revenue, it was $1.32 billion, and we're projecting for Q2 at midpoint, $1.425 billion. I think that's a brief introduction.
Okay, maybe just start with some of the near-term stuff because that's been pretty topical and top of investor minds. Your first half revenue is better than expected, and you talked about some of the cyclical recovery. Some would argue that the sub-take could be some pull-in of demand ahead of the expiration of the 90-day reciprocal tariff reprieve. Notably, one of your major wireless customers talked about pulling in component inventories in the near term to mitigate some of these near-term tariff-related challenges. Could you just walk us through the process of how you differentiate between pull-ins and just genuine cyclical order improvements?
Yeah, so I would just say, in general, we did not see a strong pull-in related to tariffs for the first quarter. We do not anticipate that for the second quarter either. We have been working with customers. For those that do not know Amkor, we have a very strong second half typically related to the communications launches that happen in the second half of the year. We have been working with customers to try to linearize some of that demand and pull that into Q1 and Q2. We have not really seen a dynamic of pull-ins related to a tariff-type environment.
Your lead times on your orders are relatively short. Yet you receive six-, nine-, twelve-month customers' rolling forecasts. Have these long-term forecasts evolved since Liberation Day in early April?
Yeah, so I would say definitely in the AI compute area, we did see some dynamics going on with swings up and down related to the forecast changes. I'd say if we look forward in time, the forecasts are starting to follow a more normalized pattern that we would see typically. I don't think we're seeing significant impacts today, but early on, we were definitely seeing some shifts.
As we look into the second half of the year, I think the companies indicated that the fundamentals remain intact, and you have some company-specific ramp expected. The Street is modeling a relatively flat year-over-year revenue profile for the second half of the year. Maybe you can outline some of the potential factors that might lead to upside or downside in this outlook.
Okay. Yeah. So, number one, I'm not confirming our second half or full-year guidance, but let me try to give you a little bit of color there. We've definitely, with all the dynamics going on with trade and tariff-type environment, there's definitely a lot of uncertainty. That is putting us in a position we want to be very, very careful with what we see for the rest of the year. If we kind of step through the different markets, on the communication side, that's our largest area of revenue. We definitely have some positive momentum there. We've seen the regaining of a socket that we were not participating in last year. We have high confidence in that. That is going to bode well for us. If we look at the computing space, we definitely saw some dynamics related to some of the restrictions on trade.
In general, now that segment is also looking pretty positive for us. If we think about automotive and industrial, and maybe we can talk about that a little bit later, but in general, what we see is on the advanced automotive segment, which are things like ADAS and infotainment, some strength in that area, whereas the more legacy devices, the MCUs in that area, a little bit weaker, still dealing with some inventory challenges with our customers. On consumer, we're still running production on a wearable device that we talked about last year. In general, if you think about positives and negatives, I'd say really it's just kind of easing, hopefully easing of the trade tensions. I think that would bode well for just consumer confidence and overall sell-through, which will help us out.
Got it. We've been in this prolonged down cycle for quite a while. Maybe you can talk about some of the pricing environment you're seeing and how that evolved.
Yeah, so I would say pricing is competitive but still rational. If we look back, there's, again, kind of talk about two different areas. If we look at the advanced packaging area, where there's less OSAT players, really less than a handful of viable alternatives, I'd say the pricing dynamic there is a little more stable. If you look on some of the more legacy or mainstream products, think of these as wire bond products in that space, there's a lot more OSAT providers in that space, a good bit of open capacity across the industry, and I'd say pricing is more competitive there.
Okay. All right, let's kind of move to your end markets. Starting with the communications segment, you mentioned about the socket loss. Regarding that, we estimate the revenue headwind from this socket loss in the current generation was approximately 2-3% of your total revenues. One, is our estimate close? As we look at the next generation phones, how should we assess the content gain and share opportunities there?
Yeah, so if we look at that specific opportunity, I already mentioned we have high confidence that we'll regain that socket this year. We would anticipate a similar level, assuming unit volumes are the same. As far as the impact, it's similar type scale. Again, if we think about how we prepare for launch cycles, obviously we're looking at how the qualification progresses, how we prepare our capacity planning, and all those are on track. We feel very good about that.
Okay, maybe just longer term, especially with this customer. Your revenue exposure with this customer has grown significantly, almost 2.7 times over the past five years, and it's almost a third of your total revenue mix. I guess, how are you thinking about share gains and content expansion opportunity with this customer?
Yeah, so it's a very important customer, obviously. We're very tightly tied to them. They are a technology leader where we make sure that we're aligned on their technologies. If you look at the portfolio today, we're across pretty much all of their main hardware segments. Those are areas where we continue to make sure we're focused on. I think we do see some opportunities and some applications within those hardware areas where we can do better and participate at a higher level. That's an area we're focused on. Just fighting every day, obviously, for market share. This customer from generation to generation of device, there's an opportunity to lose or gain share. We're very focused on making sure we maintain and grow.
Got it. Okay. Let me stop here and see if there's any questions.
Yeah, so can you talk a little bit about the advanced packaging business? What is the traction there for the AI products? And how has the recent H20 restrictions changed the broader of that business?
Okay, so yeah, the question was around our 2.5D portfolio, AI contribution, how H20 restrictions have impacted that business. Let me kind of step through a little bit of that. Our 2.5D portfolio does support some of these AI applications as well as just data center applications in general. A couple of different customers in that mix. If we look at how some of the restrictions have impacted us, we did definitely see an impact earlier in the year to where when some of the restrictions came in place that it reduced our forecast pretty dramatically. We are still running production today, so probably shipping in other markets. You will see how that continues to change over time, potentially with just this week's tariff situation. Hopefully, there will be some changes again, but still too early to see.
In general, that portfolio we see as a growing area. Obviously we'd start thinking about the next generation technologies, the RDL technologies, things like that. That's another area that we've been ramping our first products this year and then have a good pipeline behind that. Overall, we feel pretty positive on that advanced technology portfolio.
Okay, that's a good segue into the computing market. You have several growth drivers. Your RDL-based technology, the aforementioned 2.5D AI packaging solutions, co-package optics, among others. Maybe you can rank order these growth drivers in order of significance and discuss how impactful these opportunities are for your business.
Yeah, so maybe let me start by kind of just stepping through them a little bit. If you think about this compute segment in general, some of these packages are used in other markets as well. Within the compute market, there is 2.5, or let me start actually with just kind of large body standard flip- chip BGA. That is an area that is continuing to grow. For there, there are other areas that are very important, things like the thermal materials, lids, things like that to get heat out of the device. I'd say from a revenue perspective, that is an important area for us. We start migrating into multi-chip modules. That is still on a flip- chip BGA platform, again, in production for quite some time. That is an area that is continuing to mature. We start getting into the more advanced.
The 2.5D co-package optics, 2.5D, which is similar to TSMC's CoWoS-S. That has been in production. We talked about that a little bit already. The next generation would be our RDL flow. Amkor calls that Swift. For TSMC, the complementary technology would be their CoWoS-R. That is ramping for Amkor this year. We start talking about the bridge-based technologies. Our strategy there is to be a fast follower behind TSMC. That will be a little bit further out in time into next year. In general, all these technologies are kind of stepping stones for future growth drivers for us.
Beyond the H20 headwind, one of the other ones is that marquee customer transitioned to the CoWoS-L architecture. We estimate that you guys are almost generating close to $200 million in revenues for this customer in 2024. I know you guys are working on the CoWoS-L equivalent technology, which should begin to ramp in 2026. Maybe if you can discuss the partnership there and the longer-term revenue opportunities with this customer.
Yeah. So in general, that customer, we would definitely call them a partner. We're across their portfolio. I think when you look at their data center business, we don't just think about the GPU. We talk about kind of scaling up and scaling out across the server systems. For us, when we look in that pipeline, whether it's other applications from GPUs to others, that's definitely where we feel well-positioned and we have a strong pipeline. Related specifically to the CoWoS-L, obviously TSMC is participating there. Like I mentioned, we want to look at that from a fast follower perspective. The other piece that we didn't touch on too much is our advanced manufacturing is in Korea. When we start thinking about U.S. manufacturing, we've announced that we'll be breaking ground on our U.S. facility later this year.
That also gives us an opportunity with this customer to potentially do manufacturing and packaging in the U.S. longer term.
Beyond the merchant GPU market, there's been strong deployments of custom AI accelerators by the cloud and hyperscale providers. Maybe if you can discuss some of these engagements and when should we expect your customer base to kind of broaden outside this market customer.
Yeah, so I'd say that's happening today. This de-verticalization of this AI data center supply chain is definitely happening and we're participating there. If you look across a lot of these data center companies, they're looking at how they can optimize the chips to really meet their performance metrics. That's creating opportunities across for many OSATs, but for Amkor as well. We've already been running in production with one of these customers on a standard large body flip- chip BGA and working on the next generation devices, which will be RDL-based.
Any comment on next generation or advanced 3D packaging adoption of hybrid bonding and where you are versus competitors and the timing of adoption?
Yeah, the way we see hybrid bonding specifically is that that's more of a front-end application today. Obviously you may see that some of the foundries are participating in that space. For the OSATs, we feel that that process needs to mature a good bit more to really fit into our clean room environment and build that scale out. For us, we've been working with the suppliers, we've been doing a lot of evaluations, but we're holding off right now on any kind of adoption there.
Another growth driver in the compute space is the ARM-based CPUs, and that's gaining traction among the cloud hyperscalers for the data center builds as well as in PCs, just driven by the power efficiency and the adoption of AI. Additionally, these are using some of these more advanced packaging solutions you mentioned, the RDL-based technology. Maybe if you can just kind of discuss your design pipeline here and help us understand the revenue opportunity in this area.
Yeah, so it's actually interesting. We've talked about this many times. If you go back many, many years ago, obviously ARM was introduced into the mobile space. We've been participating in that for quite some time. Then we started to see ARM transition into PCs several years ago, laptops, again, for power efficiency, longer battery life. We participate with the first supplier or provider in that area and then also some of the followers. That's another area where we've been participating at a good level. If we look forward in the data center, now we're starting to see that ARM-based transition to data center also for power savings and cost savings across the data center. There's applications there. We're definitely participating, and it'll continue to grow, but it's still in its early stages.
Okay. Maybe moving towards the automotive and industrial market, that's starting to show some positive inflection, right? The team indicated that this market has reached its trough. Maybe just kind of comment on what you're seeing there. Maybe we can then talk about the different trends within that business.
Okay, sure. What we've seen, first of all, I'll start maybe with market projections. If you look at Gartner, they're projecting the areas that are going to have the fastest growth are things like ADAS, HPC, or kind of processing within the car and EV powertrains. Those are areas that we're very focused on. You think about that more from advanced packaging or more advanced packaging side. In that area, we've seen relative strength from our customers. We're excited about that area. If we look at the more legacy products, MCUs in that space, I agree with you. Most of our customers have said that they have seen the trough and are starting to climb out of that, but they all have different levels of inventory in the supply chain.
They really need to see that inventory burn off before we see any kind of a rebound that's substantial. We anticipate that'll take some time. Yeah, I think we feel that there's growth there, but it's going to take a little bit of time for this mainstream type products.
We talked about those diverging trends between your advanced product, which is 40% of sales, and then the mainstream product, which is 60% of sales. Maybe you can provide some context on the growth characteristics of these two segments over the past few years, and then how should we think about that going forward?
Yeah, so again, for this advanced area that I mentioned with Gartner and everything, definitely a lot of strength there. If I kind of think forward in time what that's going to look like, that will continue to grow. We also see, interestingly, a little bit more advanced silicon in those types of applications. If you think about the car in general, as you start to have more autonomous driving, you need more sensors in the car, you need more safety functionality in there. Those will proliferate down from premium cars into lower-end cars over time, and that will just expand this advanced package market with the automotive industry. We feel very good about that. Power modules, things like that for electrification of the car, that will continue to grow over time. I think that's still relatively early for Amkor, at least for the module space.
That's an area we're excited about. And then some of these, again, these legacy or mainstream wire bond type packages, those will come back, but it's going to take more time.
Kind of just summing all up, in general, do you think this segment is something that's growing in line with the SARs, or is it something that's going to be at a growth premium to the auto SARs?
When you say SARS, what do you mean?
Just units, auto units.
Oh, auto units. Yeah, I think there'll be, I mean, even in general, even if units stay the same, we would expect semiconductor content per vehicle to increase. I think that bodes well. Again, as you add more compute power, think about additional safety functionality, even many countries starting to require some of these safety functions be added to the cars. I think all of that drives semiconductor content within each unit.
Okay. Just maybe kind of moving on to testing business. You guys have announced some expansion plans for turnkey solutions and you're starting to prioritize some of this investment in test solutions. It's about 10%-15% of your business mix. Maybe just kind of elaborate on what opportunities are you pursuing in this testing sector? Additionally, how are you guys approaching opportunities in the AI testing market?
Okay, so let me maybe start a little bit about with the expansion we talked about. Within our Korea location, we talked about expanding within our existing building. That expansion will come online this year, so we'll see some benefit there this year. We also announced that we're going to start building another building on that campus, and that will be ready in 2027. If we think about testing in general, number one, it's definitely a focus area for Amkor. We try to focus on turnkey, so probe or wafer test, bumping, assembly, and final test. Really wafer test as well as final test. Having that all in one location is the best logistical flow for our customers from a cycle time perspective and things like that. That's definitely something we want to provide to the customers to add value.
When we think about the existing test, there's a high concentration of test in Taiwan, and we're very focused on, again, this geographical diversity to offer our customers outside just to risk mitigate and offer optionality. Korea is a good option for that. As we think about AI and just those types of applications, what we're seeing is with these chiplet type packages, the need for more testing, you're adding a lot of semiconductor content within each package. You want to make sure that you're not committing silicon prior to testing maybe the previous layer or the memory. That's adding additional assertions and also adding AI engines into things like application processors can increase the test time, driving the need for additional assets.
All those things are creating this dynamic to where we're growing the test business and need more space to support that growth.
Maybe just kind of switching to the financials. You have achieved a 20% gross margin in the past. Currently, there's some head cost winds from ramping up your Vietnam factory and then also just underutilization. Maybe just talk about the key drivers that could help you return to the 20% gross margin level.
Yeah, so let me start with Vietnam. Our Vietnam facility just started ramping late last year and into this year. If you look at Q1, we had about a burden of about 100 basis points to our corporate gross margin. If we look at Q2, we are estimating similar. Basically, we are ramping aggressively in that facility now. As we get into the third quarter and fourth quarter with that ramp cycle starting to really accelerate, we do expect that burden to start to come down. I think that is one positive over time. Other than that, really the story is just our overall utilization. If you look at the utilization we have been talking about for Q1, we have a lot of improvement we can do there.
Some segments of our business or some areas of the business, some package types have relatively high utilization, whereas we see other areas where we have very low utilization, again, related to automotive and some other areas. As we go into Q2, we should see a slight improvement in utilization. As we start going into the second half, where we typically see a lot of these communication ramps, that will just accelerate and continue to have a higher level of utilization, which will drop down to the bottom line.
Maybe just on that point on the Vietnam, just given this tariff uncertainty, has that kind of forced you to re-evaluate on how you're thinking about your geographic footprint and expansion strategy?
No, not really. Obviously, we'll see how the tariff situation plays out. I'd say our customers with the launches that we've been doing in the Vietnam facility are all full steam ahead. We'll see how that, again, what that looks like in the next months as we all know the tariff situation I'm sure will evolve, but today, no change in our strategies there.
On the Arizona fab, how are you thinking? What's the timing on that? I think you guys are planning on some constructions in the second half of the year. Maybe just update a timing on how you're thinking about the Arizona fab and when that goes online.
Yeah, so we've talked about potentially going bigger and faster. The plan right now would be we'd break ground second half of this year and scale from there. Our goal would be to finish the facility, start implementation in 2027, and then start ramping either late 2027 or 2028.
How important is the CHIPS Act to this project? If the team doesn't receive funding from it, what alternative strategies are there to fund this operation?
Yeah, so it's important. I mean, we've announced that we have a definitive agreement with the government for a $400 million grant. Obviously, that would be milestone-based. We've also had additional incentives through tax credits as well as city and state incentives of roughly another $400 million. Those are important. We've been obviously keeping in close contact with the Commerce Department to understand how their vision of the CHIPS Act can potentially change over time. With that being said, we don't have any reason to believe that that funding is at risk. We feel pretty comfortable about that.
Okay. Any questions?
Yes.
I just want to revisit your earlier comment on the recovery, and you said something about MCUs and autos. How particular are these legacy chips overall? Some of your customers have publicly said that they're seeing some improvement in the market. How do you see that playing out? Because you said that there is still some inventory in MCUs, but how is it playing out in autos as well?
Yeah, so I mean, basically, both those statements are right. I would say the customers are saying they're seeing improvement in orders, seeing improvements in the dynamics, and they've reached the bottom. I'd say that that's pretty consistent across multiple customers. They all have different levels of inventory. As far as them deciding to turn on their assembly, they need to kind of work through some of that inventory level. I think that will just take some time to burn off.
Can I just ask on your CapEx? It's not meaningfully different in dollars from where it was sort of prior to the onset of your AI packaging initiatives, and it's not meaningfully different as a percentage of your business, your intensity. Is that sustainable? As you ramp that, do you have a different CapEx profile looking forward?
Let me baseline a little bit about where our CapEx is. We're projecting about $850 million this year. Of that $850 million, roughly 70% of that is related to equipment and capacity. You're right, if we look at where we're spending that money, the vast majority of it would be on the advanced packaging side to support AI data center applications as well as some of our SiP. If you think about last year versus this year, the overall number was the same, but we spent a significantly more amount of CapEx on our facilities with our Vietnam facility. We've increased the CapEx related to the capacity expansions related to equipment. I think if we look forward, our goal is still to maintain that low teens in capital intensity long term.
We may have years here and there where we go a little bit higher, but that's our overall strategy.
Thank you. Yes, thank you for taking my question here. In terms of the AI evolution here, just in your view, in your opinion, what are investors underappreciating just about packaging in general? Where is that space going as this continues to unravel and improve?
Yeah, so underappreciate. I think complexity is one. Just all these things are, especially when you start talking about these different advanced technologies, they're not easy. It takes a good bit of time to do the development work. It's not like once you've nailed down one device, every other device is easy. You still have to make sure you're fine-tuning. Obviously, again, with more silicon content within each package, there's more expense within that package. Yield and quality are a key focus area. Because if you're throwing away a customer's silicon, that has a pretty major impact on everybody. Just making sure that you're doing those ramps in the proper profile, again, investing in the right areas to make sure you can support the customer ramps because some of these can be pretty steep. Those are some of the key areas.
Okay. We have a few minutes left. Do you have any closing remarks?
No, I think, again, just talk about some of our strategic pillars. Again, advanced technologies, obviously, we spent a lot of time talking about different advanced technologies and where we're participating there. We talk about our geographical footprint as a key differentiator across eight countries, obviously in China, in Taiwan, and several other countries offering diversity with our ninth country being the US starting breaking ground this year. Then our collaboration and partnerships with our customers to make sure that we're supporting them as market leaders and helping them advance their products moving forward. I think that's it. Thanks, Peter.
Thank you for participating.
Thank you.