Okay, great. Thanks, everybody, for joining us for the next session at our Deutsche Bank Technology Conference. I am Melissa Weathers. I lead up US Semicap Equipment and Memory Research here at Deutsche Bank. This afternoon, morning still, we are pleased to have Brice Hill, CFO of Applied Materials, joining us. Thanks for joining.
Thanks for hosting us. Beautiful conference. Appreciate you having us here.
Great. Before we begin, if anybody has any questions from the audience, feel free to raise your hand and we can get a microphone over to you to get that answered. I think to kick off, the smartest place to start would be you reported earnings two weeks ago. You talked about maybe some incremental headwinds going into the second half of the year, maybe a little bit different than what you had been thinking 90 days ago. I think just to level set everyone, can you remind us what was your messaging on that call? What are the key takeaways that you want investors to have taken from that call? What are you seeing into your October quarter?
Yeah, great. When we think about this year, if we back up 90 days and even further than that, for Applied Materials and for the industry, we've been thinking, how strong will DRAM and leading logic be this year? Will it be strong enough to offset any headwinds in ICAPS, the mature logic space? In the ICAPS space in China, there was a huge buildout in 2023 and 2024. We've kind of expected 2025 to be slower. On the positive side, with all the strength of AI, we've expected DRAM to be really strong and we've expected leading logic to be really strong. Q3 was record revenues for us, record earnings per share, and it was exhibiting that sort of behavior. We had expected leading logic to be linearly accelerating through the course of the year. When we gave our guide for Q4, it was different than what we expected.
The expected part and the headwind part was China. China's lower, lower than 2024. We've been expecting that. Still a good market, but not going as fast as 2024. What was unexpected was that leading edge wasn't linear. We model, of course, all the customers' factories and the capacities and the timing of the equipment, but the pickup rate wasn't exactly what we expected. I would dig into that a little bit because people ask us, what does that mean? Is there a slower demand? Is there a change in expectations? No change in expectations on leading logic. If we dig into it, if you look at the foundry ecosystem, the utilization on the leading edge processes is 100%. From a design perspective, reported designs are higher on these nodes and this node than prior nodes.
If you look at the performance of the process itself, a gate-all-around process, it's 20% - 30% more power-efficient compute than the prior. It's a good node to put a design on, especially in the AI space where you need more power efficiency going forward. When we listen to the cloud service providers and their CapEx forecasts, I think investors know that their CapEx forecasts have only been going up. When you look at those, the CapEx on AI compute and high-performance compute systems, what do those systems consist of? They consist of accelerators, GPUs, processors, then with stacks of DRAM, the high-bandwidth memory around those systems. That has a lot of pull for DRAM, a lot of pull for leading logic. That's what we expected just to continue. However, the pickup rate from an equipment perspective is not linear. We're a little bit more tilted towards one large customer.
It's going to be a little bit more tilted towards their factories and their timing. It will be more uneven than it was in the past from a buildout perspective. I would just leave it as that was the piece that's unexpected. I would put it, Melissa, in the construct of we expect, you know, leading logic and DRAM to grow over time and, in fact, be the two largest and fastest equipment markets.
I want to dig in more onto that leading-edge piece. Really quick, as we think about what your peers have said versus last quarter, the messaging across all the semicaps has actually been pretty different, which is strange given that you all have the same customers. How do you account for the difference in your outlook versus what your peers have talked about?
Yeah, I do think, you know, we did see some companies say they saw signs of a slower business and we saw other companies. Every company is at a different timing window with respect to China shipments, with respect to whether they can serve China customers or not, if they have more exposure to NAND, which grew a lot this year, or less exposure to NAND. There is a lot to sort out, which is your job and makes it tough for you. For us, you know, if you back up for Applied Materials, if I can just reiterate the strategy that Gary's put in place, we think DRAM and leading logic will be the two fastest equipment markets for the next 5+ years. The strategy of the company is to focus on developing new applications when the architectures change in those technologies.
What that really says is that's the time when our customers reevaluate their equipment set. They're making a change to the architecture, like the gate-all-around processor. They've got new, you know, structures to build and they reevaluate the equipment set going in. We're focusing our development efforts on when you see those architecture changes in leading logic, which would be gate-all-around, backside power, and eventually CFET. If you look at DRAM, which should be 6F squared, then 4F squared, then 3D DRAM, when you see those architecture changes, Applied already has good share and we expect to gain share once those transitions get implemented. That's really the focus for the company in terms of the fastest growing markets. When we think about AI, you know, those strategies, that strategy has been in place for a while.
We didn't know about AI when we started working on those architectures, but AI is a perfect example of the market pull for compute memory, DRAM, and of course for compute logic.
Yeah, while we're talking about AI, a lot of the high-end GPUs and even some of the ASICs, those still aren't at the very, very bleeding edge. Can you just talk about what are your expectations in terms of actual bleeding edge adoption for those processors? Could we ever get like two-nanometer GPUs? What's your view on that?
Personally, I'm not a chip architect, but we talked a little bit about this before. I would expect all of that to migrate to the leading edge. If it's not the node that's currently ramping, then maybe the node after. The reason is, there's two reasons, but the main reason is power performance. The transistor, the gate-all-around transistor, was designed to be better power-efficient compute and more reliable than the FinFET transistor. It's in the architecture. It's more efficient, more reliable, and consumes less power than the prior generation of technology. I expect all those designs, once they're comfortable with the tools and implementing and getting the design there and the process itself, I expect all those designs to go into that space. I think it would be a huge change for the whole industry if somehow that didn't evolve. Everybody knows the world is worried about power.
The way to solve the power need is to make the compute more power efficient. There are multiple different ways, including in packaging, to do that. I think you'll see all those designs move forward.
Maybe on the sticking with the near-term side, and then we can get into some of those long-term growth opportunities. WFE, I'm going to try and get an outlook from you for 2026. As semicap investors, it's very difficult for us to get a clear picture of what could happen in 2026. You have all these different moving pieces between China, between the leading edge and DRAM. You don't have to give me a hard number. As we think about the moving pieces into 2026, how would you guide us? Where should we start?
I think the dynamic will be similar from a growth perspective for 2026. We'll have strength driven by AI in DRAM and leading logic. We'll have continued growth of new packaging elements for that space in particular. If there's a headwind, the headwind will still be in the China mature nodes, which we call ICAPS. You know, the devices, ICAPS for investors, we talk about IoT devices, communications, the auto market, power, sensors. All those things that are built on essentially 14 nm and larger geometries, analog chips and the like. Because China invested so much in 2023 and 2024, even though the end markets for those mature logic devices are growing, growing at mid to high single digits, there's a lot of capacity in place. We're still adding customers. We're still adding capacity.
It's going to be slower developing for the next few quarters still than we're going to see in the end markets. Going back up to the top, we would expect growth generally over a five-year period, especially in DRAM and leading logic. No reason to expect 2026 to be different than that. That'll depend on the macros. The headwind will be ICAPS. The question is, is leading logic and DRAM going to grow strongly enough to offset that? I think the dynamic is similar to what we've seen in the past few quarters.
Okay.
I would add one thing. Sorry, Melissa. This year's growth, our semi-equipment business is growing with our guide about 4%. That gives you, assuming a constant share, some perspective on what the equipment business is doing this year. You should see WFE growing this year. I'm just saying the trends, there's not a reason to expect the trends to be different.
On the leading edge piece, that was definitely the biggest surprise to me in this most recent quarter. One of the things that you've talked about, which I found interesting, is the difference in customer order patterns and order behavior pre-COVID, during COVID, and then in this post-COVID world. Can you add a little bit more color? How have your engagements with customers, have they changed? What is your visibility going forward as to what those leading edge customers are spending?
Yeah, I think if we start in China, we have more customers and a lot of them are newer customers. If you compare them to longstanding customers that are larger, the longstanding customers have a mature process for forecasting their demand, forecasting their facilities, and sharing that with vendors. First of all, a lot more of the market has been in China or in the rest of the world with less mature processes and with the policy uncertainties and trade uncertainties and tariff uncertainties. What we've seen is customers kind of wait till the last minute before they commit to the order that they're looking for. Right. Looking at the mature, larger customers, we've actually seen the same behavior, sort of late commits to the orders that they're taking.
What we're doing is reminding everybody during COVID, we all struggled with supply chain because there was significant volatility in demand for the supply chain. Applied Materials can take more orders or take fewer orders in any particular quarter. What's really hard is if you have eight to ten thousand components in a system and you're trying to move a large number of suppliers at the same time, they really can't move up and down that much. What we're doing is reminding our customers, we know there's more volatility in the environment, which hopefully will settle out. We really need to have a perspective. What's the two-year roadmap? What's a good one-year build plan? For six months, we really have to have a clear perspective on what the builds and deliveries are.
I think we've split away from that a little bit, but I think everybody's on the same page that in order to manage the large ecosystem, we have to follow that behavior.
Are you seeing any signs of share shift?
If I look at leading-edge and DRAM where we're focused, we don't think there's any change in share. The share shift, of course, the biggest component is in China. This year, we announced early in the year that there was $400 million of business in our backlog that we would not be able to serve because of the entity-listed customers in China. We do have share loss there that's based on the rules that were put in place. I would step back and just think about that for a second. If we think about the China business that we can't serve, the memory business, whatever is leading logic there, and any restricted accounts, the question is going to be, how long will this regime of rules be in place? I think there's two big questions. The first one is a technology question.
Will there be new technologies invented in China that use different tools that are competitive? If that happens, then we don't have access to that share like we do today. If you can't really keep up without access to the equipment, at least over the next 10 years, then more of that demand should shift back to the multinational vendors where Applied Materials does provide that equipment. I think that's a large unknown in the ecosystem right now. That's one thing that could happen that could benefit Applied Materials. We recently saw that NVIDIA can sell components to China. That's going to put more demand on Applied Materials tools. That's an example of how that could happen. The second scenario would be if the government changes its stance from restricting equipment to actually, it's okay if we provide equipment and China can build on top of that equipment.
If that happens, that's another way where Applied Materials could sell more equipment to that customer base or that use base. I think that's a big unknown. Going back to your question, yes, where we see share shift is in China and we expected that it's rules-based. The question for us when we talk about will we see overall share gain or not, we think we're going to gain share on leading-edge. We expect to gain share also on DRAM. Will that be large enough to offset any share loss in China? That's what we're looking at.
There's been a lot of news in the semi-sector over the last week or two, particularly in the second-tier foundry players. I don't know if you heard the U.S. government is now a shareholder of one of your large customers, Intel. We've had some news on the Samsung foundry side as well. A lot of activity on those second-tier foundry logic players. As we think about those incremental developments, how is that factoring into your outlook for leading-edge logic spending?
Yeah, I'll start with it's not much of a factor. The reason I say that is, you know, people used to ask us, investors used to ask us, hey, if the governments are incentivizing your customers to build in the U.S. or build somewhere else, isn't your demand forecast going up? We would say no. At the highest level, demand is driven by PCs, data centers, smartphones. All that those incentives do is have a customer, instead of building in Taiwan, they're going to build in the U.S. The view was it mattered a little bit at the margin, but it didn't matter in total. Similarly, our expectation is that if the second-tier foundries are more successful, it's not going to be a major change in the way we look at the market.
At a high level, we would model the market for leading-edge logic and DRAM as an example to be roughly the same. It does matter at the margin. The way I think about it is if you move the needle all the way to one side of the equation and you say there's only one foundry, then one foundry will, you know, operate at a very high utilization. If you move it to the other side and you say there's four competing foundries, then they'll altogether operate at a little bit lower utilization because they're competing against each other. We think it matters at the margin, but hopefully for investors, it's sort of comforting. We don't think anybody's going to build extra fabs that don't have designs. When you're building a factory, you know whether you have a design or not, at least when you're ready to order the equipment.
You know, do I have a customer? Are they taping out? Are they ready to commit to volumes? Then you're ordering the equipment because there's enough time between your tapeout and design win and actually start of production that you can install the equipment. Hopefully that answers the question. It's a positive. Competition is a positive. It'll help drive performance. It matters at the margin, but we're not changing our forecast, our five-year forecast because of this dynamic.
I think that's a prudent way to handle it. Maybe last question on the leading-edge piece, and then maybe we can get into some China. On your call, you talked about 30,000 wafer starts per month of two nm. I don't recall if it was two nm gate-all-around. One, can you clarify which nodes that is? Also, as we think about the timing of this buildout, clearly maybe there's some seasonality or a little lull in spending in the second half. As we think about the scale of the two-nanometer nodes, can you talk about what that could mean for the pace of growth that you're expecting for leading-edge nodes?
Yeah, first of all, the number I think I used, bigger number, 300,000.
Sorry, 300,000.
300,000 wafer starts per month. That's what we think is a good expectation for the size of the gate-all-around node. Larger than the prior node and primarily driven by the capability, the 20%- 30% power efficiency of the transistor that you can build on that node. We think it's a good landing spot. If you look at nodes, some nodes are really good landing spots. They pick up a lot of designs. They're large. Other nodes have been smaller. I think the 10 nm node was an example of a node that was sort of an interim node. Seven was really strong in that example. Two should be a strong node from our perspective.
If you look at Applied Materials, we talk about how our SAM, our share of market, has gone up from $12 billion for 100,000 wafer starts of capacity to $14 billion of SAM for 100,000 wafer starts. That's our opportunity to grow share and to sell more equipment into those nodes. If you look over the last year, two years, we sold $2.5 billion of gate-all-around type equipment in 2024. We said $4.5 billion in 2025. There's $7 billion of equipment, 50% of that $14 billion number. That would equate with roughly, by the end of this year, 100,000 wafer starts of capacity being put in place divided across four customers. We would expect there's a long way to go in this ramp. Somebody asked me earlier today, when does that ramp end? That ramp ends when your next node that has good volume starts ramping.
Great. Now let's touch on the China piece. There are a lot of moving pieces within your China business. It can be hard to keep up with. As you look at your China business today, really strong growth in the last two years, starting to taper off going forward. How would you characterize the overall China business between you've got multinational spending, you've got some entity list restrictions, and you've also got just a normal digestion of some capacity put in place at anything above 28 nm. At a high level, can you just summarize what you're seeing in China, what your outlook is there?
Yeah. First of all, I think the perspective from inside Applied Materials is really good, market growing. I mean, you know, from the perspective of we're adding customers. This quarter, when we did our account reviews, I think there's a new customer that's building display drivers on 28 nm. There's another new customer that's building computer image sensors on 28 nm. In the last few years, the buildout in China has focused on even more mature nodes like 40, 45, 60. What we're seeing in the next couple of years is a lot more buildout is going to be on 28 nm. Applied Materials has a good footprint in 28 nm. We think a lot of these customers, when they start, they're going to use the same equipment set that the big foundries have used for 28 nm. That plays to our advantage from that perspective.
We have excellent service capabilities that we think we'll be able to offer those newer customers who are more likely to use the services as it's qualified labor and, you know, sort of fab experts that come in and can help them ramp. From that perspective, growing market, a lot of it's driven by local incentives. Those new customers I talk about, they're incentivized to build a factory in a new province. Of course, we help with that. Having said all that, yes, 2024 was a really big year. We think we won't have as big a market or as fast a market as we did in 2024 until, I call it digestion, until the end market demand sort of catches up with the capacity that's been put in place.
If it were one customer and one part called a monolith, then you would say, oh, I don't need to add any capacity for more than a year because if I'm at 80% utilization and the market grows mid to high single digits, I can at least wait a year before I need to buy any equipment. The issue is it's not one customer, it's lots of customers, lots of different products, lots of different factories. They're all scheduling construction installations. The market keeps going. I think it's complex. We've had investors ask us, who are these customers? What are they building? Are they profitable? Do they have real businesses? We're all going to learn more, but we expect that market to continue to grow, continue to be strong. The end markets itself, we're talking ICAPS, mid to high single digit growth over time.
The demand will catch up to the capacity, if you will. Last thing on the restrictions, I sort of gave you my thoughts. This environment that we're living in with the restrictions, I don't think it's permanent from the perspective of the technologies are going to evolve and change one way or the other. Either China will have successful roadmaps that are competitive, or they'll have to shift back to multinationals, or the actual rules environment will shift. You know, we don't speculate on which way that goes, but hopefully that gives you some investors a way to think about that dynamic.
Super helpful. Let's touch on ICAPS, but maybe the non-China piece of ICAPS. On your earnings call, you talked about some green shoots there. This has been a very prolonged cycle. I think we're all anxiously waiting for the analog and the broad-based earnings world to reaccelerate. What are those green shoots that you're seeing? What is your outlook on the ICAPS rest of the world part of your business?
Yes, the easiest one is we see utilization improving. We see it in the current quarter, we see it in our outlook quarter, and we expect that to continue over the next few quarters. What's happening underneath is I think the rest of the world is needing to reconcile how much capacity has been put in place in China and how much they will serve. If our rest of the world companies thought they were selling into China, there's some rationalization that has to be done on the capacity plants. The green shoots are utilizations are starting to climb even after all the capacity that's been put in place. That's good. The second is specifically in our Q4 outlook, of course, China ICAPS is lower, but rest of the world ICAPS is higher. It's indicative of some of the investments that we're expecting to see in rest of the world.
In fact, our internal forecast calls for rest of the world ICAPS to grow definitely over the five-year period that we're forecasting. We call for growth in that space. Not giving a 2026 guide, but we expect some of it's driven by 200 mm- 300 mm transitions, but the rest is driven by the underlying market is growing.
Is there any tech transition? I know the business is super broad, but what kind of innovations are you seeing in those ICAPS nodes?
I think the number one space to think about is power. Our teams work a lot on what's called compound semi, new materials for substrates or for the transistor silicon itself. I think that's where you see a lot of the innovative work, how do we make these components so they're much more energy efficient than, you know, prior technology generations. That's where you'll see a lot of the innovation.
Speaking of innovation, let's move on to DRAM. As you think about the current supply-demand balance within the DRAM market, I think some investors are concerned. There's a lot of questions on where will supply-demand shake out in 2026. What trends are you seeing on the DRAM piece? Maybe let's talk about it as a whole and then we can get into HBM.
The best trend, HBM, is AI-related components, including HBM, we think are growing at a 30%- 40% CAGR. When you look at HBM DRAM, I think about 15% of DRAM capacity wafers today are allocated for HBM production. That's growing at a 30 %- 40% CAGR. That's where you see we're going to have a record, either a record year or the second best year for DRAM this year. If you look back at 2024, we had a significant China business. This year we're at a record or close to a record, meaning the multinationals have made up all of that business we had in China in the prior year. I think that's close to 50% growth for the multinationals in DRAM. That's both a pull on the memory itself and also driven by that HBM.
We think the trend, you know, if you look across the five-year horizon I highlighted earlier, we think DRAM could be the fastest growing equipment market. It'll be between DRAM and leading logic over the next five years.
As we think about HBM specifically, that market seems to be pretty binary in terms of capacity additions. We're waiting on clarity on qualifications at certain customers. How are you, what's embedded in your forecast? How do you, as a CFO, plan given that huge binary aspect of we don't know what supply is going to look like?
The good news is this is more similar to the conversation we had about leading-edge logic. We don't model HBM at the foundry level. We model HBM at what do we expect the high-bandwidth memory market to look like? How fast is it growing? What's the macro? As far as subdividing that between the memory vendors, we don't do that. Of course, we do that when we're getting ready to take orders and ship equipment. As far as when I say our DRAM forecast is to have that process technology be one of the fastest growers over the five years, it's at that macro level, how much HBM we expect to be deployed. I would just say, of course, it matters dearly to those vendors who win those designs. For us, we would say we'll support all of them. We certainly hope there's competition in the market.
the last couple of minutes that we have, I promise I'll get you a margin question as a CFO, but maybe really quick on the services side, the AGSP. Maybe can you help us think about the two different pieces? You've got the 200 mm piece and then you've got core, which you're guiding to grow low double digits. What trends are you seeing in that business? Specifically, I think you've talked about the subscriptions side of the business or your subscription revenue is near two thirds and growing. Can you talk about the drivers, what's driving that number higher?
Yeah, I'm definitely trying to highlight that business for investors by associating and highly correlating our dividend growth with that business. We've signaled for the core part of that business, which is over 90% of what we report in AGS. For the core part of that business, more than two thirds of it is subscription, so you can view that as recurring revenue. It's the services component and the spare parts component of that business. We do say that's growing at low double digits, and we expect it to continue to grow at low double digits. If you think about the dividend, I'll hopefully exhibit that with a dividend as we go forward. That'll just put that at the top of mind for investors.
If you look at total AGS for the year, that component that's less than 10% of the business today, that's 200 mm equipment that is reported in that services business, and that has shrunk this year. I think a lot of that has to do with the transition of 200 mm production to 300 mm production. We think there will be a continued 200 mm market. There are some technologies, especially in these compound semi space, where the technology actually can't be built yet on a 300 mm. It doesn't, just from a warpage perspective and from basically the heaviness of the material, et cetera, it doesn't work on a 300 mm wafer. There will be a continued 200 mm market for as far as we can see, but it will be smaller. Back up to the top, that services business, we expect low double digits growth rate.
The dynamic that's driving that and the reason it's faster than the equipment business is every day that we ship a tool, our installed base grows. Our customers, a lot of them are being incentivized to build in different cities than they normally build in. When you do that, you are looking for labor and looking for people that know how to work in a fab. Applied comes along and says we can provide 100 service technicians that are expert on this equipment that can help you yield and offer some AI-based intelligence as to how to tune that equipment. The pickup on that business is higher in this environment because of that dynamic. Finally, the services themselves are improving because we're using AI technologies to help improve the offerings.
All right, last 30 seconds that we have. Gross margins. Didn't get a lot of attention on your last earnings call, but where do we, how should we think about the baseline of gross margins from here? Your mix is very in flux. Help us on gross margins. I do want to ask on the value-based pricing initiatives.
Okay. So 48.1% gross margin for our guide with a lower China mix, which typically weighs on our gross margins. We looked at our Q4 guide as sort of a normal, it's a normal China mix, relatively normal business mix. I think it's a good indication of where we are today from a gross margin perspective. That's improved over the last couple of years. Our Q3 quarter of high 48% gross margins was 150 basis points above where we were a couple of years ago. We've made a lot of improvements. The portfolio is shifting to more critical tools. We've got a better pricing process in place. That's helping us. Of course, we're working, there are some headwinds with the tariff headwinds, but we're working through that. We have a global footprint. We feel good about our ability to manage that.
I think that is a good baseline, Melissa, and we'll continue to work on improvements in that.
Perfect. We are out of time. Thank you so much, Brice, for joining us and enjoy the rest of the conference, everybody.
Wonderful. Nice to see everybody. Thank you.