Got it. Welcome to day two of Citi Global TMT Conference. My name is Atif Malik. I cover U.S. semiconductors and semiconductor equipment stocks. It's my pleasure to welcome Doug Bettinger, EVP, Chief Financial Officer, Ram Ganesh, VP of Investor Relations at Lam Research. I'll kick off with my five set questions first, and then we'll give an opportunity to the audience to ask their questions. If you have a question, please raise your hand, and the mic will come to you. Doug is going to open with some-
Yeah. I technically need to kick all of these things off with a reminder that the safe harbor language that is on our investor relations website is relevant to anything that I may say today. I may make forward-looking statements, but I'll be quite clear, I'm not going to tell you anything new today. It'll be consistent with everything you've heard from the company before. If you have an expectation something new is coming out, it's probably not. But please have a look at our safe harbor nonetheless. So there, my lawyers will be happy, Atif.
All right.
Now we can get going.
All right, let's get started. Doug, let's talk about the fab equipment outlook and next year's visibility. With calendar 2026 WFE expectations recently raised to the low $150 billion range.
Yep.
On accelerating AI demand, where is the upside coming from this year? Also, I forgot the adjective you guys have used in terms of super exciting, whatever outlook you have into next year. Help us understand how next year is shaping up.
Yeah. Listen, as we've gone through this year, we've moved our view of WFE up from $135 billion- $140 billion, now to low $150 billion, to your point. Basically what has happened is everybody in the industry, clean room is what's constraining the investment right now, right? The industry is fundamentally under-supplying demand, and clean room is a constraint. What I observe has happened is every one of our customers, or nearly every one of our customers, has figured out how to do a little bit more with what they had. Squeeze things a little bit closer together, perhaps. Frankly, we've done a lot of that, and we can talk about that, Atif, as well. Maybe pull some things forward, maybe buy some clean room that was there and make it production worthy a little bit quicker than anyone expected.
As the year has progressed, that's basically what has happened. However, the industry is still under-supplying demand, and that leads to our confidence in next year being a growth year. I think everybody that you listen to right now is communicating that. We say that because we see clean rooms coming online. We have described, Atif, eight to 10 new, what I would call tier one fabs coming online over the course of the next year, basically, or through the end of 2027. That reinforces our confidence in the fact that WFE is going to grow into next year. The constraint will still be there. However, it will get somewhat better.
When we look at that, when we do our own tops-down analysis, which Ram and his team do for the company, we continue to see AI demand remaining very strong, moving beyond training, into inferencing, into eventually Physical AI, and that gives us confidence also in the need to invest more. That is basically what is happening. It is a tops-down and bottoms-up assessment of what is going to be able to be supported next year. Frankly, I think the year after that probably continues to be even stronger than that.
All right. On that tops-down point that you made, Doug, a question for you, Ram. You guys introduced this kind of rule of thumb for every $100 billion of AI data center CapEx could translate to roughly $8 billion of WFE, which was super handy for the sell-siders. Recently, you suggested this number could now be $1 billion-$2 billion higher.
True.
What has driven this increase? Is it more silicon content or memory intensity, or can you explain that?
Yeah, it is a combination of both. We kind of looked at the timing of when we gave that number. It was the most pertinent thing, right? Everybody was talking about how many gigawatts of data centers are coming. We felt that was the most useful thing for you guys to think about the industry, and specifically how the tiering between leading-edge foundry logic DRAM and NAND was. So we gave that number. You are correct, since then, we upticked that number by a couple of billion. When we first came out, actually, people said, oh, you are being optimistic on that number, right? Then lo and behold, it has become even better than that. The uptick is all three. If you look at it, we used the initial flavors of the hardware that came out.
The large GPU maker had released some new versions of their latest GPU, and we used that as a benchmark on the industry pricing on capital intensity to come up with that number. Since then, things have inflected higher, both on the hardware content as well as on the intensity side. If you remember, we recently took up the NAND SAM per wafer from 1.8x that we gave at the analyst day to 2x. The complexity of these devices also is a little bit higher than what we originally included in that model. That was the gist of how we had taken that number up.
Yep. Great. Doug, you mentioned customer clean room space being the driver in terms of how much higher the WFE.
It is the constraining item, and I think will be through next year for sure.
All right. Any other supply restrictions around utilities or labor or your manufacturers capacity?
That's a great question. Listen, I think, maybe let me describe what's going on in the industry relative to conversations, just to help you frame everything. First, we are having very deep conversations with every one of our customers about, what do you need next quarter? What do you need next year? Give us an understanding of where you're going so that we can get ready. We absolutely are getting ready. The intensity of those conversations and the conviction of those conversations, I've never seen stronger, frankly, in all my years in the industry. The conversation intensity is really, really high. Everybody wants to make sure that they're going to get what they think they need. So that's happening between us and our customers.
Then we take that back and propagate that back to our supply chain with the same intensity to make sure that they're ready for what we're going to need. We're doing the same thing with our own bricks and mortar. We talked about the second facility in Malaysia. We're doing a bunch of things in existing facility footprints, densifying our ability to actually get more output from the same square footage of clean room that we previously had. Necessity is the mother of invention, and that's very much what's happening. We're trying to squeeze everything out. But we were fortunate in that we already had plans for a second facility in Malaysia, essentially a cookie cutter of the previous facility. So we're pulling hard on making that available sooner. We're hiring and training and focusing on making sure we're not going to be the bottleneck.
I would tell you, though, when you manufacture and make as many different tool types as we do that have thousands and thousands of parts in them, there's always something that is constraining us, and then you work. We've got an amazing supply chain organization that goes and works on things that pop up. It's very much manageable, but there's always something that's popping up. It's not easy, but we've got a great organization that knows how to do all of these things.
Very good.
That is very much what is happening. I think all of us also are doing our own tops-down modeling for where is this going. I am fortunate to have Ram and his team, great group of people that can do extraordinary modeling so we can make sure bottoms-up and tops-down generally makes sense, and it absolutely does.
Great. Let us talk about your products and markets. I had the experience to travel with Tim last week.
Yep.
And-
Thanks for taking him on the road. I think it was a really good set of meetings.
Yeah. For me, it was really interesting. Lam historically has been viewed as more of a memory-centric company with a lot more memory exposure, but the progress that you have made on the foundry logic side, particularly with the conductor etch tool, surprised me, based on
Akara, yeah.
Yeah.
The Akara tool is doing extraordinarily well. This is a great product.
I think the point that Tim was making was that historically, the cycles in spending have been more kind of Moore's Law driven. It is all about making incremental improvements in performance. But in this day of AI, you are getting paid. The incremental performance and latency and any kind of change you can make is super important to your customers.
It is.
That is driving the adoption of some of these newer products like Akara and conductor etch. That is what is different this cycle, that performance has become the bigger kind of-
AI compute is driving everything, and any incremental performance in the parallel compute architecture that you can squeeze out is hugely valuable. Frankly, greater tool performance contributes to greater silicon performance. We are fortunate over the last, I do not know, three to five years, Atif, we have increased R&D investment such that these new products that are coming online, like Akara, like Vantex, like ALTUS Halo, are hugely beneficial in that area. I think that is probably what Tim was communicating in your meetings last week.
Right.
If I may add one thing, if you look at our customers and the scale of AI, if they miss a product cycle or if they do not have the right products at the right time, the cost of missing that, given where industry spending is, I mean, in semiconductor industries demand is, then it is very high. Which means if an equipment company can help them with the velocity of the execution in totality, then that is a very good thing. So that is where you kind of deliver the value that our customers really kind of appreciate, products like Akara in the market.
Great. Maybe you guys can talk about, you talked about high- 30s SAM share of the market, and-
Yeah.
As we look into these products, particularly your momentum on the foundry logic side and maybe using some of these products in DRAM, how should we think about the upside to that SAM percentage over time?
Let me unpack it a little bit, and then I'll come back to your last question a little bit on why are we doing so well in foundry and logic. We had an investor day early last year where we talked about, hey, we see our SAM expanding from the low- 30s to the high 30% of WFE. Frankly, this year we're north of the mid-30s, probably north of 36%. So it's accelerated, right? We've moved forward a little more quickly, perhaps, than we had thought back then. And what's contributed? Part of it's been foundry and logic. It's also been an acceleration of different node migrations that have happened. Let me come back to your last question a little bit. Why are we winning in foundry and logic? Our SAM is expanding in foundry and logic because we're moving to gate-all-around.
Gate-all-around is more etch and deposition intensive. It's got selective etch requirements. It's got ALD requirements. Those are things we do, and things we do really well. It's got high aspect ratio etching. Akara, as an example, we're doing well there. Frankly, if you remember things we've talked about, we said, hey, for every 100,000 wafer starts to gate-all-around capacity, our SAM expands by $1 billion. Layer on top of that, when backside power eventually shows up, and it will relatively soon, that is another incremental $1 billion SAM expansion opportunity for every 100,000 starts. So that is showing up. On top of that, advanced packaging is critically important in the foundry space, as you know, Atif, right?
Our advanced packaging business is growing north of 70% this year, and a big contributor to that is what is happening in foundry and logic. Honestly, you've not yet seen the move to the panel form factor, which you're going to, and that will show up in foundry and logic. When you put all of these things together, it leads to the strength of the business and the performance that you've seen from us in foundry, Atif.
Yeah. Just on the panel-level packaging, how close is the industry to meaningful volume adoption, and what are the challenges around uniformity yield? Do you need to be looking at inorganic growth to expand the market?
We don't need inorganic growth, no. Listen, the things we do well at a wafer level, we will do well in panel. Read that to be the TSV approach, right, silicon etching and copper electroplating. We're going to be extraordinarily strong there in addition to other things that we do in advanced packaging at a wafer level. We'll continue to do well in a panel form factor. You may remember, we acquired a company a few years ago in Austria that actually did some of the panel form factor capability, and that's accelerated our R&D progress there. So we don't need anything new. We don't need anything incremental. It'll all be organic. We are very well situated for that transition when it happens. We are investing R&D right now to make sure we continue to be well positioned for this change.
Just to round up the discussion on the products, Tim spoke very highly on surface preparation, just getting-
Selective etch, yeah.
Selective etch adopted. Aether dry resist is something you guys have talked about in the past, $1.5 billion cumulative five-year revenue opportunity. Can you update us what's going on on the Aether dry resist program?
Yeah, no, thanks for asking about that. Listen, you probably saw some announcements about high-NA adoption over the last day or so from players in the industry. That's going to be beneficial for us. Listen, the pull for dry resist is really strong. I'm quite confident that billion and a half number is going to be higher than $1.5 billion. I'm not ready to give you a new number yet, but I'm highly confident we're going to do better than that. I say that based on things I hear from customers, things I see from customers, opportunity that's incremental to what we previously saw. We're going to upside that $1.5 billion. Pull is really, really good.
All right. Ram, going back to you, NAND, you do a lot of work in thinking around the drivers of the NAND market, KB cash, how's this market changing? You guys have laid out this $40 billion NAND upgrade opportunity historically through calendar 2027. Can you help us understand what's changing in the NAND market from a demand perspective, if you have an updated number, or how you're looking at the NAND opportunity?
Yeah, answering the latter part of your question, the $40 billion was given with the assumption on what it takes to go from sub 200 layers to over 200 layers. That's a fixed number, right? What has changed is not the number per se, but the rate at which we originally said that number, and people assumed, hey, we said several years, and there was a range between three to five years.
Since then, we have said it'll be done by end of 2027, which is compressing it by three years. So the pace at which the change happened was the one that is the newer, not the amount, because that's a fixed amount, right? How much you need to upgrade to the next layer. NAND, if you take a step back in the first part of your question in terms of demand, clearly, when we came out a little bit more optimistic on the NAND at the Analyst Day, people were questioning. Now it's the other way around.
Yes.
People do understand, but then there is the supply and demand balance that people want to understand. Hey, can NAND do conversions predominantly versus when do they need capacity? Look, from historically how capacity has been added, if you take that as a baseline, still 2026 and 2027 is conversion driven. There's a little bit of greenfield capacity in any given year that comes, and that is the case for 2026 and 2027. But for majority, it's still conversion driven for NAND.
If you tie that in with customer new fab announcement, we do think that beyond 2027, if demand were to sustain, which we don't see a reason why it should derail, given where AI is going and memory hierarchy is going, that you probably need some capacity addition, say maybe like second half of 2028 and beyond, right? It times with the fab announcement made by some of our leading customers.
Yeah, I'll just add on. When you look from peak capacity to where NAND is going to be at the end of this year, raw wafer starts are probably coming down 20%, Atif. You got to supplement that at some point. There are some wafer capacity additions this year and into next year. It's just not that much.
Got it. The question is around the High Bandwidth Flash of the new generation NAND. There is a lot of buzz around beyond conventional enterprise SSD, emerging high performance NAND architectures, HBF. Can you just talk about from your vantage point, are you seeing any momentum in these technologies getting some sort of volume adoption or making a breakthrough, or this more of a longer-term roadmap?
Listen, you are hearing a lot of our customers talk about this. People are working on it. You are not seeing any volume yet. But to the extent that this does ultimately show up, this is going to be great for us. You know the strength that we already have in NAND. You know the strength we have through-silicon vias. This combines both together. So if this does actually begin to show up as part of the memory hierarchy in a more meaningful way, this is going to be great for us.
All right. Moving on to a very topical question on gross margins. Doug, I almost feel guilty asking you this question because you have done so well in expanding gross margins to where they are, but my clients are greedy. They want more.
Everybody wants more. What have you done for me lately? But if you will allow me, I will point out, we just printed 52% gross margin. That was the highest gross margin in 20 years at Lam Research. So I feel great about how we are managing things. It has been a combination of new product introduction. It has added to the close to customer strategy, right? The factories being closer to where customers are. There has been an aspect of pricing in it. So we are working on everything, pulling all the levers that we can.
All right. Is there a framework on the pricing?
Yeah, I was just going to add, maybe I'm jumping ahead on your question. We also, on the last earnings call, talked about a new objective relative to the profitability target for the company. We just delivered 52, we guided 52, and then communicated, hey, we think we see a path to get to the mid-50s gross margin, continuing to exercise all those things that I just mentioned. It'll be an extension of the close to customer strategy, ramping that second factory in Malaysia, benefits from just the supply chain being close to us, that helps. New product introduction.
When you have a product like Akara coming out, you tend to be able to get paid for things when you're delivering really good results, performance on the wafer, and we're doing that. And then we're working on pricing where we can. All of those things contribute to what we're trying to deliver. Frankly, I feel great about what we've been able to do, and we see continued opportunity as we go forward.
Good. That's helpful. Doug, let's talk about the CSPG business. With more than 100,000 installed chambers, how is CSPG opportunity evolving, as tools become more complex and customers put greater value on uptime yield and productivity?
Yeah, let me unpack CSPG a little bit relative to how you should be thinking about things going forward. First, the last quarter that we delivered, third consecutive record, nearly $2.5 billion in revenue from CSPG. It's doing great. We're going to benefit from the fact that WFE is pretty strong this year, so chamber count will grow nicely this year, which will provide an incremental opportunity to do more. That's one thing to think about. Spare parts are the biggest individual component in CSPG. We're benefiting right now from the fact that utilization in the industry is basically at 100%, so spare parts consumption is high. That's going to continue, for sure. Upgrades are very strong. It's got a heavy footprint in what's going on in NAND, right?
You saw our NAND business last quarter was very strong, so upgrades are doing well and will continue to. The Reliant product line is doing pretty well. That is the equipment that goes into more mature node spending. Even though everything you're hearing us talk about right now is focused on the leading edge, the mature node investment is actually pretty strong. When you look at what is happening in analog and industrial and so forth, you're seeing strengthening there. Then layer on top of that the things we're doing strategically in service. We are beginning to offer to customers different kinds of service offerings using cobots and Equipment Intelligence and different algorithmic data techniques to change how we deliver service for our customers. This is exciting to the customer because you're delivering incremental performance capability. When you look at all those things, it moves the service opportunity.
This is all incremental to the historic way we have delivered service, which has been show up and do a task. Provide some maintenance to the equipment. All of that is still here. What the cobots and Equipment Intelligence enable us to do is go in, look at the fleet of the customer's tools, and identify different opportunities that we can do to make things better. When output is at such a high value right now, the pull for that is very strong. It changes the opportunity here to deliver service in a different way by guaranteeing some outcome. The customers really like that. When you go in, tell them you can do something, do a proof of concept, and deliver it, then the opportunity continues to grow. That's what we're quite excited about in the service area of CSPG. That's got a lot of pull right now.
Doug, on the cobot, Tim was super excited about being an early adopter on Dextro cobots.
Yes.
It would help to understand what is different about your tools or products, or you just being an early adopter to enable better serviceability and uptime for your customers by using cobots. Is it something different about your products that you guys are doing it more so than your peers?
Listen, I think we saw this before others saw it, and I think right now, because the customer has seen such value from us, they're going to force our competitors to do this as well. We're meaningfully ahead of where anybody else is with this. We identified the opportunity to do this way before anybody else did. We brought product offering way before anybody else even thought about doing this.
As always in this industry, when you do something better than your competition, your customer then goes to your competition, says, you need to do this as well. I think that's going to happen. We're quite a ways ahead of this. We're investing R&D. We've expanded from a single tool type to I think eight now, and it's growing beyond that. You're going to see us continue to provide real leadership, I think, in the industry and the space.
Good. Let me pause here and see if there are any questions in the audience. If you have a question, please raise your hand.
Yep, we got one up front here. Can we get a mic up here?
Possibly de-speccing of DRAM. Just wanted to understand on how that will impact you or- In terms of how and if you're hearing that trend, how long would that be going for?
Yeah, listen, it hasn't changed anything our customers are doing. Frankly, I've been getting this question in many say, this is a red herring. It's got really no impact that I can see. Ram, I don't know. You've done more work on this than I have.
Yeah. Look, ultimately, there is a requirement for the model, right? The models require a certain amount of compute power, and the way we look at it from our vantage point of view is the hardware totality of requirements as we look from the transition to Agentic AI to Physical AI over the next multi-year period. The compute per gigawatt, which is the most important metric for generating the right amount of tokens for these models, is very hardware favorable in totality. Equipment companies in general are very well-pegged to that trend. We see this as just a short-term noise just because of various other things that are going on. We don't see a direct impact to our business.
Good question, though. Yep, we got another one up here.
Doug, Micron, I was at a conference in August, has made this statement that 2027 would be even tighter than 20-. Sorry, yeah, 2027 would be tighter than 2026. Is that an industry-wide belief? I look at some of the industry forecasts for CapEx, for DRAM, $60 billion last year, growing to over $100 billion this year. What does it take for DRAM memory in general to get into supply-demand balance?
Yeah. We've got a ways to go to get there. You can just look at pricing and profitability. I would never disagree with any of our customers. If Micron said that's an accurate representation of what's happening. Likely an industry-wide representation. Yeah, the industry needs more clean room space, and that's beginning to come online, but it just takes time, right? You can't snap your fingers and have it show up. It takes a couple of years to bring new clean room online. So that's what's constraining things right now.
We made a comment earlier about 2028 looking like another strong year. That's something that comes up as 2028 is there's some level of uncertainty maybe because it's still a year and a half away. But what makes you optimistic at this point about 2028?
Just our top-down modeling, the bottoms-up conversations with customers about where things are heading, and our view of the timing of clean room showing up. Again, I'm not going to put my neck out quite yet on 2028. It's still a long ways away. Lots can change. But the strength of AI demand continuing, the fact that you hear comments from some of our customers, like you just mentioned, that things are still going to be undersupplied in 2027, leads you to think 2028 is probably a pretty decent year as well. Yep. We got one in the middle here.
Thank you. Just following up on those same questions. I think a lot of the memory players have talked about 20%-25% capacity growth year- on- year within DRAM, for example. They're doing that, and the constraint on that is that one of the constraints is the clean rooms, as you've been saying.
Yep.
You mentioned earlier about your supply chain, though, as an example, and that even when clean rooms come online, that does not solve all the problems automatically because Lam supply chain and your suppliers would also need to increase that.
Yes, for sure.
I am wondering if you can just elaborate on that a little bit more. If clean rooms did magically sort themselves out within a year or 18 months longer, what are the challenges for Lam, or more generally within Semi Cap, in order to be able to meet that new demand from the memory players, if they were to try and grow capacity 600,000 wafers per month or something like that, more than the 400 that they are talking about?
Yeah. Listen, we have a pretty good understanding of clean rooms that are showing up when they are showing up because they cannot just magically come from nowhere. Maybe what would help is I will describe the conversations that are happening to make sure it is pretty clear. There is someone at Lam whose job it is, and in fact, teams of people, to know what every single customer plans to do over the next year or two years and beyond, right? We have got account teams. That is their job.
When I listen to them describe to me what is going on, the conviction of those conversations has never been higher, right? All of our customers have a huge motivation to make sure we are able to get them what they need. I would say that is where everything starts. Every single customer, what do you think you are going to need next year?
What's the point estimate? What's the upside? Tell me when. What's the error bar around it? We then take that and make sure we're going to be ready with the same thing. The good news from our point of view is, generally speaking, our lead time to get ready is shorter than our customers' is because we can just do things quicker. We're not building as big a clean room and so forth. We were fortunate in that we had another facility coming online in Malaysia. I've talked about that. So we're getting that ready. The lead time we have there is to facilitize it, hire, and train people that can actually build product. Then we take all of this information that's coming back from our customers to us, then go back to our supply chain.
To make sure they're going to be prepared for it as well. I think the intensity of all of those conversations, because none of us want to be the constraining item in the industry. That's not a great place to be. That conversation is happening with an intensity that I can't remember in the past. I think all of us, if you go all the way back to our supply chain, even maybe a layer or two behind that, everybody sees this AI demand. It's obvious that it's here. Everybody just needs to know to what magnitude. So those conversations are quite robust right now to make sure we're all getting prepared for what everybody needs. Did that help?
Yeah.
In Malaysia, too, you were going to say. Malaysia phase two.
Yeah. I talked about, we've got another facility in Malaysia that's coming online, and some of the, like I said, necessity is the mother of invention in this industry. We're going through a process of trying to get more output from the existing square footage we already have. We're densifying things. We're getting more output from the same square footage. You might say, hey, why weren't you always doing that? Well, you didn't know you needed to until you needed to. So there's a lot of creative things we're doing with our own capability to try to squeeze a little bit more out. Frankly, I think my customers are doing the same. Our customers are doing the same thing.
Questions?
Doug, last one from me. In terms of your China sales exposure, you guys are a bit unique that you do have a bit higher international fab exposure in China versus peers because of your NAND market share.
And DRAM.
And DRAM, too. Fundamentally, if the non-China market is growing faster than the China, should we be thinking about your China sales exposure roughly stable around these levels this year, next year?
Listen, I think this is kind of a numerator, denominator thing. The China investment continues. It is not like it is gone away. However, when you look at where all of the growth is showing up, it is with the leading edge customer base. It is DRAM, it is NAND, it is going into AI compute from the leading foundry and others. That is where that growth is really happening over and above where it was last year.
It is not that China has gone away. It absolutely has not. But everything else is just growing faster. I think you are going to probably see over the next several years, a continuation of those trends. As a result, I think for everybody in the industry as well as for us, you will see China as a percent of overall revenues declining. Not because China's going away, it absolutely is not, but everything is just growing much faster.
There will be quarter- to- quarter variability, Atif. It's not going to monotonically decline, right? It's always lumpy.
Yeah.
Yeah.
Awesome. We're almost out of time. Doug and Ram, thank you for coming to Citi Conference.
Of course. Thanks for having us, Atif. I appreciate it, and thanks for taking time.
Yeah.
On the road last week.
Yeah, thank you.