Lam Research Corporation

Good day, and welcome to the LAM Research Corporation September Quarter 2025 Earnings Conference Call. All participants will be in a listen-only mode. Should you need assistance, please signal a conference specialist by pressing the star key followed by zero. After today's presentation, there will be an opportunity to ask questions. To ask a question, you may press star, then one on your telephone keypad. To withdraw your question, please press star, then two. Please note this event is being recorded. I would now like to turn the conference over to Ram Ganesh of Investor Relations. Please go ahead.

Thank you, and good afternoon, everyone. Welcome to the LAM Research Quarterly Earnings Conference call. With me today are Tim Archer, President and CEO, and Deb Bittinger, Executive Vice President and Chief Financial Officer. During today's call, we will share our overview on the business environment, and we'll review our financial results for the September 2025 quarter, from your outlook for the December 2025 quarter. The press release detailing our financial results was distributed a little after 1 p.m. Pacific time. The release can also be found on the investor relations section of the company's website, along with the presentation slides that accompany today's call. Today's presentation and Q&A include forward-looking statements that are subject to risks and uncertainties reflected in the risk factors disclosed in our SEC public filings. Please see accompanying slides in the presentation for additional information. Today's discussion of a financial resource will be presented on a non-GAAP financial basis unless otherwise specified. A detailed reconciliation between GAAP and non-GAAP results can be found in the accompanying slides in the presentation. This call is scheduled to last until 3 p.m. Pacific time. A replay of this call will be made available later this afternoon on our website. And with that, I'll hand the call over to Tim. Thanks, Ram, and good afternoon to everyone on the call.

LAM delivered a solid September quarter, highlighted by record revenues of $5.3 billion, gross margin of 50.6%, and record operating margin of 35%. We also achieved record combined spares and services revenue, and growth in total CSBG revenue outpaced the increase in install-based units. Inclusive of our guidance for the December quarter, we expect to close calendar 2025 with three consecutive quarters of greater than $5 billion in revenue. Our performance reflects strong company-wide execution and the critical role that our products and services portfolio plays in enabling the industry's technology roadmap and in addressing the rapid increase in semiconductor manufacturing complexity. Our December quarter guidance does contemplate roughly a $200 million revenue impact from the recently announced 50% affiliate rule, restricting shipments to certain domestic China customers. Currently, we expect this rule to impact our calendar year 2026 revenues by approximately $600 million. This impact, together with the strong growth anticipated in worldwide fabrication equipment or WFP spending, leads us to expect the China region to represent less than 30% of our overall revenues in calendar year 2026. Turning to WFE, spending in calendar year 2025 is shaping up to be slightly better than our prior view of $105 billion, predominantly due to better than expected high bandwidth memory or HBM related investments. As we look ahead, we see a robust setup for equipment spending in calendar year 2026. AI related demand should support sustained strength in leading-edge Foundry logic and DRAM, as well as continued manned upgrade spending. The strong leading-edge growth we see across all three device segments is forecasted to be partially offset by a decline in domestic China-related investments. We plan to provide our detailed 2026 WFE spending outlook and sub-segment color on our January call for our usual practice. AI and its impact on the semiconductor industry continues to be a major topic of interest, and the data center CapEx investments already announced are expected to drive significant expansion of manufacturing capacity over a multi-year period. In recent months, we have seen an acceleration of activity. AI data centers require the most advanced CPU and accelerator capabilities, low latency, high bandwidth memory, and high speed ESSD storage. all integrated through 2.5D and 3D advanced packaging. We estimate these needs translate to roughly $8 billion of WFP spending for every $100 billion in incremental data center investment. Most importantly, deposition and etch, the area of LAM's core product differentiation, play an increasingly critical role in enabling the higher performance, more scalable semiconductor devices required for AI. We see the surge in AI data center demand creating billions of dollars of served available market expansion and share gain opportunity for LAM in the coming years. I will share a few areas of strength we are seeing. In NAND, customers are continuing to upgrade existing FABs to meet the need for higher layer count, higher performance devices. We have estimated that these conversions will require $40 billion of WFE spending over the next several years. And as we have previously said, LAM should capture a high percentage of this conversion spend due to our large install-based position. Our upgrades business is projected to remain strong into 2026, as NAND bit demand looks to be trending higher than prior expectations. Device makers have already announced enterprise-grade SSDs with 256 terabytes of storage capacity to satisfy growing data center demand for high-capacity storage. Availability of clean room space will likely act as a limiter to the pace of NAND supply growth, but we believe that capacity additions to meet rising BIT demand may be needed sooner than previously thought. LAND is in a great position for both upgrade activity in the near term and new capacity builds in the future. We have the industry's largest installed base of NAND systems, and our comprehensive NAND product portfolio features several industry-first advances. Notably, LAM recently earned the 2025 SEMI Award for our pioneering LAM Cryo 3.0 dielectric edge technology, a process that has quickly become the industry standard for advanced NAND devices. We're also seeing solid demand for our atomic layer deposition, or ALD products, including a recent key win at a major NAND manufacturer for a critical high aspect ratio dielectric deposition application. LAM's differentiated conformal fill capability using a higher temperature process was fundamental in securing this win. On the metal side, LAM's Halo MOLLE ALD tool has been selected as the tool of record for three consecutive nodes at a leading customer, including for devices with more than 500 layers. This further reinforces our leadership in the 3D NAND word line application, a step that is fundamental to building the higher performance devices required for ESSDs. The performance demands of AI devices are also spurring investment in foundry logic and DRAM manufacturing inflections. Over the last several years, we have focused on expanding our product portfolio to target these opportunities and believe we will benefit as the technology transitions unfold. For example, LAM's Ether Dry Resist EUV patterning solution has demonstrated the ability to resolve features of less than 15 nanometers at the highest density and pattern fidelity. Ether also enables a more than 10% reduction in EUV exposure dose, boosting scanner productivity and reducing the cost of patterning per wafer. LAMP's Ether technology is already ramping in the HBM high-volume production line of a major memory manufacturer, and we see more opportunities ahead as we look further out on the road map. For instance, we believe high NA EUV in combination with ether for single patterning of sub 10 nanometer features will be critical to addressing the complexity and cost challenges associated with the transition from gate all around transistors to CFET in Foundry logic, as well as the anticipated migration from 6F squared to 4F squared in DRAM. In September, we announced a key partnership with JSR Corporation an innovative semiconductor materials company, to collaborate on the integration of our ether technology with novel EUV patterning materials and metal oxide resists. In addition, LAM and JSR are partnering to explore new precursor materials for advanced ALD applications, which we believe can further enhance our capabilities and differentiation for future technology inflections. In the case of low K ALD films, LAM's high productivity single wafer solution are enabling our customers to move past traditional furnace-based approaches. As logic transistor sizes scale down to achieve greater compute power, higher capacitive coupling in the gate module degrades overall performance. Similarly, as DRAM devices shrink, there is a detrimental increase in capacitance between the bit line and capacitor contact. To resolve these issues, deposited low-K films must be very thin, 5 nanometers or less. and conformal in high aspect ratio structures. Furnace-based films at these thicknesses are often fragile and unable to withstand the harsh chemistries used in subsequent process steps. LAM's Low-K ALD solution employs a unique single wafer remote plasma reactor and a novel precursor to deposit thin, defect-free films with the desired silicon-carbon bonding structure. As a result, LAM's ALB films have demonstrated superior durability on the remainder of the chip-making process. And we recently secured critical wins at Foundry Logic and DRAM customers for low-K applications using this process. Beyond traditional device inflections, LAM is also benefiting from healthy growth in advanced packaging. Our Sabre 3D plating and Cindi and Edge systems are industry leaders and should continue to see strong demand in 2026 as AI-related spending grows. Looking further ahead, we are investing in new, advanced packaging opportunities. Today's packaging production lines primarily use 300-millimeter diameter wafers. But as AI and high-performance computing demand larger chips to integrate more accelerators, memory, and interconnects, panel-level packaging is emerging as a scalable solution. By processing multiple units on larger format panels, It significantly improves manufacturing efficiency and supports the integration of increasingly complex and larger semiconductor devices. LAM's Sabre 3D, Callisto, and Phoenix tools are being engineered to meet future panel packaging needs. We are collaborating across the ecosystem to drive industry-wide standardization and co-development, both of which are essential for scaling high-volume manufacturing and next-generation integration solutions. We expect to end this year with tools shipped to or installed at 20 customers worldwide. Our growing installed base of panel packaging tools is rapidly building experience and maturity that should prove valuable as this technology becomes mainstream in the future. So to wrap up, LAM is poised to close out a record calendar year 2025, and our setup is strong heading into 2026, where we expect solid WFE growth. The technology requirements of AI play extremely well to LAM's product strengths, and we are excited by the breadth of opportunities we see ahead for the company. Now, here's Doug.

Thank you, Jim. Good afternoon, everyone, and thank you for joining our call today during what I know is a busy earnings season. We executed well in the September 25 quarter, delivering gross margin performance of 50.6%, which is a record in the post-Novellus period. Financial results for the quarter came in above the midpoint of all of our guidance ranges. We also delivered many financial records throughout the P&L. The company truly performed well in the quarter. Let's turn to the details of our September quarter results. Revenue for the September quarter came in at an all-time record of $5.3 billion, which is up 3% from the June quarter. The referred revenue balance at quarter end was $2.77 billion, up slightly from the June quarter due to increases in services and system-related transactions where revenue recognition was not yet complete. This was partially offset by approximately $100 million of reduction in customer advance down payments. We do expect to see these down payments continue to decline in the December quarter. From a market segment perspective, Foundry accounted for 60% of our systems revenue in the September quarter, up from 52% in the June quarter. This marks our third consecutive record quarter, underscoring the strength of our strategic focus and execution in Foundry. Foundry's strength came from investments at the leading edge, in addition to mature node spending in China. Memory was 34% of systems revenue, which was down from 41% in the prior quarter due to the timing of customer investment plans. Within memory, non-volatile memory contributed 18% of our systems revenue, which was down from 27% in the June quarter. The trajectory of the NAND spending this year is broadly consistent with our expectations coming into the year. And as the industry transitions to devices above 200 layers, we continue to estimate over $40 billion in upgrade spending will be required over the next several years. DRAM increased from the June quarter, accounting for 16% of systems revenue compared to 14%. Investments in high bandwidth memory continue to remain strong, driven by AI-related customer demand. We're also seeing traditional node migrations to the 1B and 1C nodes, enabling the transition to DDR5. The logic and other segment came in at 6% of systems revenue in the September quarter, roughly in line with the 7% we reported in the June quarter. Let's turn to the regional breakdown of our total revenue. China came in at 43%, an increase from the prior quarter level of 35%. While the multinationals in China remained steady, the domestic Chinese customers grew, and the majority of our China revenue continued to come from them. The next largest geographic concentrations were Taiwan at 19%, which was flat sequentially, and Korea at 15%, down sequentially from 22%, again, due to the timing of customer investment plans. The customer support business group generated approximately $1.8 billion in revenue for the September quarter, slightly higher sequentially and year-over-year. This is being driven by continued strength in spares and upgrades. CSBG remains a key part of our growth strategy, given the expanding installed base and our innovation in advanced services. We expect CSBG to deliver year-over-year growth in 2025. And I just mentioned that in the 13 years since we brought Lamb and Novellis together, CSBG has grown every year except for one. Let's look at profitability. Gross margin of the September quarter was 50.6% at the higher end of our guided range and improving from the June quarter level of 50.3%. The increase is primarily driven by favorable customer mix, partially offset by the impact of tariffs. I expect the impact from tariffs to continue to increase somewhat in this December quarter. Operating expenses for September were $832 million, which was up from the prior quarter level of $822 million. The increase is primarily due to increased headcount and incentive compensation, which is tied to the company's improved profitability. R&D accounted for 68% of the total operating expenses. We're investing in innovations like Vantex, Acara, Halo, and Dextro to continue our leadership in providing a differentiated product portfolio for our customers. The September quarter operating margin was 35% at the high end of our guidance. This operating profit represents a record level for LAM in both dollars as well as percentage terms. The non-GAAP tax rate for the quarter came in at 14.2%, generally in line with our expectations. We continue to see the tax rate in the low to mid-teens for the near term. We do expect, however, with the increase in the GILTI rate in the United States, as well as the advent of the global minimum tax regime outside of the United States, you will see a slight increase in our effective tax rate as we get into calendar year 2026. Other income expects for the September quarter was approximately $8 million in income, compared with $4 million in income in the June quarter. The slight increase in OINE was primarily the result of increased interest income tied to a higher cash balance. As we've talked about in the past, you should expect to see variability in OINE quarter to quarter. Let's look at capital returns. In the September quarter, we allocated approximately $990 million to share buybacks through open market share repurchases. Our average buyback price in the quarter was approximately $106 per share. Year to date, we've repurchased nearly 30 million shares at an average price of a little more than $88 per share. We also paid $292 million in dividends in the quarter. I'll remind you that we increased the dividend from 23 cents to 26 cents per share earlier this month. Moving forward, we remain committed to returning at least 85% of free cash flow to our shareholders over time. The September quarter diluted earnings per share were $1.26 above the midpoint of our range. The diluted share count was 1.27 billion shares, which was a reduction from the June quarter and consistent with our guidance. We have $6.5 billion remaining on our board authorized share repurchase plan. Let me pivot to the balance sheet. Cash and cash equivalents totaled $6.7 billion at the end of the September quarter, an increase from $6.4 billion at the end of the June quarter. The main reason for the cash increase was cash generated from operating activities, which was partially offset by cash allocated to capital return, as well as capital expenditures. The sales outstanding was 62 days in the September quarter, which was up slightly from 59 days in the June quarter. September quarter inventory terms improved to 2.6 times compared with 2.4 times in the prior quarter and up from the levels two years ago of 1.5 times. We've remained focused on driving asset utilization during this timeframe and I was pleased to see us deliver this outcome. Our non-cash expenses for the September quarter included approximately $97 million for equity compensation, $89 million for depreciation, and $13 million for amortization. Capital expenditures in the September quarter were $185 million, which was up $13 million from the June quarter. Spending was primarily focused on lab investments in the United States, along with expansion of manufacturing sites in Asia. This remains consistent with our global strategy to be close to our customers' development and manufacturing locations. We ended the September quarter with approximately 19,400 regular full-time employees, which was an increase of approximately 400 people from the prior quarter. Headcount increases were in R&D to support our long-term product roadmap. Additionally, we've had increases within the field organization to support customer growth and a higher volume of tool installations. Let's turn to our non-GAAP guidance for the December 2025 quarter. We're expecting revenue of $5.2 billion, plus or minus $300 million. We expect a decline in China revenue, offset by stronger spending from the global multinationals. We're expecting gross margin of 48.5% plus or minus one percentage point. I expect customer mix and tariffs will be contributing to the sequential decline in gross margin. We're expecting operating margins of 33% plus or minus one percentage point. And finally, earnings per share of $1.15 plus or minus 10 cents based on a share count of approximately 1.26 billion shares. I want to give you a few things to think about as you build your 2026 models. While we are not yet quantifying the level of growth in WFE, I will tell you that calendar 26 looks somewhat second half weighted as we sit here today. The newly restricted China entities would have been weighted to the first half of next year. Customer mix will be a headwind to gross margin a bit next year as the China mix normalizes. And the tax rate will likely pick up very slightly, as I previously mentioned. We'll give you a better color on all this during the December quarter call. So let me wrap up. LAM delivered another strong quarter, highlighted by record levels of revenue, record gross and operating profit. Inclusive of our December quarter guidance, we're on track to deliver calendar year 2025 at an all-time high watermark in financial performance. closing with three consecutive quarters of revenue above $5 billion. We remain focused on strategic investments that extend our technology leadership, operational efficiencies, and long-term value creation. Operator, that concludes our preferred remarks. Tim and I would now like to open up the call for questions.

Thank you, sir. Ladies and gentlemen, if you would like to ask a question, please press star then 1 on your telephone keypad. If you are using a speakerphone, please pick up your handset before pressing the star keys. If at any time your question has been addressed and you would like to withdraw your question, please press star then 2. At this time, we'll pause momentarily to assemble our roster. And the first question comes from the line of CJ Muse with Cantor Fitzgerald. Please proceed.

Yeah, good afternoon. Thank you for taking the question. I guess first question, very helpful guideposts for thinking through incremental WFE tied to AI infrastructure spending. But I guess over the last six, eight weeks, would love to hear how your conversations with customers have progressed. Are you seeing expedited meetings? Are you seeing actual orders? Would love to see how the announcements around infrastructure spending is translating into visibility on your business?

Sure, CJ. Let me take that. I mean, obviously when we talk about announcements that are made recently, there's not physical space. There's not near-term demand for those. Those are things that, you know, give you guideposts as to where demand is going, you know, further in the future. I think to look at the conversations that we're having today about near-term needs for equipment, it's a lot of the things I talked about, which is, you know, you know, enterprise SSDs and the impact on NAND. You've heard some of our customers, I believe, speak to BIT demand that might be a little bit higher than expectations. I just said that our view of NAND upgrades is well on track for a good 25 and a strong 26. So it's really down. When we talk about our equipment demands, it's near-term needs with those longer-term announcements as opportunity in the future. But they're along the same technology transitions. Those data center investments are going to require faster GPUs made at smaller nodes for Foundry Logic. They're going to be made with higher capability HBM. And that's where all of our products come into play. How we participate in gate all around, our ALD tools, high aspect ratio conductor edge on the DRAM side. the work we're doing in HBM to enable higher stacking of HBM devices. And so, again, we're seeing very robust demand, as we mentioned, going into 2026, but it's for things that are real and here today.

Very helpful. And then maybe taking through your relative outperformance to WFE, you know, based on your guide, it looks like you're tracking on a tool shipment basis up 40%. And, you know, I think many investors think we're growing 10% overall. So tremendous outperformance. I guess, how are you thinking about 2026? And as part of that, what would be the critical drivers to drive relevant outperformance? Thanks so much.

Yeah, sure. I'll let Doug add in here as well. But I guess I would just say that you have to remember that as fabs get built and equipment is brought in, there often are timing issues. And so it's a little bit hard to speak to any near-term outperformance, underperformance, but certainly because it depends on what other suppliers are going to be doing and when they're shipping and what their lead times are. But what we can say with quite a bit of confidence, and I think we laid it out on our investor day earlier this year, is that Over the longer term, LAMS markets, etch and deposition, will outgrow WFE because of nearly every trend that's taking place technology-wise in semiconductor manufacturing, whether that's 3D devices in Foundry Logic, in NAND, whether it's smaller, higher aspect ratio devices in stacking using advanced packaging in DRAM, whether it's advanced packaging itself. They're all deposition and etch-intensive products, and therefore, I think over the longer term, confidence to outperform WC is very high.

Nothing to add, Tim. You nailed it. Thanks, CJ.

The next question comes from the line of Tim Arcuri with UBS. Please proceed.

Thanks a lot. Doug, when we talked three months ago, you were thinking that December would be like 4-7. You were saying it'd be sort of back to where March was. and now it's coming in at 5.2. So the incremental $400 to $500 million, where did that come from? It sounds like maybe a little bit of it pulled in from the first half of next year. Can you just sort of give us a sense, like, what's actually better than what you saw three or so months ago?

Yeah, listen, I think Tim pointed to the fact that WFE is a little bit stronger. We think high bandwidth memory and DRAM is a little bit stronger. Maybe everything else was just a little bit stronger, Tim. it's not any one thing that I would point to. I would tell you, though, honestly, it would have been even higher if not for the restricted entities in China. So things did strengthen for sure. You know, as we look into next year, clearly next year is a growth year. And whether we pulled anything in from the first half, I honestly don't think so, Tim, because as we sit here today, I think the first half of next year is flat to maybe slightly up from the second half of this year. So it's going to continue to be pretty good.

Got it. Great. Thanks, Doug. And then on the 2026 WFE in China, I know you and everyone else has the same message that it's going to be down. But to be honest, that's what everybody said at this time last year, too. And you're growing like 20% this year. I know that there's some others that are more close to flat, but I mean, you're up a ton and virtually everyone else is up a ton, too. So it seems like we keep thinking that China will digest and that it'll be down, but they keep finding ways around these bans. So Like, why would China be down next year? I guess I'm just trying to figure out, is there something different next year that you're seeing that maybe gives you the confidence that finally China is going to be down because it hasn't happened yet? Thanks.

I guess what I'd say, Tim, it's twofold. First, the global multinationals outside of China are going to be pretty strong next year. So that's part of it, right? Everything else is going to be stronger in 26, I think, than it was in 25. And honestly, as we sit here right now and look at the stackup of everybody's plans in China... it's going to be less. That's the best I can tell you. And, yep, you're absolutely right. A year ago when we were sitting here, we would have seen the same thing and then it strengthened through the year. I just right now don't see where that's going to come from next year, Tim.

Okay, Doug.

Thank you. Thanks, Tim. The next question comes from the line of Vivek Arya with Bank of America Securities. Please proceed.

Thanks for taking my questions. For the first one, Tim, you gave this interesting statistic, $8 billion of WFE for, I think, every $100 billion in data center. How much of that $100 billion of data center spend is in semiconductors? Basically, what is the WFE intensity in an AI data center versus kind of the mid-teens, you know, WFE intensity for all semiconductors? And then off that $8 billion, what is LAM's opportunity?

Okay, well, I guess just to clarify, the $8 billion was WFE or $100 billion of data center investment. So that would represent the equipment portion that we could target. And again, given that data centers and especially those focused on AI applications require leading edge across all three device segments, Again, that's an area where SAM as percent of WFE continues to increase at every technology node. And so all the things I'm talking about, whether it's high aspect ratio etches, it's ether, dry UV resist, it's ALV, all of those are growing our opportunity in that $8 billion. And so that's where a lot of our focus is these days, is the products that are required to do well through that spend.

Vivek, was that your question? Did that answer the question?

I'm not sure we got it. My question is different. You know, because if you look at WFE overall as a percentage of semiconductors, it's about mid-teens percent. And when we talk about $100 billion in data center spend, there's a lot of non-semiconductor spend as part of that. So my question is off that $100 billion, how much is semiconductor spend? And what does that imply for WFE intensity in an AI environment? And of that $8 billion, how much is LAM's opportunity?

Yeah, listen, Vivek, if I'm honest, I don't know the precise answer to your first question, how much is semiconductor content as part of that $100 billion. It's a decent amount. It's GPUs, it's HPM, it's enterprise SSDs. I don't know the precise number, but I know it's a decent amount. And then relative to our intensity in all these things, it's very similar to what you've seen from us across the rest of semis, which is you've got the move to gate all around. You've got high bandwidth memory. You've got a growing stack in NAND. It's part of the contribution of our share of WFE going from the low 30s to the high 30s. This would be representative of it to the best of my ability to answer your question.

Yeah, I think that would be the best way to think about it is at the investor day, we said that as you move to these leading edge nodes, LAM, SAM, as percent of WFE, our opportunity would grow from the low 30s to the high 30s through those transitions. So assuming these are at the leading edge, then you're starting to create an opportunity that's at that high 30s level.

Okay. For my follow-up, I think in the past you have mentioned the $40 billion TAM for NAND upgrades. How much of that will be completed by the end of the year? Like, will a third be completed? You know, will half be completed? Just any rough sense of where we are in the journey of that conversion. And as you look at next year, I know you're not giving a specific WFE view, but what would, you know, cause NAND growth to be different, you know, higher or lower than what you saw and what we have seen so far in 2025?

Well, here's what we said. So when we were at the investor day back in February, we said that $40 billion would be spent. And obviously, we didn't exactly put a date on it. We said over several years to satisfy the upgrade of the installed base to the 200-plus layer level. What I said in my remarks today is that with demand being a little bit higher than prior expectations, we've likely seen a little bit of an acceleration of that upgrade. I also said that in 2023, our upgrade business in NAMD would remain strong. We're not going to tell you exactly how far we are through that several-year period, but compared to February, it's accelerated. And as I also mentioned, I think that if this demand for high-capacity storage continues, the $40 billion when we gave that in February targeted kind of a mid-high-teens BIT demand. And I think that what you're hearing publicly right now is maybe demand that's a little bit higher than that. So that would suggest it's being accelerated. And all we can speak to is our demand would suggest the same thing.

Thank you. Thanks, Vivek.

The next question comes from the line of Harlan Seur with J.P. Morgan. Please proceed.

Hey, good afternoon. Thanks for taking my question. Maybe as a follow-up to CJ's question, you know, given all of these data centers, infrastructure announcements, you know, I think, for example, Sam Altman's recent trip to Asia, he had a view that, you know, DRAM and advanced foundry supply requirements over the next several years would be X amount, right, which positively surprised all of us. And it actually does imply significantly more capacity requirements across advanced foundry, memory, and advanced packaging. you would think that your customers would want to start to put this in place as quickly as possible, maybe starting next year. But, Tim, I think you did bring up a good point, right, which is on NAND, for example, that growth might be limited by tight cleanroom space. So do you think that overall growth in calendar 26 WFE might be limited by availability of cleanroom space, not only in NAND, but across DRAM, advanced foundry, and advanced packaging?

Well, I would say that... Look, I can't speak for the customers. This would be a much better thing to ask them. They can do amazing things. But I would say that generally, you know, there's a time that it takes to put in physical infrastructure. We suffer from the same thing, whether we're expanding labs or expanding manufacturing. And so, you know, depending on what the demand is, it could be limited. I think what we're trying to respond to is the point of how much could you accelerate? And that's a function of probably more physical space than it is – ability for the equipment supply chain to respond. And that's simply because our lead times are generally within the lead time of building a facility. That was kind of my general comment. Maybe we're not going to be the bottleneck, but clearly we've seen some accelerated demand as a result of the current demand, but also in anticipation of those future opportunities I would imagine this year you'll see some of those plans come to fruition, but I would suggest you talk to the customers and find out what their physical investment plans are.

That's a fair point. And then maybe for Doug, good to see the CSBG dynamics still on track to drive growth this year. I think the first nine months of this year, CSBG was up 7% year-over-year, but this includes Reliant. I assume that Of course, spare services and upgrades are growing at a faster rate anyway. Maybe quantify how much faster it's growing. And then maybe if you could just throw us up, I believe CSBG has been neutral to accretive to your overall operating margins, but you've had similar cost benefits as you've moved CSBG support closer to your customers, especially with the Malaysia build-out. Given all of this on a relative basis, where do CSBG up margins currently sit relative to corporate average?

Yeah, Harlan, let me impact that a little bit. Just to remind everybody on the call, and Harlan, I know you know this, but for others, there's four components of CSBG, spare service upgrades and then Reliant. Three of the components of CSBG are clearly growing. One is not, which is Reliant, largely because of maybe mature node spending across the whole world. Tim mentioned in his scripted remarks that you know, record spares and service combination. I said in my script remarks, hey, upgrades are pretty strong given what you got going on in NAMM. So that's the way to think through all of the kind of ups and downs, and CSBG is going to grow this year. You're right. CSBG is a creative operating margin. I've never quantified that for anybody, but that continues to be the case.

Appreciate the call. Thanks, Tim. Thanks, Doug.

Yeah, thanks, Harlan.

The next question comes from the line of Jim Schneider with Goldman Sachs. Please proceed.

Good evening. Thanks for taking my question. I was wondering if you could maybe just give us a little bit of color on the NAND market and what you're seeing. I think clearly given all the announcements that are out there on the market, there's the expectation that NAND orders could accelerate at some point and I'm sort of wondering your view on whether you're seeing that yet, whether you're seeing any kind of initial signals from customers in terms of longer-term forecasts, and when we might start to see that show up in the numbers. And maybe as a follow-on to that, maybe comment on whether you expect 26 growth in NAND to be led by upgrades, or whether you see any potential for new tools starting to lead that. Thank you.

Yeah, thanks for the question. I think that... you know, I addressed some of that in my comments about NAND, but just to kind of go back and say that of the $40 billion of conversion spend that we had anticipated, I think the demand signal right now is a little bit accelerated to what we originally saw, and that's based on the fact that BIT demand is, people are speaking about BIT demand that's a little bit higher than probably prior expectations. I think our business is going to continue to be predominantly upgrade-focused You know, not only in 2025, but through 2026. And that's primarily because there was a very large install base that had not been upgraded for a number of years. So there are quite a few tools that can still be upgraded to provide those higher layer count devices. Now as you do those upgrades, you tend to lose wafer out capacity. And so at some point, if demand remains as high as maybe people anticipate with these data center announcements, then I would think you would transition to capacity additions. But my comment about floor space, physical infrastructure, again, a better question for our customers, but I would anticipate that everyone will remain focused on upgrades since it's the lowest cost and likely easiest way to achieve higher performance growth in BITS. through 2026. And then beyond that, you know, it's, again, a better question for all the NAND providers to speak to their clients.

Great. Thank you very much. Thanks, Jim.

The next question comes from the line of Krish Sankar with TD Cowan. Please proceed.

Yeah, hi. Thanks for taking my question. Tim, I just want to follow up on the NAND thing. I understand clean room space is limited. Maybe Keosha is the only one that has a new FAB. But it also seems like the NAND utilization rate for most of your customers is heading towards 100%. So I'm kind of curious, in that scenario, should we assume that there might be a quarter or two where your NAND orders or shipments drop off? Or do you think it's going to be not like in lock and separate, it's going to be pretty smooth transitions?

You know, maybe I'll jump in and then, Tim, you can add. Chris, nothing goes up into the right every single period. In fact, if you looked at the most recently reported quarter, the net was actually down a little bit. It's going to continue to kind of trend towards that $40 billion, maybe a little bit more, but every quarter will have some level of variability depending on who's investing and what. You know, all these guys, or most of these guys, also have DRAM investments, so they're going to modulate what happens in what quarter. But, you know, things have strengthened somewhat relative to what we were describing a quarter ago. And, Tim, I don't know if you want to add anything.

Yeah, no, I think that's fair. I mean, I think the important thing to remember is that as we move above 200 layers, you know, the drivers for our business aren't just the increased bit demand. It basically is that to produce each of those bits, the intensity of LAM's equipment actually rises as well. Again, above 200 layers, we've talked about the fact that we start introducing several new types of products to deal with wafer stress, to deal with the higher gap fill requirements from the higher layer count devices. I talked a little bit about MOLLE in my prepared remarks. For us, as we see more interest and perhaps a bit of acceleration in those upgrades, we think that it's a combination for us of upgrades to existing install base and the addition of some new tools. You'll see it within our business, both in systems and in upgrades.

Got it, got it. Thanks for that. And then a follow-up for Doug. Maybe it's a hypothetical question for you. You know, you said next year WFE is going to grow. Kind of makes sense. Seems like most folks are in mid to high single digit growth in calendar 26. I'm assuming in that setup, I understand LAM's revenues are going to grow year over year. But with less China being a gross margin headwind and higher tax rates, Can LAMs EPS also grow year-over-year in that situation or outgrow revenue?

You're funny, Chris. You know I'm not going to answer that question. Listen, but you have it right, and I don't want to over-position the tax rate. Tax rate's going to go up just a little bit, okay? Don't go too far with it. Instead of low mid-teens, maybe it's approaching mid-teens or maybe a little bit towards the higher end of the low mid-teens. Don't go too far with that. The reduction of customer mix, though, will be a headwind for gross margin, right? We just took you down to 48.5%. Depending on how each quarter progresses, we're probably in that range for a while. You know, as things grow, that's going to be beneficial from a fixed cost standpoint. We don't have huge fixed costs. There's going to be a customer mix headwind and then probably an extra tariff for a little bit of a headwind, too. So, I don't know, anchor yourself. plus or minus where we just guided you in December, I think, and that'll be a good spot for you to start your models.

Got it. Thanks a lot, Doug. Appreciate it. Thanks, Chris.

The next question comes from the line of Stacy Rasgon with Bernstein Research. Please proceed.

Hi, guys. Thanks for taking my questions. Doug, my first one, I wanted to zero into something you said about next year. So you said second half-loaded year or But then you said the first half would be sort of flat to maybe up a bit versus the second half of 25. But then if it's the second half of the year, it feels to me like the second half ought to be up, you know, more materially than that. Am I sort of characterizing that trajectory correctly?

Well, I haven't given you any numbers for the second half, Stacey. I just said it's the second half of the year. And I said the first. It says second half weighted, Stacey, and that the first half was flat to slightly up from the second half of this year.

The second half weighted to me means at least the second half of next year should be higher than the first half of next year, correct?

That's what that means, yep.

Okay, got it. So then I want to dig into the implications of that with regard to China. You said China drops below 30% next year. It's probably going to be, I don't know, 36% or something next year. like this year. So that drop would have dropped to like 29. It'd be something like a billion and a half dollar headwind, maybe more. It's probably a high single-digit headwind to revenue growth. But from what we just heard, like revenue overall should be growing. I mean, it feels like it should be growing okay, given the trajectory you just laid out, at least qualitatively. Again, I just want to know, do I have that dynamic correct? And like, can you give us maybe a little more color on the non-China offsets? that are enabling you to overcome like a headwind from China that, like I said, has to be at least a billion and a half, probably something in that range?

Yeah, Stacey, what you just described is largely consistent with what I believe is going to happen next year. So, yeah, China's going to be down. Your numbers probably aren't too far off. The global multinationals, though, are going to offset that, right? More than offset that is our, unless we sit here today, right? And just think about what's going on, right? We've been talking about NAND a ton on the call. We've We've been talking about high bandwidth memory. We've been talking about accelerators and all that kind of stuff going to more advanced nodes. That's what's going to be offsetting it, Stacy.

I was actually just hoping to get a little more color on the granularity there, but maybe you're saving that for next quarter.

Yeah, let us leave a little bit in our pocket for next quarter.

All right. All right. Sounds good. Thank you, Doug. I appreciate it.

You bet, Stacy. Thanks for trying.

The next question comes from the line of Blaine Curtis with Jefferies. Please proceed.

Hey, thanks for letting me ask questions. I just want to ask on China. Maybe I had it wrong. You didn't really answer it last quarter, but I thought the strength that you highlighted for September was going to be multinational spending in China. That's clearly not the case. So maybe you could just walk through why such a big bump to China revenue in September now that it's done.

No, Blaine. If it came across that we were suggesting it was the multinationals in China, that wasn't what we intended to communicate. If we did, apologies for misrepresenting it. Listen, the multinationals in China stayed relatively steady, so call it flat-ish. The growth in China was largely driven by the domestic Chinese customer base. Gotcha.

And then just the driver of the second half weighted, is that across all your segments, or is that more of a Foundry Logic comment?

Listen, we'll give you more granularity. I know everybody wants it now, but we'll give you more granularity on the December call. It's just a description of what we see across the totality of the spending in the industry, WFE in total.

Okay. Thanks, Seth. Yep. Thanks, Blaine.

The next question comes from the line of Melissa Weathers with Dorsha Bank. Please proceed.

Hi there. Thanks for letting me ask a question. I wanted to check in on some of the new products that you introduced at the start of the year at your Analyst Day. Now that it seems like we're getting a little bit more momentum on the WFE side, have you seen any acceleration and engagements on those new products, the ACARA and the Altus Halo products?

Yeah, it's a great question. And we have. I mean, ACARA and Halo, these are both products that are very focused on inflections that are taking place in FoundryLogic DRAM and NAND. ACARA for conductor-edge high aspect ratio, very well suited as we scale gate all around and also heavily used in DRAM. We've talked about a couple of wins since February in some key DRAM conductor-edge applications. Again, Remember, we introduced some of these products specifically to improve our performance and revenue growth in Foundry Logic and DRAM because of the drivers there. Halo, I talked about on this call, again, continuing to make progress in securing 3D NAND word line applications, which is an important step for the ESST performance. And so I would say where I sit right now, There's a long way to go to deliver on the investor day full model, but I think from a product perspective and how we see the transitions playing out, we're feeling pretty good about the progress we've made since February.

Thank you. And then maybe one more on the backside power side of things. It looks like backside power nodes are going to start to ramp in volume next year or maybe the year after. So has anything changed on either the timing or the magnitude of what you're expecting for backside power contributions in the next couple quarters or years?

No, no, not really. I think in terms of change, I think, again, as you move forward and you hear all about the requirements and challenges of power in these very compute-intensive devices, it just gives us further confidence that you need solutions like backside power. I mean, it directly addresses some of the issues that customers have in scaling performance of high-compute devices. So it's an edge-depth intensive inflection, and therefore it's important for us when we're focused on it, and I think we'll do well as those nodes ramp.

Thank you. Thanks, Melissa.

The next question comes from the line of Mehdi Hosseini with SIG. Please proceed.

Yes, thanks for taking my question. All the good questions have been asked, so I have a quick follow-up. Going back to your slide number six. I know, I know. Going back to your slide number six, interesting observation. I understand the facility construction lead times are in a one to two year, and that's almost in a ballpark as leading as FAP. But I would also argue that there is increased concentration within that $8 billion of WFC for every $100 billion of incremental AI. And that increased concentration would, in my opinion, give your customers some leeway. They don't have to rush to secure capacity or to release all their POs. And I'm just wondering if you have any additional thoughts to it. Is that a factor?

Manny, that's a good question.

Not exactly sure how to answer it. I think Tim has been meeting with a lot of customers over the last couple of weeks and having conversations about, okay, what do you think next year looks like? Where are you going? And so forth. I don't know that lead times have been all that different, so to speak, at least not yet. I'm not sure I'm answering your question. I'm just kind of rambling here a little bit.

I think your customers don't have to worry about running out of capacity among their suppliers.

Well, I think that, look, whether it's us and how we work with our customers or I'm sure our customers, how they work with theirs, everybody wants to make sure they have what they need. So, you know, we spend a lot of time with our supply chain making sure they have the capacity, their ramp, they understand our plans. You can anticipate that our customers do the same thing with us to ensure that when they take an order, they can deliver it. I think we're in a period right now, as we talked about, of some acceleration. I don't believe most people anticipated the number of announcements that have come in recent months for AI infrastructure. It will take time for those, but I can guarantee when people hear those announcements, it ripples through the supply chain to make sure that capacity is going to exist and I think everybody goes to work. And if there's one thing that Lamb has been good at, is executing to the needs of our customers, and that continues to be our focus.

Got it. Thank you.

Thanks, Matty.

The next question comes from the line of Vijay Rakesh with Mizuho. Please proceed.

Yeah, hey, Doug and Tim. Just a quick question on 2026. As you look at the strength that you mentioned into next year, is that being driven by memory or foundry as well? Because it looks like DRAM and pricing have been especially strong. So just wondering what you're seeing on the memory side as well.

I think it's probably going to come from both. And again, we'll give you more color on the December call.

Got it. And then as you look at 26, obviously there's a U.S. It looks like a U.S. ITC that kicks in December 31st for higher investment tax credits, like 35%. And it looks like some chipset money is starting to flow again. Are you seeing that as a tailwind for WFP into next year?

Maybe only on the margin, Vijay. No, what's driving WFP next year is end demand. At the end of the day, yeah, I'm sure the investment tax credit is going to maybe influence a little bit of the geographic distribution of that, maybe a little bit. But then demand is what matters right now. Got it. Thanks. Thanks, Vijay. Operator, we will take one more call or one more set of questions. Sorry.

Okay. And the final question will come from the line of Brian Chin with Stiefel. Please proceed.

Thanks for taking our questions. Appreciate it. Maybe the first one, going back to that popular slide in the slide deck, of the $8 billion in WFE spending, on a rough cut, could you partition that across, on a percentage basis, Advanced Logic, DRAM, and NAN?

Brian, more than half of it is coming from memory.

Enterprise SSDs and high bandwidth memory. And clearly the great big GPU accelerators, the ASICs and whatnot are an important part of it, but more than half is memory.

Okay, that's helpful. And going back to the $40 billion mass in terms of upgrades over several years, do you view that as the industry having to sort of exhaust that and then capacity spending occurs? Or can they be somewhat concurrent maybe towards the back end of that maybe a kind of more customer-by-customer basis as opposed to... Yeah, I'll take that.

Yeah, I think just as you said there, it's going to be a customer-by-customer situation, and already today we're seeing some capacity additions, and that has nothing to do with end-to-end. That was the customer's plan all along, and so I think you find different customers at different points of where they are with their installed base, where they are with the needs of their customers and end markets, and... You know, the great thing for LAM is that we can work with customers in a very agnostic way as to whether they're gaining the bits and performance they need through upgrades or through capacity ads. We participate in both in a meaningful way. And so I think you'll see both. However, what I said was upgrades to install base tend to be the fastest and lowest cost means of achieving upgrades. it's at the higher performance. And so I think that most customers will prioritize that first before they get to capacity, but there will be some concurrence and maybe, as you said, towards the back end of that upgrade rollout.

Great. Thank you.

Yeah. Thank you, Brian. Operator, with that, we're going to conclude the call. Thank you, everyone, for joining today. I know Tim and I will be talking to a lot of you during the remainder of the quarter before we get into the quiet period, but thanks for your interest in lab research.

Thank you. This concludes today's conference. You may disconnect your lines at this time. Thank you for your participation. Thank you.