Good morning, everybody. Welcome to the Goldman Sachs Communacopia + Technology Conference. I'm Jim Schneider, the semiconductor analyst at Goldman Sachs. It's my pleasure to welcome KLA Corporation to the stage today. With us from the company, we have CFO Bren Higgins, and President of Semiconductor Products and Customers, Ahmad Khan. Welcome, guys. Thanks for being here.
Thanks for having us.
Bren, I think to kick off, you might want to give the audience a brief recap of the quarter and sort of the current business outlook as you maybe currently see it.
Yeah, sure. That's great. Just to frame things up a bit, 2026 is setting up to be a very good year for the company. If you just look back over the last three years, we've had a really impressive period of growth. In 2024, the company grew at about 12%, in 2025, we're 19%, 2026, based on some of the guidance we gave, would be somewhere in the low 20% growth range for the total business. Over that timeframe, our incremental operating margins have been at the top end of our 40%-50% target. Perhaps maybe even a little bit better than that, and so that translates into over 400 basis points of improvement operating margin over that timeframe. So the business is performing very well. We've seen share of market grow over that timeframe.
We're excited about that, and the operating leverage in the model continues to drive what is one of the best business models in the industry. We're pretty proud of the accomplishments to date, but we're pretty excited about what's to come. 2026 is a period of time here where the momentum continues to build in the industry. If you just go back to our Investor Day in March, we thought the industry would be somewhere between $135 billion and $140 billion. Today, and in earnings, we talked about a low $150 billion range. That translates into a mid-20% growth rate over 2025 for the industry. Over that timeframe, we've seen not only strengthening in our views of 2026, but also that momentum continuing into 2027.
One of the things we said at Investor Day is that we thought that 2027 would have a growth rate at the same level or faster than 2026. 2026 has gone up, but 2027 has gone up. I still think the statement is accurate. I still think that as we look at next year, given the backlog that we have, and we disclosed our backlog in our 10-K that we filed back in August of over $12 billion, and my expectation that in the funnel, we'll continue to see backlog build moving forward. That's leading to very strong visibility, and it's driving us within the company to really focus on three things. The first being to continue to execute on the opportunities that are out there. We have to drive and keep our focus on next-generation development.
We've got a lot of programs in the company to deliver new capability to support, which is, I think, a compelling market and a compelling roadmap moving forward. That allows us to also deliver new capability to customers, better cost of ownership, and allows us to share in the value of that moving forward. I think that's very important. Then finally, supporting our customers with service resources that we can keep their tools running at high utilizations. Process control is really critical. Yield translates into printing money in a lot of respects, and certainly with some of the pricing dynamics in the industry today.
Application support, how we get work with our customers to drive and deliver value out of our systems, and doing the things that we need to do just in terms of overall supply chain engagement, installation, all the things we do to continue to deliver our systems and execute to help them deal with what is a very strong demand environment where supply is being outpaced by the demand levels that are out there. Maybe I'll stop there. We'll get into your questions. Very happy to have Ahmad here. Ahmad runs the systems and customer channel of the company, so systems across process control, but also in our non-process control businesses and also the primary interface with our customer channel. You'll bring some, I think, some perspective that a little bit more insight and context into what's kind of happening out there today.
Fantastic. Welcome, Ahmad. Maybe just to kind of follow up on what you just said, Bren, there, in terms of the Investor Day, you talked about same or faster growth in 2027. Given backlogs being up as much as they are, how would you sort of handicap the chances of 2027 being sort of decisively faster growth? Then maybe as we look further, I mean, I think backlogs are quite high, so how would you handicap the chances of that strong growth sustaining itself into 2028 as well?
Well, to be talking about 2027 at this stage here in September, talking about the next year with the specificity I have is a little unique. 2028, the only thing I'll say about that, and I'll get to the 2027 part in a minute, is that there are multi-year investment cycles for our customers with new fabs and new fabs that are in process. We feel and see in terms of how we're sizing our factories and planning is that, I don't want to comment on the level of growth, but I see a sustainability in investment continuing as we move beyond next year. Certainly, the order profile would suggest that. The backlog is providing a good level of visibility, and it helps. I mean, look, we stay very close to our major customers in terms of their needs.
But getting the orders solidified, both from a plan and a slotting point of view, where what our customers want from us in terms of expectations for deliveries. It does help provide some additional clarity that will then go and drive the decisions we need to make to make sure we're in a position to support that. I think in general, we feel it's pretty constructive. Now we'll see what happens in terms of the clean rooms coming online, the ability of peer companies to be able to support different growth levels. We'll see how that translates. I think I've probably been a little more specific than some of my peers in terms of talking about next year. I think that's just predicated by the lead times on our business and the strength of some of our products in the market.
We'll see how that plays out, at least in terms of how we're managing the company, is we're managing it in a way that we've got to be able to deliver to some of the more bullish scenarios. I talked at earnings, I said, well, it's 190-ish billion kind of growth rates into next year, that would be a mid-20s kind of growth rate from this year. But there are scenarios and views of strength beyond that, and so the way we are running the company is that we've got to be able to deliver that. We don't want to be the bottleneck, and so we're taking actions to ensure that we're not. We'll see how it goes, but certainly the strength of the environment's pretty clear and there's a lot of pressure to deliver.
Great. Now, a couple of high-level questions for you for a second. I think we're in a situation where your customers are incredibly profitable right now, as profitable as I've ever seen them in my career, for sure, and that's probably a long time. But some of your customers' customers who are driving the ultimate spending on chips are spending well over $1 trillion per year, and that seems to be moving a lot higher. Those companies are actually tapping capital markets to sustain even their current levels of CapEx. How do you think about the ultimate returns for AI for the supply chain? Do you see any kind of systemic risk that could be building up that gives you any kind of pause?
Look, I will comment at our level, and certainly our customers and our customers' customers, is their profitability levels are unprecedented. There's a ton of profitability in the system. It contributes to the outlook I just talked about. I think given the need for everyone's short compute and the need for more compute, and the supply being deployed to meet the demand that's out there, and how that's translating back into our customers' profitability, is that they can afford and will sustain this investment. That certainly, as we look at it over the next couple of years, how we're thinking about it. Over the long run, we'll see how it plays out in terms of ultimate returns on AI. We certainly see a number of compelling opportunities within the company, some of the things that we're doing to take advantage of it.
At least as it relates to our outlook and the profitability to fund that outlook, never seen anything like it, and I think we're pretty compelled about our ability to deliver to it.
Yeah. Great. One other high-level question on AI. I think every conversation at this conference has touched on AI in one way or another. But one thing that I like to do is try to understand what companies that are presenting here are doing internally with respect to AI, how you're leveraging AI. Is this something that's still just kind of cost reduction and operationally focused, or there are things that are driving the top line as well?
I will let Ahmad weigh in as it relates to some of our product development efforts. One of the things we talked about at Investor Day is the use of AI in our systems. We had our first engagement with AI back in the middle of the last decade. We put it in our first product back in 2019. Now we have it deployed across the portfolio. In terms of product development efforts, and I will come back and talk about some of the operational things in a moment, but why don't we let Ahmad weigh in on that?
Yeah. AI is deeply deployed in our systems, primarily because KLA takes the most amount of images of semiconductor devices probably in the world. Just to give some context, an average high-end KLA machine can take about one petabyte of images per day. In order to process those images, we have completely changed our compute architecture, and we utilize GPUs and memory to do these calculations. All of that is driven by AI, and we use novel algorithms to determine what defect matters to the customers. That is how we use AI, but it is very deeply engraved in our systems.
One of the things we have been really compelled with is how strong the compute is relative to the alternatives. As we move the architecture of our systems, and we talked about this Investor Day, to a GPU-based architecture, is that we find that we can, against the baseline, we can drive better cost performance, we can deliver more capability, better power consumption. We had to do a lot of work to reprogram our compute architecture to be able to deliver that. But we have been very compelled by what we have seen in terms of the benefits we see. We can understand, in terms of how hyperscalers are investing in AI servers, that they are realizing similar benefits.
As it relates to the infrastructure build-out, it is very interesting and we have seen some really good advantages in terms of the operating system relative to the previous CPU-based structure that we had in the past. Within the company, obviously in engineering, we are seeing a lot of advantages in terms of just the efficiency of writing code and debugging code or doing experimental code. Our teams are. We are seeing, like others, we are spending more on tokens than we thought we would spend. That is a factor in the business. But we are starting to see some efficiency in there, understanding the context of the problems is really critical in terms of being able to write and get productivity out of it.
Operationally, though, as you look at SG&A and so on, we're finding that where we can use the capabilities to get access to data quickly, we're able to do that. We have a number of repetitive processes that have a human element, and how do we replace humans in that? We're doing those things. We're using AI and machine learning techniques to do supply chain optimization and so on. I wouldn't say that we're driving returns yet, but we've got a ton of focus on driving adoption, just trying to change fundamentally the way people work. I think it will be a collection of a lot of small things you'll do that will start to change the nature of a lot of roles. I think we will get there.
We think, as I said at Investor Day, that it will ultimately be a factor in how we're driving leverage in our operating model, how much we're spending on SG&A in terms of leverage on the growth that we expect over the next several years. We're focused a lot on driving adoption. Everybody in the company has a Copilot license, and we're using multiple tools, again, across engineering. Every engineer has licenses as well. I think we're on the path, and so we think that the capability will deliver returns over time.
Great. I want to take a regional cut to your business for a second and talk about the U.S. semiconductor industry, because clearly the U.S. has lagged the world in global semiconductor production for the last 25 years plus. It seems like that could be changing a little bit now with TSMC in Arizona, Samsung Foundry, potentially Intel. Maybe talk about your market position specifically in the U.S. market. Does your market share in the U.S. reflect your global market share, broadly speaking? Or are there areas where you feel like you're actually punching above your weight in the U.S.?
Yeah. I would say market share-wise, our overall engagement in the U.S. semiconductors is pretty deep, with the two top leaders, TSMC being in Arizona and doing a pretty wide footprint there. KLA has always been very close with TSMC and all of our customers, and we're deeply engaged there. That same engagement goes from Taiwan over to the U.S. Samsung has had expansions in Texas, and we're deeply engaged there. Micron, being the premier memory U.S. company, has had many expansions in the U.S., and we're engaged with them and growing there. There's new companies coming out. Intel, of course, we are deeply engaged with them and driving yields for them. That's been a challenge for them, and we've been very closely engaged with them to drive that. There's new customers coming in as you know, like Terafab and others.
But if we step back, I think five years ago or so, the compute architecture changed very significantly, where the number of applications reduced, but the compute per application went up pretty significantly. That is what has changed. Therefore, the GPUs came on board, and then onboard memory came on board, advanced packaging came on board. All of this is driving increase in overall compute, overall need for compute, and then, of course, process control intensity is going up for the same reasons. So we are deploying our BCAMs at U.S. customers and elsewhere.
Yeah. I think as our customers expand the footprint, it does create an opportunity. Now, we have to invest in it and talked about applications and service, and we have got to make sure we have got parts to be able to support a broader footprint. So there are investments that happen. A s these operations start to scale, there is a higher reliance level, in my opinion, on a scale provider to be able to provide support, to deploy resources, to bring resources from Taiwan to help with what is going on in the U.S., for example, in Arizona, that over time, we are making those investments today. As those operations scale, we will start to drive more leverage opportunities on those investments.
It creates an opportunity where there is a higher reliance on us to help drive and execute similar performance in the new fab operation in a different location to the other fabs in the primary location. So, I think in a lot of ways, from a share point of view, it is at least sort of flattish, and it does create opportunities for us to collaborate more, and hopefully that translates into more business for us.
Ahmad, I want to ask you about Terafab, which you raised. Can you maybe say anything about the business engagement that KLA has with Terafab and the kinds of ramps or opportunities you see for the company as a customer?
Yeah. I wouldn't go into extreme detail about it other than the fact that I think everyone knows that they're engaged in an R&D line first, and we're engaged with them on making sure that they're able to develop those processes. They are bringing process transfers from other customers that have developed nodes, and we are deeply engaged with those customers. So we deeply understand how to drive those yields up. So as Terafab starts building that initial factory, we would be engaged with them on that, and then as they go forward with HBM fab, we'll be engaged with them. But prior to Terafab, of course, we were engaged with them on other parts of their businesses that need advanced packaging and chip integration, so we have a relationship with them from the past.
One of the things about go forward that is particularly exciting is the broader base of participation at the leading edge. Over the last several years, there's really effectively been one player who's been driving the leading edge, driving a lot of efficiency in that part of the market. More players chasing what is a pretty exciting opportunity will be good for broader collaboration, more investment. I think that in terms of how that translates back into wafer equipment intensity is a positive, because a lot of that investment will be strategic, and takes time, and you've got to be in a place to be able to deliver to win business. I think that there's an efficiency element of that that will be favorable over the next five years to what we've experienced over the last five.
Every time when a new player joins semiconductors, process control becomes a critical element because unless you yield, there's no revenue. I think the engagements with all new customers have with KLA has been always very good. We ensure that we help yield, do process transfers and things like that. I think this engagement would be good just like any other.
Great.
There's this really strong design environment, right? If you think about, I think KLA and process control more broadly benefits from aspects of high-performance compute chips that we benefit uniquely. The fact that there's a lot of designs moving through fabs create a more higher mix environment. That doesn't necessarily change an opportunity for a process tool, but changes process control a lot. Our customers managing a more dynamic environment, delivering different designs to specific windows, not starting too much inventory or starting too many wafers or too few, but being able to deliver what they need. Design rules that test designs in different designs that have different design rules that test the design rules in different ways. The value of the die, the size of the die, bigger die are harder to yield. They're more expensive. You spend more money to monitor them.
Those are all kind of unique benefits to process control that makes the next several years, and we talked a lot about this Investor Day, that this particular part of the market, and it grows faster than the rest of the market, creates some unique opportunities for driving higher process control and KLA relevance.
More players doing it contributes to that.
Overall inflections in semiconductors are increasing over time, right? If you look at the 2 nanometer design rule with gate-all-around coming in, it's a very complex transistor architecture. KLA plays a very important role with our optical inspection systems and electron beam inspection systems to drive that yield. High-NA EUV is coming. That means that you are able to print smaller lines and spaces, but the customers will also print smaller defects. So KLA will have to help with driving those yields up. These are all very large dies, and larger die has a large process control element because if you have a wafer with two dies and one killer defect, then your yield is 50%. If you have 100 dies, your yield would be much higher. So as die size goes up in logic and also in memory, process control intensity goes up.
Also because of bandwidth, memory has to be coupled with GPUs very closely. Now they're integrated on a system. In some cases, people are thinking of putting them on top of each other. So that drives a lot of process control intensity because you have losses as you build an HBM. HBM has 12 dies. The base die is logic. The rest of the dies are DRAM. As you stack them together, you would have losses. So that's part of the reason why in advanced packaging, KLA is doing quite well and because process control intensity is going up. Then with the agentic workflow, we see that the CPU portion of the market and the NAND portion of the market is also growing, and DDR market is growing. So all segments are growing and KLA is playing an important role in driving yields up.
Great. I want to unpack a few of those things. Maybe just to kind of go back to your Investor Day, Bren, for a second. You laid out a path to a 17% revenue CAGR through 2030, process control outgrowing WFE, and then assumption of kind of 60% split between foundry logic and memory, 60/40. The profile of WFE going into next year or at least over the next 18 months or so is starting to get a little bit more clear. What is your level of confidence in the near term being able to outgrow WFE if one of the bigger increases is coming from memory? Maybe touch on some of the points that Ahmad just made.
Yeah. As you look at the setup for 2027, I think there are some things that I think are pretty constructive for the company. First is that I think from a supply chain point of view, we are continuing to add supply. We have certain components that have longer lead times, and so the industry momentum into 2026 changed very quickly, and so we have seen that kind of slowed us down a little bit in the first half of the year where we were somewhat supply constrained. I could ship more if I had more parts even today. But we are seeing that improve and we see the second half of the year up 20% from the first half and would expect to continue to see our ability to scale that moving forward as we move into 2027.
So we have this very strong, I think, leading-edge environment with broad-based participation, the ability to ship into that. In the legacy parts of the market, including China, I think you will start to see some growth again. It is a smaller part of the business, maybe 25%-ish of overall WFE, but I think you will start to see some improvement there, although it will not grow as fast as the overall. So as a percent of the total maybe comes down. Within memory, you will see more greenfield investments, so new fabs. So when new fabs, they have to tool the whole fab. So that creates an opportunity that is greater than what you have in a more of a upgrade environment where customers are doing tech migration to the next node to add incremental bit supply. So in a greenfield environment, that is positive.
HBM is driving intensity, and so we have talked a little bit about some of the HBM drivers, but in some product lines, we have intensity in HBM that matches what we see in advanced logic. So that is good. The continued introduction and more layers that are EUV layers in conventional memory. Conventional memory can be a bit of a headwind for intensity, but it is improving over time. So I think in general, when you look across the board here and also some new node investment that will happen at the leading edge next year, that the setup is pretty good for KLA, even despite a very strong memory environment. We feel very good about our ability to continue to drive what we said at Investor Day that we would drive a multi-year improvement in share of the overall market. Driving share of market improvement is difficult in this industry.
It means that something is fundamentally changing, either in the device or the market, that drives the need for one thing over something else. The last five years, we've seen that happen, and we were one of two markets that really drove share of market over the last five years. Over the next five years, we think that because of what's happening in high-performance compute, that it will contribute to that trend increasing more over time. But in the nearer term, I think the setup's pretty good. As more supply comes online, given the backlog we have, we feel very good about our ability to perform and continue to drive that thesis.
Great. Now, advanced packaging is one area that's clearly outgrowing the market, given everything in AI-related devices. You've grown that business 40% over the past couple of years, on track to probably grow even faster in the near term. I think you said your share of market was over 5% last year. So what's your confidence in being able to take incremental market share in that market? Specifically, if we think about the more specialized players in that market, such as Amkor, Camtek, et cetera.
You want to start?
Yeah. So advanced packaging has had a very interesting lifetime. If you look at in the past, it used to be fan-out packaging, where you took a final die and you did fan-out, and then you provided that chip to the customer, and the customer put it on a PC board. What has changed in the industry is really the bandwidth requirement for semiconductor compute. So everything has to be co-located next to each other, and the transfer of electrons has to be very, very fast, and this is what drove advanced packaging. We first started with advanced packaging about five, six years ago, brought in our initial systems, and have done really well in market share gain. We had 10% share about five, six years ago. Now we are above 50%, 56%, 57% share in advanced packaging. The main reason is the capability requirements.
Now, this is still bump packaging. What is happening now in logic with SoIC is hybrid bonding, die-to-wafer bonding, and that has a 10x change in specifications of defects that kill the die. Therefore, our most advanced systems that are used in the front end now are being used in advanced packaging, or older models, but the high-end capability. Then with the introduction of hybrid bonding in memory, the spec tightening would be pretty significant as well, and that is going to drive, again, our front-end systems to go into advanced packaging. We have developed a full portfolio of our front-end systems and now made them capable for advanced packaging, and that is why we think that the growth continues in several years from here.
In 2023, we were 2% share of advanced packaging market. This year, it will be between 7% and 8%.
We have seen the relevance, given the dynamics that Ahmad has talked about, in how important the package is and how much front-end processing is now happening in the package has driven the share of market higher. So it is an intensity gain, very high sample rates, but it is also a share opportunity. The market is moving to the need for more capability, and so that will then tap into the existing portfolio, the front-end portfolio that exists to support it. We did a very good job in engineering as we started to develop the handling capabilities in the primary product servicing packaging many years ago, and we have been able to then reuse a lot of that engineering across the portfolio.
To Ahmad's point, we have the portfolio ready, and so as our customers continue to drive to density shrinks of the lines and spaces in the package, that we have that incremental capability available. So we are pretty excited about it, and it is not that we have to go develop new products to be able to deliver to it.
Just to add, the economics of advanced packaging is very interesting, right? Our customer has spent a lot of money making that initial wafer. That wafer that comes out of the front-end fab for logic, you guys know what the prices are, but $30,000, $40,000, $50,000, a memory $50,000, $60,000, $70,000 wafer that comes out. Now you're going to package it, and if you have losses during packaging, you not only lose the package for the customer, but you also lose that initial revenue. Therefore, our customers are not going to take risks as they integrate these devices, and therefore, advanced packaging becomes very critical for process control, and customers want to inspect every step to make sure that the yields are going to be good.
Yeah.
The economics make sense.
Okay, want to wrap up on two questions. One is an operational question. Bren, you talked a couple of times about supply chain constraints. One of those clearly has been DRAM. You talked about the pressure on your gross margins you're seeing tactically from that. First of all, is that one getting resolved soon? Do you see other supply chain bottlenecks beyond memory? Just how do you think about your overall production capacity given the very strong uptrend that we're seeing in the industry right now?
Yeah. The memory issue has been more of a cost issue than it's been a supply issue. We initially, when things started to ramp, we went out and secured supply to meet our needs for 2026 and 2027. We started to see pricing increased. Momentum has increased, so we've had to go out and secure more supply, and we've had to secure more supply at much higher prices. So it has had an effect on the gross margins of the company in terms of a headwind. I've talked about that, roughly 100-ish basis points of impact against the baselines from, we'll call it, 12 months ago. So I don't think it's necessarily an issue of being able to secure the supply we need, but we're dealing with the cost dynamics.
As it relates to the rest of the supply chain, I think, one of the things, and I talked about at Investor Day, is our supply chain is unique to KLA. We are a highly differentiated offering to our customers. A lot of our parts are customs, and with our key suppliers, we work very closely in long-term partnerships. It is much more of a strategic supply relationship, less. We spend a lot less time on commercial issues and a lot more time on capability requirements and roadmap. When you look at optical components, for example, they do carry longer lead times, so you have to go secure and put that supply in place. We have been working through that, and our suppliers have been responsive. We are working with them, and we will make investments with them.
Those investments show up in different ways, whether it is large prepaids or sharing of CapEx and so on. We have been making investments not only to support where we are today, but also our expectations over the next several years. We are having supply discussions around our needs for 2029 and 2030. As I said, we have to be in a position to be able to support the more bullish cases. We do not want to be the bottleneck. Because these decisions take a while to action, then we have to think about having the headroom that is required to be able to meet the different scenarios. It is a core competency of the company. We feel very good. Our supply chain is pretty resilient.
Typically, when we can commit and understand a volume level, they can deliver that volume level, and then we can go manage all the other issues you have in a strong growth environment within the window that gets established by the longest pole in the tent, if you will, as it relates to optics volume. I think our own capacity is, we can react to that, and we are investing around the world at all of our facilities. In November, I am going to Singapore to do a groundbreaking on an expansion of our facility there. We will do a ribbon cutting on a groundbreaking I did two years ago, and then a groundbreaking on the next building. It will be a cut the ribbon, scoop the dirt kind of meeting.
Yeah.
I think we are excited about what is required, but we are doing what is necessary. I am not concerned about our ability to do that over time, and we are trying to be as forward-looking as we can to ensure that we have what is required to support the industry.
Maybe a quick lightning question. We've only got about 30 seconds left, but I did want to ask you, if we're back on stage here in five years, what do you think the one thing that investors are going to be surprised about in terms of KLA's performance, looking back in time?
Well, we laid out a plan for five years from now, and it's interesting, as we laid out that plan back then, I generally felt like the plan was about right, in terms of industry spend, because you had some people say, "Well, that's too aggressive," and other people said, "Well, that's not aggressive enough." Maybe it's about right. I would say today the view on that plan would suggest that it's conservative. Certainly given the momentum we've seen as it relates to expectations for semiconductor revenue growth and how that translates back into wafer equipment, some of the dynamics we talked about, both in terms of what's happening in memory, but what's also happening at the leading edge in logic, could potentially advanced packaging fuel additional upside to that.
I think when you take a step back and you look at these plans, what matters is, you think about the long-term assumptions and do those make sense. But as you look underneath those, what really matters is, that we've established, we've made commitments, put these plans in place, and we beat and delivered to them. We've proven we can scale the business, and the fundamental, the best example or source of credibility for the operating model that's underneath is what you've done in the past against the commitments you made. I think we stand up very well with that, and our expectations moving forward are consistent with that. I'm a big believer, I think the next five will be better than the last five, and for all the reasons we've talked about. Thanks for having us.
Thanks, Bren. Thanks a lot for being here.
Thank you.
We appreciate the time.