I'd like to get started. Thank you all for coming. I'm Toshiya Hari. I cover the semiconductor and semi cap equipment space at Goldman Sachs. Very pleased to have Ben Higgins, Executive Vice President and Chief Financial Officer from KLA as well as Ahmed Khan, President of the Semiconductor Process Control Group.
Bren, I wanted you to kick us off by kind of reflecting on 20 Sure.
Well, you. Yes. No, thank you, Toshiya, having us. So over the course of the session here, we'll make some forward looking statements. And those statements are subject to risk, and you can see a full list of our risk factors in our SEC filings.
It's quite a list. So I encourage you to take a look at it. Very happy to have Ahmad here too. So for those who don't know Ahmad, Ahmad runs all of the semiconductor process control business for KLA. So I'll be throwing some questions to Ahmad as we go here to get his perspective.
So 2019 was a good year for the company. If you just take a step back, I mean, we did a branding change, which is an exciting thing for the employees and certainly part of how we're thinking about the next sort of step of our future as a company, both in terms of opportunities within the parts of KLA that we had in process control, but also exposing us to more opportunities outside of front end process manufacturing. And so we're excited about that. We closed the Orbotech transaction, gives us that exposure, which we're excited about. And we've seen a nice progression through the year in terms of strength in logic and foundry supported by what we believe is a very strong competitive position, a portfolio approach to delivering products to market that makes it a differentiator.
And as we move through the year, we saw considerable strength in the logicfoundry part of the business, mostly around foundry. Memory turned out was weak and stayed weak, but the foundry intensity was very strong. And I think the strength of the seven nanometer node maybe surprised some people, including our customer in terms of overall strength. And so we were able to participate in that. And so against the backdrop of an industry that at the beginning of the year, people thought would be down 20% to 25%, memory was down in excess of 30%, ended up down what we think is about 10% or so given the strength of Logix Foundry in terms of how it offset that.
So for the year, our business finished and we were slightly organically, just slightly above flat, so about 1% growth in a down industry year. And I think that reflects our exposure to those markets, both in terms of process control intensity with a high mix environment for foundry, technology progression at seven nanometer with EUV coming in, which I'm sure we'll talk about, that we think we're very well positioned as we
go
into 2020. Service business also performing very well, 70% plus of the service revenue stream is contract based. And so we had a little bit of weakness in the first half of the year in service driven by lower utilizations as memory customers were idling capacity, but then we saw that strengthen in the second half. So we believe that that's well positioned to grow double digits, 9% to 11% type growth as we move into 2020. So we think we're pretty well positioned.
I mean we're excited about what's in front of us.
Thanks for that. And then for 2020 WFE, I don't think you guys necessarily gave a specific number on the call, but you did talk about a very healthy outlook. Can you remind us what kind of growth expectations you have for logic and foundry, NAND And and for NAND specifically, is it one or two customers? Is it broad based? What sort of participation are you seeing at the customer level?
Sure. So I think when you look at WFE we think WFE is going to finish somewhere around 49,000,000,050 billion dollars in 2019. And so as we move into 2020, we talked about a plus 10% sort of plus or minus type growth year this year, driven by what we expect to be a broadening of foundry and logic investment, both from other foundry customers, more intensity in the logic space and a stronger year in China from logicfoundry. So all of that collectively would drive about 10% to 15% growth year to year. Three months ago, I would have said flat, but given the strength of some of the announcements we've seen from some of our major customers and as we've seen the funnel in China sort of fill out, we think that there's growth there.
On the flash side, again, after a pretty significant down year, we would expect to see some correction there, probably somewhere in that 15% to 20% type range of growth from 2019 and maybe more middle to second half of the year. On the DRAM side, not expecting to see much I mean, of a flattish outlook and so we'll see how the year progresses. Mean one thing with memory customers is we tend to see things turn off fast and turn back on fast. So we'll see how it goes as we progress through the year. But we're excited about that outlook.
Given the process control intensity of memory, you would expect with more memory that, that puts some pressure on that, but we have a number of new products in the company that we're bringing to markets. And so we would think that the contribution from those products should lead us to a year where process control intensity should be flat. We think there's some minor share opportunities. And against that industry backdrop, we would expect the company to grow in line with the industry, maybe a little bit better. So you take that, drop through in terms of the operating leverage model, the company should drive 1.5 times revenue growth rate in terms of earnings growth.
That's how we're sizing and expecting running the company.
Bren, the broadening of the spending profile that you spoke to for Logic and Foundry in China, I'm guessing that's kind of lagging edge, maybe leading edge for them, but lagging edge in relation to your leading edge foundry customers. Are those technology buys or capacity buys? Or is it a little bit of both? What's driving that sort of broadening of demand in China?
Yes. So I'll start and I'll let Ahmad weigh in here. But yes, it is more trailing edge. It's more IoT centric. Certainly, you have industrial and automotive and some of the broadening of demand that is now consuming some of the trailing edge nodes as well.
And so there's demand domestic demand that they're shipping into, but also other demand out there. So given the roadmaps, we're talking about the leading edge roadmaps are seven nanometer into five as we progress through 2020. You're looking for maybe in sort of 14 nanometer and above type technologies in China today and a lot of activity at more lagging edge. Now as you're new into that, you're trying to progress a road map and that those customers rely as those dynamics are that high mix drives higher complexity in devices and so more support from companies like KLA in terms of supporting their road maps. So I think it's a good opportunity for us from a participation and a share perspective, and it reflects the broader demand that we believe that exists out there for and diversification of demand for semiconductors.
Yes, think you've covered it well. What I would add is the infrastructure spend on top of that, for example, for mask houses, there's infrastructure spend for reticle inspection to ensure that you are able to generate the best reticles and clean reticles. On top of that, wafer houses that need to set up shops to ensure that they are able to generate wafers for the local market. And for all of those, the inspection metrology requirements are pretty tight. So it's not it's from a KLA point of view, we still sell a lot of the leading edge equipment into that market because the specs are quite difficult to attain.
The rest of it, think, is well covered by what Brent said earlier.
Understood. I think one of the issues in China and when the memory ramp of 2017 and 2018 was the ability of the customers to get wafers. So a lot of that wafer supply was consumed by just the general leaders out there. So there is a desire to build some internal wafer capability to be able to produce bare wafers. And so when you're looking at memory devices and particularly vertical memory devices, you start to have more stringent, to Maad's point, stringent requirements on specs, both from a flatness and wafer geometry point of view, but also just general cleanliness and other attributes.
So it's an aspect of our business we're exposed to that I think sometimes people forget about, but it's very strong product position and something that sort of augments our broader growth profile.
So just to confirm, so these are local Chinese bare wafer manufacturers? The coronavirus, obviously, a very fluid evolving situation. I think you were probably at least probably the first company in my coverage to quantify the potential impact in your guide. It must have been a pretty tricky process. I guess it's been about a week since you reported, maybe a little bit more than that.
Any new findings in terms of potential supply chain disruptions or changes to the demand outlook? Yes. I don't think it's
I mean, obviously, it's a very fluid situation. I think from a financial point of view, it hasn't really changed all that much. I mean, logistically, it's a challenge with people trying to get people in and get them back out and that supports our worldwide customer base quarantine periods around people and so on. What we tried to do and maybe I have the good fortune of being able to go tool by tool in terms of how we drive a bottoms up revenue plan and identify what we thought as tools that were in Hubei province across the broader business that we thought would be impacted in terms of just delay in timing. And so that we adjusted our midpoint of our guidance to reflect that.
Now more broadly, supply chain, other dynamics here, we expanded the range to try to compensate for what else could potentially happen. We do source some components out of China. And so those components as those go into our factories to support other shipments. So we wanted to see how that would play out. Everything started.
I wouldn't call it normalized, but we did see some improvement this week as people returned from the Lunar New Year. Facilities starting back up, maybe not at 100% capacity, but certainly facilities that are starting up that we believe will catch up. So not much to say beyond what I said last week at earnings. So we'll play through this quarter. As you think about our 2020 view, given our lead time, given our customer plans tend to have some longevity to them, unless there's some sort of broader issue that ultimately impacts their demand profile that affects how they think about supply, Our 2020 view is basically consistent that I wouldn't expect it to have any lasting effects.
But over the course of the next couple of quarters, we're scrambling a bit to manage and meet our customer demands and working our way through it.
Okay. But the base case is whatever you kind of removed from your March guide, I think it was 3% to 5%, would show up in the June. Is that sort of the base case scenario Of In terms of process control, the market as a percentage of total WFE, I think it's been sort of in the low to mid teens for a while. It definitely ticked up in 2019.
The market was down double digits. You guys Brent, to your point, you were flat to up organically. What were some of the drivers that drove that increase? Sure. Why don't I
let Mon take that?
Yes. So I think as you know, process control intensity in logicfoundry is higher and we did well there. We released a few products a couple of years back. Gen5 was a product that we had mentioned quite a bit. And that product started to ramp in a big way in 2019 serving the logic foundry case and in this case both logic and foundry and did quite well.
We have seen cases at multiple customers where Gen five has been able to solve problem not only for the node that they were purchasing, let's say five nanometer node in one case, but also we're back porting Gen five to previous nodes to solve defectivity problems and therefore driving improvement in yield. So that was one major factor. And I think that drove to the overall growth in the business. At the same time, the mask infrastructure setup was important. As you know, the number of designs that were predicted a few years back for seven nanometer were not very high, but we have seen that that's not the case anymore.
The foundry partners' customers have done a really nice job pulling all the customers, their customers to go to advanced nodes. And for that reason, customers are building several mass shops to ensure that they are able to produce all the reticles that are required. So this really drives process control intensity because process control intensity for a mask inspection is quite high. You cannot ship a reticle with a defect, not even one defect to the wafer fab because as you know, it will print on every single wafer, and that defect would be not considered good. So mass shop infrastructure, was quite important to set up.
So overall, those were this happened in foundry. We also drove very close engagement in, logic to do the same, for Gen five and drive Gen five, utilization in logic, and we saw the growth there also.
Yeah, one thing about the reticle inspection is you would have thought we would have seen more of the demand that surprised so many people because it does behave like a process tool and that every reticle has to be inspected before it leaves the reticle house. And to Maud's point, it's a very high end inspection. It's a 100% defect detection and it's a market that we have a very strong share in. So for every design sort of drives a significant reticle set and so there's a lot of demand for those tools. So that's a business that inflected strongly, I would say, at least 20% year And to I would expect to continue given what we expect to see follow on in terms of seven nanometer activity, but also expectations around five.
So it's one of those areas that given what we're now seeing there that gives us confidence around the growth expectations for logic and foundry, both from the leader, but also from other suppliers.
And just to add one more point with the introduction of EUV, the defect modes on the wafer and on the reticle change in a significant way. On reticle, it changes because the line widths are shrinking and for that reason, there is new investment required, new machines required in order to control that defectivity. But on the wafer side, your previous know how doesn't really work. And this is one of the reasons why Gen5 did quite well in the wafer side also and several other systems because the defect modes changed completely and for that reason, we saw growth. Moving forward, with implementation of EUV, further implementation of EUV in five nanometer and three nanometer, we see again process control intensity continuing at a high rate because the defect and the metrology modes would change quite a bit.
And for that reason, your previous DOHA doesn't always work, you need new tools, you need to upgrade your fleet in a significant way and that drives revenue growth. And then eventually with sub three nanometer with nanosheet and gate all around coming in, that changes the entire front end transistor structure and that changes defect and metrology modes and we're developing new machines now to ensure that we will be able to meet needs for that. Quite a lot of pull in from all the R and D locations and customers for that type of close development. We have a very good infrastructure internally for close partnership with our customers early in the phase of their development to ensure that we are able to help them with this development, but at the same time this ensures that when they make their HVM decisions that they choose KLA.
Just on that point on EUV, how should we think about the growth in your opportunities as some of the memory companies perhaps start to embrace EUV in their processes? I guess one in particular, relatively near term, one to two years and then broadening out potentially to two, three, four customers.
Yes. So what I would say is similar to Logic, it's all about, as you mentioned earlier, one in particular. And then so there's a gradient in memory as to who's going to adopt first, who's going to adopt second and potentially third. And then the next gradient is how many layers the first person would implement, second layer is third. But first, going back to the infrastructure spend, there is almost a fixed infrastructure spend if you're going to do EUV.
And KLA gets that benefit, meaning you need to set up a mesh shop that requires radical inspection tools, you need to do inspections in the wafer fab. At the recent Investor Day, I spoke about a new product that we're developing called Gen5 EUV extension. And this is a product that really focuses on ensuring that the reticle defects in wafer fab are also monitored using wafer inspection. So these products need to get introduced even if you do one layer or three layers or five layers or 10 layers. So there's an initial infrastructure spend for R and D and setting up the line, we benefit from that.
And then of course as EUV scales with the number of layers, would benefit from that. So we are again partnering closely with customers who are looking at going into memory EUV and helping them with that structure.
I mean if you just back up a little bit and think about what EUV enables, it enables linear scaling again, right, significant linear scaling. And for the last number of years, the ability to scale has been really challenged by some of the lithography challenges to print smaller feature sizes and multi patterning schemes and other things were developed to do that, but also architecture changes and so on for customers to continue to deliver what I call the Moore's Law attribute. And so when you add EUV, you enable scaling again. So it means that small defects matter. You see more aggressive scaling, smaller feature sizes.
And so those are things that when they weren't happening, some of the capabilities that KLA would offer wouldn't necessarily see the same value. But now as you start to press that technology roadmap, customers have to find smaller defects, defects that at one node maybe didn't matter, now they do. And of course, when you have super sensitive technology, you find all kinds of stuff. So trying to figure out what the signal versus the noise is also more challenged. So they have to buy new equipment.
It's very hard for them to reuse their equipment as they move to the next node and so on. So it really is, we believe, a strong driver both for our business, but also for driving what we would expect to see more WFE investment in logicfoundry moving forward. So yes, what happens in DRAM is a technical challenge, but also in the other parts of the market.
Got it. I guess based on your position in process control, you have a lot of insight into what's happening at your customer fabs. Just going back to EUV, do you feel like EUV is production ready for memory? I know, obviously, it's a mass production today for foundry, but for memory, throughput is clearly, an important factor. So what do you see in the market?
What's your view there?
Yes. I would say that, there are trade offs that customers will make in logic versus memory from a throughput perspective for two reasons. One of them of course has to do with redundancy differences between logic and memory. They're different. So customers can take choices of not implementing pellicles on the reticles.
This, of course, will enable faster throughputs on the scanners. So there is that element, and I believe that first implementation in memory would look like that. Of course, there are customers who make both logic and memory devices. If you are making a logic and memory device, you're setting up an infrastructure spend for logic anyway. In that case, you're able to depreciate that infrastructure spend for memory because your requirement for, goes down significantly.
As I said earlier, if you're gonna set up an EUV line, you have to do that initial infrastructure spend. So customers who dedicated memory will have a deeper thought before they implement EUV.
Got it. That makes sense. You guys introduced a new e beam inspection platform. Obviously, you've got two competitors who've been in the market for some time. I know KLA was in the market pretty actively back in the day.
You kind of dropped off leaderboard, if you will, for some time. You'll be back on it. To the extent possible, can you speak to sort of the points of differentiation of your tool relative to your competitors? So
I think one of the things that we spoke about in the Investor Day was the KLA operating system, and we really spoke about codifying that operating system across the organization. And one of the things that the operating system calls for is that we don't just go into any market unless we can really solve very difficult customer problems, and at the same time provide differentiation for the product that the customer would value and therefore value share with KLA. So we have been watching the e beam market for quite a bit. We still continue to believe that optical inspection is the workhorse of the industry and no amount of single beam tools or multi beam tools will be able to change that. Our success in the Gen five product line shows that optical at five nanometer and sub five nanometer would be served very, very well with optical technology.
At the same time, there are class of defects that may
or
may not be catchable in the first instance with optical and we wanted to create a product that would be synergistic to our optical portfolio. And we spent about four to five years investing in a new platform that would be differentiated both from a signal point of view and also signal and noise point of view and be able to be very synergistic from a software connectivity and everything else with our optical platform. And that's the essence of the product that we created. It's called platform. We have brought it out in the marketplace.
The initial feedback from customers is very good and that connectivity that I speak about is quite synergistic to finding small defects. With the introduction of EUV, as you know, the EUV has a much shorter wavelength than the 193 scanners. It can print smaller things on the wafers that are defects. So you do need high resolution, and for that reason we went with an architecture that has extremely high resolution, a very high beam current density and at the same time it has very high landing energy. And therefore what it can do is it is the fastest and highest resolution product that we believe is in the market.
But more importantly, as I said earlier, by itself, point product by itself would not be attractive for KLA. What we wanted to do is bring a product that would be very synergistic to our optical portfolio and support it to solve the five nanometer and three nanometer defects. And so far, the engagements we have done with customers show that we are able to do that.
Not substitutional markets, complementary, right? So if you think about total pattern inspection, 80% to 85% is served by optical capabilities because of just the requirements for speed in a production fab. So to Amad's point, on a point product basis for the market, not all that attractive, but trying to drive and use it as a reference tool, given the setup challenges at the most advanced nodes on your optical tools to try to drive yield learning in production, being able to couple the tools to be able to drive higher relevancy of your total inspection capability was the big driver strategically for us. That dynamic isn't going to change around the market, but because we had this optical based platform and two types of tools to be able to meet customer needs, trying to use it as reference tool to drive faster and more efficient setup and recipe optimization was the big driver for us. So market reception has been pretty good.
We're excited about the opportunities that
And based on sort of the customer feedback and what you know from a customer technology road map standpoint, do you feel like that 8080% to 85% optical versus 15% to 20% e beam kind of balances here to stay for a couple of years, several years? Or do you still see?
I think it stays in there. I mean it depends on how the customers are buying, so you see it move around. But I would expect to see more growth that in the underlying pattern inspection space, which again, without scaling, some of the technology that drives for smaller and smaller defect detection was not necessarily as relevant that you would see in an EUV environment that's driving scaling. So I would expect it to stay complementary as I described it in that ratio.
Got it. Your services business, great business. It grew 11% your semi process control service business grew 11% last year in a year where obviously wafer starts were down and there were challenges on the memory side. Is that outperformance simply a function of roughly 70% of your business being very recurring in nature? Is there kind of more to it?
Well, it's part of it. I mean the one thing that's, I think, surprising to people is the useful life of these tools as they go into the field. And our business behaves a little bit differently than a lot of our peer companies because customers the complexity, first of all, is one of the issues. It tends to be an inelastic buy. So what that means is customers buy what they need and they try not to buy more.
And then they buy what they need and they run the tools really, really hard because of the ability to monitor that process with long cycle times and lots of wafer starts. And so as a result of that, they rely on us to drive uptime in the tool. So it's a little bit different than if you've a bunch of process tools and you can offload capacity or you could service them yourselves because they're not that complex. So that tends to drive more of a captive stream and because of that reliance, they tend to like the service contract structure because it's very hard to develop your own capability to support that kind of tool set across multiple generations. So that tends to be the biggest driver to it.
And then these tools live on for a long time. I mean, we are servicing tools we shipped in the late 80s and early 90s still. So there's a nice stream to that and then customers like fab wide support contracts. So even in environments where CapEx is lower, they still run the install base to be able to deliver market windows. Sometimes they idle capacity.
We saw it in memory a little bit in the first half of the year, which means the tools don't break as fast. You don't have to fix them as often. But over time, the installed bases tend to drive profit. Those are the things that have been already depreciated by customers. And so there's a lot of demand across the trailing edge nodes, and that drives a lot of profitability for our customers.
And so
that drives the value proposition of service.
Okay. So you guys are still servicing TENCORE branded?
Oh, sure. I mean I throw out the ratio. It's about 80% of what KLA and TENCORE independently have shipped since they started is still in service today. It doesn't mean we service every one of them or that everyone's under contract, but it does show you that they run these things for a long
Got it. Shifting gears, Orbotech. It's been about a year since you guys closed on the deal. Looking back, any surprises to the positive side, negative side?
Yes. I think we're excited about a couple of things that we've been able to talk more about, and we spent some time at Investor Day talking about the specialty semiconductor market, which is a great market that Orbotech had, servicing packaging markets, MEMS markets, RF power supply. So you've seen they designed uniquely for these markets that were nichey but have inflected. And so that's been a good driver for them both from a top line point of view, but also from a profitability point of view. It's got process margins and we do segment reporting around this, but in the mid-50s.
And so it gives you a sense of when you look at the process guys, the gross margins are versus this one about what differentiation is. I always point people to when people say, are you differentiated, what are your gross margins? I think that that's the surest sign of differentiating the market or differentiation. So it's been a great business. It was impacted a little bit by some of the trade dynamics last year given those customers and who they support related to Huawei.
But expectations moving forward are stronger. We'd expect automotive to recover in the second half of the year, and that's been a good driver for them. On the printed circuit board side, they do imaging, which is the leading technology that images, and you see finer and finer pitches that are happening there, which means tighter density of designs. And you're starting to see this evolving integration between the board and how the board integrates with the IC package and the IC substrate layers that are in between. And so revenue synergy opportunities, we believe, for KLA, given some of the things we're already doing plus what they're doing, and especially already having some channel in that market, so how we optimize that.
It does have a flat panel business that's been in a trough. And so there's a lot of focus for us right now in improving the profitability profile of that business. They do have a couple of products that we believe are pretty well positioned, but we've got to make sure we can deliver value and get paid for it. So there's a lot of focus on cost, and we'll see where we go from there. I don't expect a lot of growth in that business this year, but would expect to start to see some recovery as we move into 2021.
So synergy progress has been good. We should be on target for our $50,000,000 commitment by the Q4 from a run rate perspective by the fourth quarter of this year. And I think there's opportunities above that, as I had outlined. There are things we need to do structurally, which we're working on in terms of leveraging low cost manufacturing versus high cost and so on. But over the long run, we should be able to drive, think, meaningful upside out of that.
Okay. Got it. We have a couple of minutes left. I wanted to pause there and just question Mike in the front, please. Is there a mic?
Sorry. Yes. There's one second row.
Recently there was some news reports about ASML being pressured not to ship EUV device to the Chinese semiconductor maker. Just curious if you take a step back, what's the U. S. Government's attitude towards the Chinese semiconductor manufacturers? And how would that kind of impact or how would the U.
S. Government try to use the SPE providers as potential kind of leverage against them?
Well, think one of the I mean certainly there's a lot of noise there and we've been engaged with industry trade groups plus Lam Research and Applied Materials in terms of how do we engage with the government to try to educate in these areas. I think that one of the sort of core issues is where leading what's the intention, what's the goal, right? Where leading edge roadmaps are versus what's happening in China today, what's progress against those and then how do you best sort of address what are legitimate concerns around IP and other things that the U. S. Government is interested in.
I think one of the things we have to be careful about is that we don't do things that are defensive in terms of undermining our ability to compete and our ability to deliver leading edge capability to customers. On the leading edge. And anytime I think you start to restrict and there's the economic dynamics, there's enabling of competition and all of those issues. So we're talking through those issues. And I think so far, I think we've been able to get people to think about these things.
There's also multinational versus unilateral kind of approach to things or multilateral. And so anything that impacts just U. S. Companies doesn't necessarily solve problems or foreign alternatives. And so I think the government is also thinking about that as well.
So I think that's where we're at. It's sort of fluid and we spend, as Rick likes to say, he's been to Washington more times in the last year than he has in the previous twelve years as CEO, right? So but I think it's important for us to articulate what's really going on here about procuring chips versus manufacturing chips, leading edge roadmaps, what's really happening and so on.
One last one.
Yes, second row. Is
there a gross margin tailwind from the advanced process nodes and EUV in the mix?
I would say that our gross margin, as we outlined in our model, and a lot of that has to do with the mix of products that we sell, not every product has same margin profile, and we sell a suite of products in the process control space. So over time, I would expect to see new platforms that we're delivering to market to support the EUV transition for us to drive better cost profile and better optimization of both cost, supply chain, efficiency of build and so on. As it all blends through, I would expect on growth for the company, we should be able to deliver 60% plus incremental margins on revenue growth. And I don't see how that would ultimately or that would change in any way. But you do have some of the mix dynamics that tends to impact how we perform in any given quarter.
But over time, I think 60% is a reasonable way to think about it.
With that, I think time is up.
Thank you, gentlemen, for joining Great. Thank Thank you for having us.
Thank you.