I'm going to go ahead and get started. I'd like to welcome everyone, I think, to the last session of the morning. It's my pleasure to introduce the management team of KLA Corporation for our fireside chat. To my immediate left is Bren Higgins. He's the Executive Vice President and CFO of KLA. We are in a fireside chat format. It's a pretty intimate group. If you've got questions, please raise your hand. We have mics that are traveling around the room, and we will definitely get to you that. With that, Bren, first and foremost, I want to thank you for coming down to the desert after the holidays.
Yeah
to support the conference.
No, thank you for having us.
Greatly appreciated. I always find it helpful, kind of in this fireside chat setting, to make the first question kind of an open-ended question to allow you to position the company. The core IP, the core strategy, the markets you exploit, kind of help set the investment table for the room.
Sure. Before I get there, I'll just make a couple of comments. First, I'll do the safe harbor that I'll be making forward-looking statements. Those statements are subject to risk. You can find a list of the risk factors in our SEC filings, and they're available on our website. 2019 is a very eventful year for KLA. It started the year as a slower period of time for the company, and certainly expected the industry to be down 15%-20% on the semiconductor process side, wafer fab equipment. As we progressed through the year, we saw sequential growth through the year as we expected, and saw the second half come on pretty strongly, really related to a couple of factors. Certainly, the leader in the foundry space came on very strong, much stronger than we expected, stronger, I think, than they expected as well.
That drove nice results in September and a guide up into December. Also, some improvement in the China forecast and shipments into China, where despite the headline risk and noise, we did think that there'd be a little bit of digestion in 2019 compared to 2018 business levels, and it looks like it's going to be roughly flattish. That was positive as well, and I think as we're finishing the year, we probably see the industry down 10%-15%. Against that backdrop, KLA, if you take the guidance for Q4, should have a year of slightly better than 2018. A down year for the industry, and a slightly up year for our process control business. We're excited about that. I think it's indicative of a couple things.
First of all, product positioning, but also the investments in foundry and logic that are happening and how they spend on process control, which is a sweet spot for our business. We closed the Orbotech transaction in March, and so we've been working pretty aggressively over the last six months or so in integration planning. I'm excited about the opportunities in those businesses. We outlined a lot of opportunity at an analyst day we had in September, talking about not just our semiconductor process control business, but also the opportunities in the Orbotech specialty semiconductor business, printed circuit board business, and the combined sort of service opportunities that are in all of it. So we announced some capital structure announcements. We raised the dividend for the 10th year in a row back at that time, so we're at $0.85.
It was 13 per quarter, 13% increase, and our board also authorized an additional share repurchase authorization, signaling, I think, the strength of the plan that we presented and the business model we put up as we talked about long-term growth objectives of 7%-9% top line growth and 1.5x earnings, 1.5 the revenue growth rate in earnings growth. We feel really good about where we're at. I think in terms of the December quarter, things are shaping up as we expected, so I don't have any real changes to that. I think from a top-line perspective, margin cost, everything's lining up as I thought, so no real update there. 2020, and I know we'll talk about it a little bit more, but we see an environment where logic foundry has some stability and sustainability into next year.
We see an end market environment that's attractive for those customers. We see competitive dynamics that are attractive. I think in China, we see a flip where more logic investment than memory. We think that it's a great year this year for logic foundry, and we think there's some sustainability to that into 2020. For memory, I think it's a better year for memory, as we said on the call. I think it's probably more second-half dynamic, probably more NAND flash-weighted, I think, versus DRAM in terms of expectations. I wouldn't say we have any really unique insight there, but certainly some of the indicators we're seeing in terms of utilization rates in the service business and so on give us some comfort that we'll start to see some recovery in memory spending sometime in 2020. Overall, flat to up mid-single digit type outlook for 2020.
Against that backdrop, given the foundry logic composition, I'd expect the company to probably do a little bit better than that. I feel very good about, I think, the outlook. I think things haven't changed all that much since where we were at earnings. In terms of your question, I like to characterize KLA, maybe I'll just walk through our strategic objectives as a company. The first thing is about market leadership, and if you know KLA and our history at KLA, it's about market leading positions. Our market share is 4x our nearest competitor. We spend more in our space than our competitors' revenue levels. We introduce products faster, so we're a moving target.
Introducing products that we believe are high-value problems for customers, that we can help solve them, is how we think about markets and market opportunities, and we invest to ensure that we can deliver on that. Differentiation, product differentiation is another objective for the company, and we measure that by market share. People talk about having innovative businesses or having differentiation in the market, the first thing I do is say, "What's your gross margin?" If you look at the gross margin of the company, gross margin is the clearest sort of signal of differentiation in the market. KLA gross margins are in the low 60% on our process control business, 63% this year. I think that reflects the strength of the products.
Look, process control is a business where elasticity of demand is sometimes customers buy what they need, and so you have to have a deep understanding of the value you bring, the returns your customers get, and be disciplined about trying to share in that value. I think the company does a very good job of doing that. Operating excellence, again, the business model is driving. We measure that in two ways. We measure it around operating margins, but also incremental operating margins. We think on revenue growth, for every dollar of revenue growth, we ought to be able to drive at least $0.40 of operating income growth. We're more than a great products company.
I think how we execute the business operationally and how we look at opportunities in acquired businesses to get more out-of-market leadership positions, we have a history of doing it, and we believe we can do that with Orbotech. Finally, talent, and it all starts with the people at the company, creating opportunities for those people. I think we have a global approach, both in terms of manufacturing, but development. We're opening a new site in Michigan. We're excited about that, to tap a new R&D pool. Attrition rates for us in Silicon Valley, despite the competition, are about half of what Silicon Valley rates are, and I think that's illustrative of the opportunity and the complexity of the problems that people solve at KLA, we're pretty pleased with that. I think that's our competitive mode, if you will.
I think that's all part and parcel to the IP of the company, and we're excited about the prospects moving forward.
Kind of a great positioning of KLA. Bren, as you know, with semi investors, depending on the week, day, sometime hour, sometime minute, sometimes bad news is viewed as good news and good news is viewed as bad news. We're kind of in this weird condition where all the strength that we've seen in foundry logic this year is leading to a lot of concern about sustainability for next year. I guess I have a couple questions around that. TSMC, the largest foundry spender out there, they clearly significantly raised their CapEx for the year on their Q3 conference call. If you just kind of annualize their Q4 spend, you kind of get to a $20 billion CapEx number for them. If you scale that up to WFE, you're sitting high 50s, close to $60 billion of WFE.
Clearly, those sequential sort of rates can't be sustained, I don't believe. I'd love to get your opinion on that, one. More importantly, there's a concern out there that when you look at logic foundry intensity, we're back to where we've historically seen kind of peak-ish levels. I kind of want you to focus more on that side of the answer, because you guys have talked a lot about the resumption of scaling with the insertion of EUV, I want to get a better understanding of how that benefits your business.
Yeah, no, those are great questions. To your point on the math, I just talked about an outlook for WFE that's flat to up mid-single digits, and certainly, the math implies something different. Look, I think the beginning of the year was fairly weak, so we have seen some strength in the second half of the year, and it's been mostly isolated to one customer. I think as we look at 2020, we do feel like there's a broadening of the investment profile, the competitive dynamics of the leading edge, some of the trailing edge activity. Look, I think there could be some moderation to the point you just made. Overall, we think that not only in the shorter run, just sustainability of the investment levels sort of on an annual basis, but over the longer run, that given the scaling opportunity that's now happening.
If you go back and you look at the last, from 2013 or 2014, industry achieved node transitions through means other than scaling, right? These architecture changes. multi-patterning wasn't really a scaling dynamic. It drove more process steps. The lack of scaling has, I think, impacted how end markets have thought about transitioning from node to node, just in terms of power and capability and cost, but also how customers look at buying process control. If the end market adoption is not strong, then customers will try to optimize their capacity and try to use tools at the next node and do things like that. Certainly, 7-nanometer has been a very strong node. Lots of design starts, and we've seen that strength in our reticle business, where each unique design has a unique reticle set.
Those reticles have to be inspected before they're used in the scanners to start to print the wafer. We've seen strength there, and that's been good. I think it surprised the customer a little bit too, in terms of the strength. As we move to five-nanometer, you start to introduce scaling again. We start to see the adoption of EUV. End market drivers, we think, are pretty compelling that influence, I think, the adoption profile we'll see from an end market perspective. I think a big part of how we thought about the go-forward plan that we talked about at our investor day was the introduction of scaling increases the relevance for things that KLA has always gotten paid for, is finding smaller defects in chips. Optical inspection capabilities have, I think, underperformed the market.
WFE really hasn't grown in logic and foundry. Our expectation moving forward is that you'll start to see scaling that will drive end market adoption. It will drive mix towards some of the traditional KLA flagship products that have very strong market positions. We'll see WFE growth over time that's going to mimic somewhere around where semiconductor revenue is, and that's a real positive for us. There are new product offerings we got to get in the market we think help enable that position. A big part of our 2020 plan is to be able to feed the market with those new products we talked about at our investor day.
As we start to move through 2021, 2022, and beyond, and you start to see significant EUV adoption, significant scaling, you start to move into even High NA or higher resolution EUV, that it will create more opportunities for the company, and drivers that are very different than what we've seen over the last few years, where all the growth has been very memory-centric. We've improved our position in memory, but logic foundry spends a lot more on process control per WFE dollar than memory does.
Bren, I was fortunate enough to host you and Rick a few weeks back in New York for some investor meetings. One of the things that came up was this idea of reuse and the rate of reuse. I think it's an interesting dynamic because in a world where scaling kind of paused, it seems like your business in particular was hit harder than most with reuse of your tools as you went down process nodes. It seems with the resumption of scaling, that the reuse aspect is coming down dramatically. Am I characterizing that correctly? Is there any way to kind of quantify what reuse had been for your inspection tools and where is it going?
Yeah. No, it's a good question because it was scaling, but it was also adoption beyond a few customers with high volume products of those nodes. As they moved on to the next node, and there wasn't a scaling driver to drive higher levels of sensitivity at commensurate throughputs, that those customers were able to take tools and try to move them to the next process node. In process control, certainly how they use tools, they don't sort of set a tool line and then things never change like they do on the process side. The customers are constantly moving tools around, and depending on where they are in the maturity of a process ramp. Given those dynamics, given the fact that there weren't many end market, that the design starts were pretty low, each design start carries its own unique defect characteristics.
Our customer has to deliver to a tight market window and deliver volume in those windows. That drives a higher level of investment. What we've seen, and certainly saw at 7, is a lot of end market adoption, which has been good. As the customer starts to prepare for 5 nanometer, A, the challenge of the fact that that capacity is in use and will have, I think, a long tail as design starts are only increasing at that node and incremental capacity being added there, but also at 5 nanometer, they have the EUV challenges associated with scaling. It'd be hard anyway because just the new capability on the tools, but it's even more challenging given the scaling dynamic coupled with the sort of overall demand for the node.
Is there any way of quantifying that? If we were at reuse levels 4 or 5 years ago that you're expecting to see now, would your logic foundry quarterly revenue have been 10% higher, 5% higher, 20% higher?
That's a good question. It was pretty significant. Think about 20 nanometer to 14, 16 nanometer. You basically had a transistor architecture change, no real change in lithography. I would say, given that you only had a few products up there, reuse was probably greater than 60% on some of the capacity purchases they had already made. I think today, we're probably down 25%. I know we haven't really sort of thought through the actual math because a lot of it's dependent on the number of wafer starts that ultimately get added. I think it's sort of in those ranges.
As you pointed out, the upside in foundry logic this year has been mainly one customer. As we think about the broadening out, and I'm thinking more of logic, how do you think about logic kind of tracking from here?
Well, the last few years have been a little more challenged in the logic area. There's a number of customers there. I'm not talking about anyone in particular, it's been weaker, I think the challenges have been pretty well chronicled out there. As we move into next year, I think some of those problems are getting resolved, we're starting to see more activity. We're seeing more activity in terms of use of capability to debug older node processes, certainly on higher-end designs and die, also more activity on the development front for the next node. It's one of the factors, I think, given the competitive dynamics there, that we believe that we'll see more strength into next year. I think given what we've seen and hearing publicly, some changes in strategy and investment to get on a tighter cadence.
We're encouraged by the potential. With a certain amount of spending in the past, that gets spread over multiple years, right. As you start to see that start to tighten up, that same level of spending, and with some of the process challenges we've been talking about, we start to see it in a tighter timeframe. We're encouraged by that. In other parts of logic, I think sometimes people don't realize what's happening in automotive, industrial, and IoT, and those markets. We have some exposure through the SPTS market, but even on the process control side of the business, that there's a pretty healthy level of business that's out there. If I think about our business historically, I'd say 75%, 80% of our orders were always about whatever the leading-edge node was.
Today, I'd say it's probably 65%-ish, plus or minus, our leading edge. There's a pretty solid level of business that's coming from these efforts. Obviously, China's part of it, but there are also other customers that are also making investments. What I think the thing that's exciting about sort of the stratification of demand from leading edge to trailing edge, and as semiconductors become more and more significant in all of these markets, is creating a lot of opportunity, and it's great for our service business also because there's tools in the field that they're augmenting, they're upgrading, but also that they're running harder, and those require more service and support. It's a pretty healthy environment as we sort of think about it kind of broadly like that.
Can you talk a little bit about your memory exposure? If I'm doing the math right, 50% is almost half. Yet there seems to be a perception in the investment community that you're not all that levered to the memory market. Can you talk about the growth prospects for KLA in memory?
Particularly in the flash market, we've seen from planar flash to 3D flash, some improvement in process control intensity in a couple of areas. I'd say it's probably a couple of points of intensity. Let's say it was 8%-9% to maybe 10%-11% of WFE was spent on process control. Our market share generally across the different segments is pretty consistent. When you think about that, we're at 52% of the market. In the markets we compete in process control, we're probably in the 65% range. Of that 2% increment, we get 65% of the pieces that we're in. We saw a step up as you went to 3D structures in a few areas, in metrology requirements, both around shape and profile of the stacks, the wafer flatness.
As you're starting to build layer stacks, the flatness of the wafer matters a lot. You don't want the stacks to tip over. Defectivity control was a challenge, and so customers are running more monitor wafers through unpatterned inspection systems to make sure the tools are good and clean before they start to process. Across all those businesses, we saw a nice level of improvement there. We are introducing a new product that addresses an additional challenge that we think could probably improve another point and a half of intensity to do channel hole measurements. There's a channel hole that goes up and down the stack. Getting that precise is a real challenge for customers. Because it's a measurement that's within the film stack, it's opaque and it's very hard for other more traditional metrology techniques to do the imaging.
We're developing something that we've been working on for a long time that's coming to market. There's been a couple tools in the field already that we talked about in Investor Day, that's an X-ray metrology product. Today, the way customers make that measurement is they do destructive techniques. They basically cross-section the wafer and offline go make that measurement. You might imagine that's a slow process, takes a while to get the kind of results. Being able to do that in-line is going to be, we think, pretty compelling and will create some opportunities there. On the DRAM side, an introduction of EUV is coming. I think that will be a driver for intensity. DRAM's been a little bit more unique in terms of a market with redundancy and repair and so on, that the intensity levels haven't changed all that much.
I think share position's been pretty good. Over time, as you introduce EUV, we think it's going to drive some of these scaling dynamics and high-end inspection requirements we think will play well for our business. We're doing better there, and I think that even in China, I'd argue that the intensity is even higher with those customers trying to progress their nodes perhaps a little bit faster, are monitoring and sampling more, that's been a good driver for the business as well. I think all those factors have influenced what we'll call the process control intensity of memory.
Bren, I think one of the optical concerns that are out there is if you look at your two U.S. equipment, process equipment peers, their quarterly revenue is still running on the equipment side, roughly 20% below prior peak, which makes them appear to be really geared up for when memory starts to come back. KLA, if you do the same sort of analysis, I think your revenue's kind of flattish with the prior peak, and I think that speaks volumes to kind of your leverage in the logic market, and I think that's more than just cyclical. I think there's some structural drivers there. Can you remind us from peak where your memory revenue sits today, and if memory starts to come back, what kind of growth you think you'll enjoy in that business?
We've had some level of investment in memory. As customers continue to do technology migration, which is shrinks, they still invest a certain amount of process control. We have had maybe some exposure that the process guys haven't, because they haven't added any capacity. I would say that memory right now is pretty light, certainly in the first half of the year. I don't expect it to change very much in that timeframe. Year to year, my memory business is down 35%, which is sort of in line with overall industry memory decline. I think we've had some exposure, but it's down quite a lot from where it was. We're encouraged by the things we're talking about here and what that will mean as memory starts to turn back on again as we move through 2020 at some point.
The overall semi industry's been fighting the rumor of the death of Moore's Law. You've been fighting the rumor of the death of optical. Can you talk a little bit about the Gen 5 optical product?
Sure.
It seems like that's a robust new product introduction for you. Also, at the Analyst Day, you did introduce some new tools not using optical, E-beam, X-ray.
What's the strategy there? Are we finally at the end of optical over the five and three-generation node, or do you see these tools more complementary?
Yeah, no, it's a good question, and it's funny. I do get that question a lot from investors, and sometimes I think it gets fueled by our customers because our customers will say, "Well, I think optical's dead because I don't have the next extension." We introduce the next extension, and they go, "Okay, this works great." Every day of the week, I will choose an optical inspector for production requirements because it's so much faster than alternatives. When it doesn't meet the need, whether it's signal-to-noise or resolution issues, and there was a period of time that that was the case, that they seek other alternatives. You use an E-beam technology, which is complementary to optical.
Gen 5 came to market, and we're into the second iteration of Gen 5, and we're now preparing for Gen 5. It's been a nice uptick this year, but to be more of the flagship product of the business with 2020 revenue being over 50% Gen 5 versus Gen 4. We're seeing lots of use case opportunity and application for it, and if you look at Gen 4, we extended it six or seven iterations. There's a fair amount of opportunity or headroom there. The challenge I have, and I've said this for a while, my challenge hasn't necessarily been a lack of light. It's been more about signal to noise and being able to take small defects and figure out what matters from what doesn't.
There's been a lot of investment in the company to increase relevance of the inspection input that we're getting, what we're seeing on the surface of the wafer, and how to use different either statistical capabilities or other data analysis capabilities to try to figure that out and speed that time to results, understand what's really a problem from what isn't. We did introduce a new product, an E-beam, because we think that there's a complementary dynamic there that as the tools work together in a scaling environment, will create opportunities for us to increase the effectiveness of the optical inspection tool. We're excited about that opportunity. I think on a standalone basis, the E-beam market is a smaller market, but in conjunction with the broader portfolio of products, we think it creates a pretty compelling opportunity or advantage with competitors and opportunity for our customers.
We're making investments in Gen 6. I think that we think that the extendibility here is quite significant, and the ability to run a wafer an hour versus a wafer a week is a pretty compelling opportunity or situation for customers, particularly in a production environment. If we can now speed the time to results by adding and augmenting with this additional capability, we think it's a good opportunity for us.
Bren, I know the other question or concern you get from investors around your product technology roadmap is actinic inspection tools for EUV pellicles. What's your view of the market, and where are you internally on your investment strategy?
Yeah. Controlling the reticle, which is, again, the stencil, if you will, there's a reticle set for every design, and that reticle is put into the scanner and drives the pattern print on the surface of the wafer, for those that don't know. There is a challenge that with traditional optical technologies in an EUV environment, that you don't have the resolution you need. Certainly today with what's being done today with contact and vias, the resolution is enough. It's working, and so what customers are using today is our existing optical platform. As you move into EUV patterns, it becomes a challenge. Then the issue of pellicle, which is a protective layer on top of the reticle that protects it against particles, becomes a challenge for customers to be able to control that defectivity.
Adoption of pellicles not clear and getting it right because it absorbs energy and there's a lot of other physics issues associated with it is kind of a challenge in the market. One of the challenges with actinic is there was a challenge, actinic means an EUV inspector, is the same challenge as you had with EUV that source powers can be a challenge to get to resolution. You still have those issues. That's, I think, one of the biggest problems is can you actually get resolution that beats whatever the alternatives are.
We talked about and we're launching this year, we'll ship this year, actually expect to maybe even see revenue on the first tool this year of a multi-column E-beam tool to do the inspections in the mask shop of pre-pellicle, before the pellicle is installed, if it gets installed. In a lot of cases, it won't. We feel very good about that roadmap between the Teron system, between what we have with this multi-column E-beam tool that has a resolution for it to do the critical sort of first couple of inspections in a mask shop before the reticle ships out with that roadmap. We think there's an actinic need in the market. We're investing in it, but we don't think that need is really out there until probably the timing around High-NA, which is high-resolution EUV. That's probably in the 2022, 2023 timeframe.
As we know with EUV, sometimes this technology is very complicated. The schedules tend to push out more than pull in. We think the roadmap that we have is right. We'll have this tool that'll be out there solving these customer problems. We have some unique capability around guided database inspection, which is an important aspect of these tools in terms of how they inspect reticles for customers. We feel very comfortable about this tool and the roadmap over time in terms of solving the reticle sort of ecosystem requirements of EUV. In a fab, they will do print check. What they'll do is they'll use a Gen 5 tool with some new algorithms that take the design information and then inspect and qualify the reticle after it's been put into a fab. There is a solution in place.
We think it'll be a driver for Gen 5, and there's an increment for it. The reticle house sort of challenge gets solved by this capability we're talking about. We'd expect to have an EUV-based offering sometime in the timeframe that I talked about.
I wanted to give you an opportunity, a little bit of talk about the Orbotech acquisition. You guys have always been actively seeking out M&A. I think you've had some concerns finding the right targets, given how strong the core business is. I know you've talked about process intensity in other industries not being as high as semis as you look to kind of move left or right. What was appealing to Orbotech? I guess more importantly, as you integrate the asset, I know you'd rather underpromise and overdeliver, but when I look at the target model you put out at the Analyst Day, it seems relatively conservative with what your historical operational acumen would suggest.
Well, that means there's upside. Yeah. What we've tried to do is be pretty straightforward, I think, with the way to think about the plan and a plan that was less dependent on market drivers and more sort of self-help or dependent on our ability to go execute. We're excited about the Orbotech business because you have rising complexity in some markets where they have market-leading positions. If you look at the specialty semiconductor market, that's a market that's inflecting. You have products that have been designed specifically for RF, power supply, MEMS markets, packaging markets. There's opportunities for us to, I think, augment that product portfolio but also drive a higher level of engagement between them and some of our key customers, and we've had some strong reception there.
In the PCB business, it's imaging and inspection. Also you're starting to see a harmonization between PCB boards and how they blend in with advanced packaging and IC substrates. There's opportunities for us to leverage technology in both parts of the company to create new SAM. We talked about that at Analyst Day. They also have a flat panel business that we're doing a lot of work on improving the cost structure. People ask me and say, "Well, the ratios are different." We don't really manage the business for the ratios necessarily. What we're looking is for market-leading positions with market increasing complexity that we believe we can drive more value out over time.
Part of our operating system is not just pricing, but operational execution. I do think in the Orbotech business, we ought to be able to drive the same level of incremental operating margins as we drive in our core scale of the process control part of the business. I think there are opportunities there. We're focused on them. I'd like to see that we should be able to do that. I talked about a synergy plan that was ahead of the $50 million that we talked about. We talked about revenue synergy for the first time. Some of that is structural cost improvement by moving things from high-cost regions to low-cost regions, but things we've done in the past. We have proven that we can improve. I talked about an example at Analyst Day, that we can improve acquired business profitability.
A particular business we bought in 2008, over the last 10 years, we improved the gross margin 14 points. Seven points were operational, seven points were from pricings and sort of top-line dynamics. We're pretty confident about it. I'd like to think that we can for sure deliver on what we've said, but also maybe even do better than that.
Perfect with that, we've ended our time here with the session, but I want to thank everyone for joining us, especially Bren, for spending some time with me on stage this morning. Appreciate it.
Thank you, John. Appreciate it.