Hello, everyone. My name is Mehdi Hosseini, senior analyst at Susquehanna International, covering technology hardware. Today, as part of our tech conference, we have a fireside chat with the team from KLA. I have Oreste here to help us with the several topics that we have prepared. We also have the IR team. The format here is that we will go through these topics and questions that we have prepared, and if there is any question from the audience, please feel free to shoot me an email, mehdi.hosseini@sig.com. With that, I want to thank the entire KLA team for giving us an opportunity, and Oreste, thanks so much for carving out some time to go over some of these very exciting topics. Perhaps wanna just dive in and put things in perspective.
It has been about three years since KLA acquired Orbotech, and two years since Orbotech and SPTS were consolidated into the organization, and we now refer to it as EPC. Can you give us an update as to how the integration of these assets is going on?
Mehdi, thanks for having me. I'm really happy to be here with you guys today, and happy to tell you a little bit more about EPC and where we are with integration and the progress as explained. As you said, we completed the acquisition of Orbotech and SPTS exactly two years ago. It was February 2019. Time goes very, very fast these days, and we are very, very happy with the progress that we have made since then. If you look at the revenue and profitability targets that we outlined at the time of the acquisition, and we presented Investor Day, New York City in September 2019, we are ahead of these targets.
Another thing that I'm very, very impressed with all the organizations that we acquired and eventually integrated under KLA operating model inside EPC is the way how the employees embrace the KLA value and all the processes we have. I believe the integration has been successful for a few reasons. First of all, we have been very, very clear with what we were trying to accomplish with the acquisition of Orbotech and SPTS. SPTS, I would remind investors, was an independent subsidiary inside the Orbotech. It came with the Orbotech acquisition. We wanted to expand our SAM. We wanted to enter new markets. We wanted to enter new markets in the electronics ecosystem, in particular printed circuit board and display.
We wanted also to strengthen our portfolio in packaging, especially semiconductor, with the additional process tools like the SPTS vacuum-based PVD, CVD tools, in addition what we have developed for many, many years at KLA, that means inspection and metrology. We were very, very clear. The second thing is we wanted to buy organizations that were leader in their sectors and a organization that can benefit from KLA core competencies or even a market presence. For example, since then, we have done some work together to improve the technologies of the former Orbotech divisions, Orbotech product lines. We introduced, for example, optics designed by KLA in the next generation of PCB automated optical inspection tool. We also use some of the KLA simulation and modeling capability to help with direct imaging photolithography development.
We initiate collaborations with top semiconductor customers in packaging, including some offerings from Orbotech and SPTS. These organizations, of course, previously had no interaction with these customers. All these things that we have done in the last two years, three years made this integration more successful. Of course, we are still along the course of the integration. We are not done yet. Finally, I would like to mention probably the secret sauce of the integration is the KLA operating model that I explained during the Investor Day in 2019.
Okay. Now, one thing that I asked you a couple of years ago, I'm gonna ask you again, it's KLA's core competency in front-end, in process diagnostic inspection and metrology. What I wanna get an update on, like if you were to focus on like a packaging yield management, how are you going to take the leadership that you have in the front-end, in process diagnostic, and apply it to advanced packaging and basically standardizing the back-end processes? Is that even possible?
Yeah, that's a great question, actually. One of the main focuses of the EPC organization is, of course, advanced packaging. Before answering your question, let me say a little bit about, first of all, how the advanced packaging is progressing and our role in advanced packaging, and then I'll respond, if you don't mind, with a direct answer to your question about standardization packaging. First of all, let's talk about overall packaging. As you know, the ability of scaling transistors like Moore's Law is slowing down and it's becoming very, very expensive. It doesn't mean that Moore's Law is dead. In fact, it's not.
In fact, our advanced customers are progressing towards 3-nanometer or even beyond 3-nanometer, introducing a lot of very innovative technologies like EUV photolithography, going to the next level of transistor architecture, gate-all-around. We still see scaling in the front-end of wafer fabs. However, it's not enough anymore. That's the reason why advanced packaging is being used as a augmentation of a front-end technology roadmap to make sure that the entire semiconductor industry can advance. Advanced package become a key enabler. We see a lot of top semiconductor companies now heavily investing in this area. Let me give you two examples. First of all, heterogeneous integration. And this is at the front and the center of advanced packaging technology roadmap right now.
With the heterogeneous integration, we allow the aggregation of multiple chips on a single package. In this case, everything gets integrated in a single package with the help of very advanced substrate. This is leading a lot of significant process control and process integration challenges. In particular, when you try to integrate multiple dies, multiple chips on a single package, and you want to do like the next generation, heterogeneous integration will be bumpless, so will be direct copper-to-copper bonding, what we call hybrid bonding architecture. You will see even more challenges in terms of small defects detection, die placement overlay, warpage measurement, and overlay die level cleanliness.
This has, of course, a significant impact to our business, positive impact to our business, because now the challenges that we have seen in front-end semiconductor process control for a long time, now they are converting into similar challenges in packaging as well with all these, heterogeneous integration, in particular the hybrid bonding architecture. The other thing that is happening in packaging that is something I like to talk today because I believe has not been really understood, is the role of IC substrates. IC substrates, just to give you an idea what the substrates are. Whenever you place multiple chips in a package, you need to land the package on a substrate before connecting to the PCB. These substrates are acting as an intermediate layer between the chips package and the PCB. These substrates are evolving a lot.
In particular, they are important for FPGA, CPU, GPU, for all the computing chips. These substrates are done in PCB fabricators right now. I want to say this is like the high-end PCB technology, and we have seen an unprecedented tightening of substrate supply right now. Sometimes we talk about chip shortage in the world. Sometimes the chip shortage is very relative to substrate shortage, and you don't have the substrates where you land your packaging chips. Subs are becoming extremely important because they are becoming much tighter in terms of geometry and also the complexity overall in fabricating the substrate is becoming really challenging for the PCB fabricator.
With Orbotech, we have an opportunity because Orbotech, of course, is in the PCB space and now is fully integrating PC group to play in this segment, developing very interesting products to help with the fabrication of these substrates. This is the second portion of the packaging, the challenge I want to discuss. The third one has nothing to do with the computing chip, but it has to do with another important trend that is 5G. The adoption of 5G network communication protocol in smartphones is driving significant innovation as well in packaging and boards. In particular, I think maybe you do a lot of research, I know you are very well aware, there is a big inflection of what we call SiP, system in a package.
Whenever you transition, for example, the 5G communication to millimeter wave, there is a new integrated RF front-end module that is designed including many components, power amplifiers, switches, transceivers, antennas in a single package. In addition to reduce the form factor that is by itself a big challenge, you also have a lot of complexity in the design of this packaging and the boards supporting SiP, where we see our ICOS business and PCB business getting more and more critical. You see like an incredible inflection in packaging and subs and PCB driven by all these trends happening right now in electronics. Now, going back to your question, it is a very good question. As you know, I joined KLA a long time ago, 23 years ago. Before KLA, I was working in semiconductor fabs. I worked specifically in Texas Instruments and Micron.
I started working in semiconductor almost 30 years ago. At the beginning of my career, when you talk about process diagnostic control, there is nothing. There was nothing. I remember we were in the clean room, looking at wafers with optical microscope, try to find the defect. When we find the defect, everybody was happy or unhappy, depending on which process area you get the defects from. It was not real. The semiconductor industry 30 years ago, 25 years+ ago, was not really set up to monitor the line and do a proper process diagnostics. KLA 25 years ago invented in semiconductor wafer fabs the concept of inline monitoring and yield management system. It was, in reality, a more rigorous way to split the fabrication process in multiple modules.
At the end of each module, we have an inspection point or a metrology point. This concept of process diagnostic control or inline module with the management system became a standard in semiconductor front end. Now, standard doesn't exist in packaging. Standard doesn't exist in substrate and PCB. Can we do something? Yes, we can. When you look at what happened, for example, in wafer fabs, this transition from optical manual inspection to fully automated feedback, feedforward control loop based on process control, was gradually done. It didn't change in a day. It was done in the last 25 years. We are starting now partnering with the top five customer packaging companies in implementing the same concept in terms of data control and data analytics.
As you know, we have our software platform that everybody's using in the wafer fabs. It's called Klarity and 5D Analyzer. These are software solutions that KLA has been developing for many, many years to make sure that all the data are aggregated in semiconductor fabs and are used for process control optimization, feedback and feed forward control loop. Generally, on top of the software infrastructure, we build AI and machine learning analytics. We are starting to have this conversation right now with the packaging customers.
First of all, to standardize the output of the tools, and then after the output of the tools is standardized, like SECS/GEM or something similar to what we do in semiconductor wafer fabs, we can create a software suite, and then we can build algorithms on top to have much better control of the line in packaging. Eventually, we will build a similar thing later for SPTS and PCB. This is a fantastic question, Mehdi, and this is something that, frankly speaking, I'm spending a lot of my time and my money, because I'm really passionate about it.
Sure. Essentially, these opportunities presented relates to non-good die, and non-good die has historically been the challenge. Now, I don't wanna deviate from well the topics, but I think one way to better understand opportunities for KLA and as a follow-up to that standardization thought process, who would be your customer, especially with hybrid bonding, advanced substrate, silicon package, and we're dealing with a non-good die improving the yield. Does that create opportunity as you would sell systems and solution to multiple customers? Or you basically are offering data analytics to connected data, but ultimately, there is one single customer.
No, it's not a single customer. If you look at what I said before, inflections in packaging and substrates driven by all these trends happening in the digital society, this is, first of all, we sell products to our customers, and this is not only one. These are many customers in packaging, many customers in subs and PCB. From the data analytics point, we will sell to the same customers because eventually the advanced packaging line or the substrate fabricators, they need to control their lines, they need to control their quality, and they need to make these automated. We will sell these software solutions to the same customer we sell our hardware product to. On top of that, there is an opportunity. This is a very good point, actually.
There is an opportunity to broaden our collaboration with the entire ecosystem. For example, fabless. Fabless, they design their own chips, and generally, the wafers are built in foundries and packaged either in foundries or in OSATs. There is an opportunity for us, and there has been for a long time because we work with the fabless, not only with direct customers, to understand how we can help the entire ecosystem to talk to them about these new software solutions to make sure that they may have some value, or they may see some value from the data that are generated in the fab. Of course, this is a part of their agreement with their suppliers. KLA is absolutely agnostic to their agreements.
there is a possibility by developing some software analytics and some analytics that can create more value out of the data coming from the tools in the packaging line, that the fabless will have a different type of agreement with their suppliers, either foundries or OSATs, and eventually they can leverage this data.
Sure. When it comes to how KLA could leverage relationship collaboration with customers, these additional services or solution becomes critical. To that end, it was AMD that finally brought the concept of SiP or chiplets to the headlines. We have known about SiP for maybe more than two decades, but one fabless company commercialized it. Now, looking into next year, the other leading logic IDM based in the Bay Area wants to do that. This is where like KLA could be. I could argue that KLA could catalyze the chiplet opportunity for this other logic manufacturer. Actually you can become a part of success stories as they catch up to AMD. Am I thinking about this the right way?
Absolutely. In fact, it's a matter of fact, again, we have two roles in all this innovation and inflection. One role is as providers, and the other one is, I would say, solution providers through our data analytics platform. Again, the data analytics platform, of course, if you can leverage the data out of a fab, a packaging line or assembly line or a substrate fabrication line, so you are better in determining what the problems are in the line, how you control it, how you monitor it, how you solve. Eventually AMD or other fabless or other OEM will take advantage of it. Frankly speaking, as I said, we have collaboration. We don't sell tools to our fabless. We don't sell tools to mobile OEM. We don't sell tools to automotive OEM.
Maybe we can talk about automotive later on. We leverage the collaboration to make sure that they can get the best out of the tool, the best value out of the tool that we ship to their suppliers.
Sure. Just a quick follow-up here as we move on to the next topic. If an IDM has a hybrid strategy utilizing their own fabs as well as foundries, and also trying to scale chiplet, will managing the non-good die give you additional opportunity? Like if they're fabbing GPUs at TSMC, but they wanna do the backend in-house, is there an opportunity for KLA or is this usually-
Absolutely. If they want to do back-end, they start opening their own back-end fabs, they will become customers of ours, where KLA is kind of sort of priority because we have already implemented the same system and the same, let's say, standardization at their current suppliers. Yes, absolutely it's an opportunity.
Okay. One other end market that doesn't get as much attention with traditional semi cap, but everything else is in the headline, it's electrification in auto industry, automotive electronics, which requires new material, new substrate, new way of putting more silicon into the package. Can you help us understand how those opportunities are presenting to KLA? Any detail will be great because I think this is one area that there is a shortage, but I think we're in an inflection point in supply chain changing to accommodate electrification in auto industry. How does KLA can capitalize on it?
Yeah. I'll respond to your question again in a broader way, talking a little bit about automotive electronics first and then the role of KLA, that is, quite unique in the automotive industry. Let me talk about automotive electronics, first of all. As you know, automotive industry is really at a singular shift. Right now we see implementation of electrification, autonomous connectivity. All these trends will require a new approach how to develop or qualify modules, semiconductors and boards. When you look at electrification, for example, it's no secret that the automakers are making significant investment. They want to accelerate the electrification. We hear almost every week that the automotive OEM, they want to have like a full line of electric vehicles by 2030.
They want to get pretty much this transition to electrification as fast as possible. This is not easy because the electrification, although I still believe the electric car is much easier to produce than the combustion engine car, but there are very big challenges from power efficiency point of view. Of course, when you talk about power efficiency in the car, the first thing you think is how can we have different material than silicon to provide better power efficiency in the car? Of course, you think about silicon carbide. Silicon carbide is more power efficient. You can operate at higher temperature, you can operate at higher frequency. It seems to be a perfect compound semiconductor material to handle this challenge.
However, these materials are very hard to manage, and I'm talking about silicon carbide, but I can combine with gallium nitride for some onboard charging. Also, gallium nitride is very, very well used for a 5G infrastructure as well. These materials are new. We have been working with silicon for 50 years. These materials are exotic, are new, are hard to manage, and they are still a very low yield right now. However, they are going to progress. When I look at the automotive power right now and for electrification, I expect that it's going to be a mix between silicon power and silicon carbide power inverter for the next five years, 10 years, and then eventually will shift to silicon carbide for the vast majority of power devices.
When you look at all these compound semiconductor, this is a great example where the new group, DPC, that we formed after the acquisition of Orbotech SPTS, is very, very helpful to the industry because SPTS has been the market leader for PVD technology serving power semiconductor for a long time, and we see an adoption increase for our deposition and etch solutions for both silicon carbide and gallium nitride. Now going back to your question, apart from silicon carbide, gallium nitride, our solution either for process control or process tools for this kind of compound semiconductor, let's talk much more in a broad sense about quality and reliability control in automotive. As you know, I want also to make a link to the known good die question on packaging you asked me before.
When you look at automotive, it has always been the most demanding industry in terms of reliability of quality of semiconductor, of course because very high liability, expensive recalls and so on. For that kind of reasons, the level of scrutiny that you have in automotive industry, you don't have it in any other industry of electronics. The chips are tested multiple times. Sometimes you try to replicate the harsh environmental condition, doing burn-in testing and so on. Despite all this testing that is done in automotive semiconductor, you still have marginal parts, marginal chips that can still escape and entering the car, entering the automotive supply chain. What happen?
What happens is the chip can fail and the airbag can fail, and you have millions of vehicles that are recalled, causing a huge cost and also, frankly speaking, severe legal liability. That's the reason why KLA started getting really interested in automotive semiconductor even before the acquisition of Orbotech SPTS. Now, with Orbotech SPTS, we grew our presence in automotive semiconductor, but even before them, we thought that we have a role to play because we are dealing with quality and reliability. Like we drive the standard with quality and reliability overall in the wafer fabs, and we thought in automotive it was much easier for us to make a big impact, a big presence. What we have done, three years, four years ago, we started talking to everybody in automotive, including the car makers, the Tier 1.
The supply chain of automotive is extremely complex. I think maybe you know. There are automotive OEM, the car makers, and then the Tier 1 preparing the boards, and then the Tier 2 that are semiconductor supplier, and there are the fabless, there are the OSATs. It's so complicated. Also, that's another reason why the automotive was the first industry to pay the price for the chip shortage at the beginning of last year. What happened was we decided to partner with everybody, and we came up with an idea that we called IPAT. IPAT means Inline Parts Average Testing. It's really a clever way based on machine learning methodology to screen the potential reliability failures in fab.
Instead of waiting for the chip with reliability concern getting into a car or getting into automotive supply chain, we came with some ideas how to create a new solution in between electrical test and inspection that creates pretty much it's a software and a hardware solution that uses some of the KLA hardware tools with some machine learning methodology and a software to make sure that you test exactly the chips, exactly at the steps, exactly at the parts of the die that is more exposed to potential reliability failures later. How do we know? Because we create this suite of software algorithm solution that can collect all the data from many, many different sources, from inspection tools, metrology tools, process tools, from reliability test, from failure, from electrical test and so on.
You put all together, you create a huge database, you apply machine learning methodology, and the outcome is whatever is the best solution for screening chips in the fab by diminishing, by reducing the potential reliability failures that may happen later on. That's what we have done. We introduced this IPAT. IPAT now is a standard in the Automotive Electronics Council guideline and specs for automotive semiconductor devices. In other words, it's been written in the spec, and you use our machine learning methodology, proprietary methodology, plus the KLA tech or KLA tools for inspection. That's the reason why we are part of the automotive ecosystem, and we are a big influencer in the way how we spec or the industry spec automotive semiconductor for reliability and quality control.
On top of that, of course, we build these other products after the IPAT a couple of years ago.
Although electrification in auto industry may present unit opportunity for broader semi cap like units of chips, like, logic chips going into car industry is good for everyone. When it comes to KLA, perhaps the TAM opportunities are bigger given the way you described it. It's more than just etch and deposition for traying the chips that go into an EV.
Correct.
Right.
Correct. It's many things.
Right.
Getting all these tailwinds from electrification of vehicles, silicon power and silicon carbide power in our SPTS business. That is process tools. It's like process control tools that are becoming more and more important and relevant.
Mm-hmm.
because more semiconductor content in the fab, but also more advanced technology in the fabs. We have 5 nanometer controller now in that car.
Sure.
The third thing is our ability to create a standard in terms of software, other solutions that
Right.
the semiconductor industry gotta follow in order to make sure that they comply with the specs and eventually they minimize
Sure.
the odds of reliability failures at the end of the line.
Finally, you wanna give us a preview what these incremental TAM opportunities could be like? Is it like 500,000 , more than a billion? Any way you can quantify the incremental TAM opportunities?
Well, it's definitely very, I would say, double digit. You know, we don't give like big numbers in this context and so on, of course.
Yeah.
I'll tell you one thing very, very clear. In 2019, in the investor day in September in New York City, I presented. At that time it was kind of awkward because nobody was thinking about automotive. At that time, automotive semiconductor industry was not really an important stepping stone of semiconductor. We presented a plan on all these ideas that, by the way, get this idea materialized in the last two years, three years. And then we put up a plan to say, "We will reach 4x the pretty much the revenue by 2023." Well, I gotta tell you, we are ahead of that plan.
If you look at how much revenue we had from automotive semiconductor in 2019, and you see like how much revenue we'll have in 2023, we have more than 4x increase. That is quite significant. Of course, with the inception of the Orbotech and the SPTS as well. We are talking about impressive growth. Impressive.
What was that revenue in 2019?
Don't let me say numbers publicly.
Right. Okay.
We don't publish these numbers. I don't want to.
Yes, yes.
Talk about the numbers now. The growth has been, is going to be 4x, depending what you have before the acquisition of Orbotech SPTS, after acquisition of Orbotech SPTS. Together with all these machine learning methodology and new products that we developed and released in semiconductor process control, we said we increase our revenue 4x in four years. That is amazing.
Gotcha. Okay. I just checked my inbox. There is no follow-up question, and we're reaching the time limit. I wanna thank the entire KLA team and Oreste. Safe travel, and thanks so much for providing an insight. I think these are very exciting opportunities beyond the traditional front-end process diagnostic.
You are right, Mehdi. We have a lot more to say in the industry.
Okay.
Together with automotive and packaging subs and PCB and whatever we have done in the last few years to diversify and make our pipeline of tools and the presence in the market stronger. Thank you. Thank you for having me.
Thank you, and I wish everyone a great and healthy weekend.
Thank you.