All right. Let's get started. Good afternoon, everyone. Welcome to the 26th annual Needham Growth Conference, 2nd day in person. Glad to see so many people here in the audience. My name is Charles Shi. I'm a semi cap analyst here at Needham. Joining me today, right now, is Onto Innovation, and thrilled to have here on stage, Mike Plisinski.
Very good.
as the CEO. We also have Mike Sheaffer, Head of IR. He's in the audience. So Mike, really wanna thank you for joining me today. I know you've been coming here for many years, but it's probably my first time hosting you as a covering analyst of Onto. I'm looking forward to more of more like this, hopefully-
More interactions.
Yep. All right. So let me start with some basic questions about Onto. Onto was created after a merger of Rudolph Technologies, which you were the CEO of Rudolph, right? And Nanometrics in 2019. I think back then, the prevailing view about this merger seemed to be that there were simply not many revenue synergies between the two businesses, and now we know that that was probably an inaccurate view, I mean, to say the least, right? And can you kind of tell us, over the past four years, how has the merger transformed the company, and what kind of revenue synergies have materialized-
Mm.
as a result of that merger?
Sure. So that may have been the prevailing view, but of course, that wasn't our view. When we brought the companies together, the cost synergies were sort of the obvious play, but that's not really enough of a reason, strategic reason. So we always saw the synergies. I think what we said publicly is it's hard to quantify, so we didn't really put a number out for it. And then, as we got through the cost synergies, which we achieved, a couple of years ahead of schedule, we started to outline some of the revenue synergies. So those came in, in several forms.
The first was just when we merged, and we started meeting with customers, this was right before COVID, customers would talk to us, the large IDMs would talk to us about how strategic all of a sudden this combination was to them.
Rudolph Technologies was driving several of their initiatives in the packaging area, which you can start seeing now is driving a lot of growth. Then also, Nanometrics was critical to some of the roadmaps driving the advanced nodes in the OCD. The combination made us much more important, and you can see that reflected in the VPAs. We've talked about it over the last several years. The first set of VPAs we got were roughly 20%-30% larger than we would've normally had as standalone units. There was a-
Just wanna clarify, VPA means Volume Purchase Agreement.
Thank you for that. Yep, and so we just saw an increase there 'cause now we had a broader portfolio of products we could horse trade with. In addition, the next year, we announced doubling of that, so really significant growth. That helped propel us to number 3 in process control overall by 2022. Yeah, 2022, we were number 3 behind KLA and Applied Materials. The next set of synergies that we were working on and talked about were through the channels to market. So Nanometrics was very focused on the advanced nodes, the leading edge, and we saw opportunities to bring their technologies into power semi, specialty devices, our filters, things like this. So that was another area of revenue synergy.
Especially the last few years, that part of the market has been growing quite meaningfully.
Yep, exactly. I think, you know, after year one of that, we talked about $40 million of revenue synergies achieved through the channel market opportunity. The other area was through technology, so combining the technologies from Rudolph and Nanometrics to go after the planar films market, a market that's been dominated by KLA-Tencor for decades. By dominate, I mean 98%+ share. Customers basically said to me, "Look, we understand the Rudolph technology. We understand your focused beam lithography. We understand your OCD. Bring those together, and you can drive a real meaningful opportunity for yourselves in films." Films is roughly half of a $1.2 billion optical metrology market, so it's quite a good opportunity for us. We released that tool.
We've talked about, over the last couple of years, penetration in several of the top five semiconductor manufacturers. We've also gotten customers in the specialty device area as well, power semiconductor in particular. So the progress for our films business is still early, but it's definitely progressing well. And there's more to come.
Great, so I heard, commercial synergies on the commercial, right, the VPA? The synergies on, on the channel, the go-to-market, the synergy on technology product R&D.
Yep. So one of-
... talk about your metrology business. I believe it is largely the, I mean, the root of this metrology business is probably largely the Nanometrics from the Nanometrics. Correct me if I'm wrong, though. I mean, I used to follow Nanometrics as a research associate-
Mm-hmm.
I mean, prior to the merger, and remembered that the Nanometrics business was heavily driven by memory, and specifically NAND. So when investors really think about the 2024 WFE, the puts and takes between different buckets, which one's gonna grow, which one's gonna have a little more challenge, should people still think your metrology business a memory play?
No. I think even before the merger, the Nanometrics team, the former team, did a really nice job of diversifying their position in advanced nodes. So, during the merger or after the merger, we've talked about roughly a third, a third, a third. So they had roughly a third of the revenue at any given quarter was through logic, advanced logic, DRAM, and NAND. And we can talk more about that. I mean, NAND, there's more customers, less capital intensity, but more customers, so that's how you get to a third. DRAM, the capital intensity is higher, but of course, you have three customers. And then in logic, advanced logic, again, you have a small number of customers, but the capital intensity is even higher. So roughly a third, a third, a third.
So if you add that up, two-thirds is memory, so yes, we're still heavily tied to memory. But still, when we talk and when we look to the future, you hear us talking more about Gate-All-Around and the node transitions to the more complicated 3D structures that the OCD, our OCD, has some unique capabilities to be able to measure. And again, you may remember, before we entered the market and started pushing the OCD to these levels, our competitors were talking about leveraging X-ray, and you needed to shift to CD-SAXS or to CD X-ray solutions for this kind of technology, much slower technology. We were able to show that our OCD can be used beyond 2 nanometer, and we're actually doing applications and proving that.
So we've made that clear to the market, and we've gained, you know, significant share and opportunity there.
So, I mean, you kind of touched upon that Gate-All-Around, right? So I think let me dig into that. I know this is something that investors care. What exactly is your play in Gate-All-Around? Can you kind of give us a rundown? What are the opportunities there? I mean, especially probably on the metrology side, what are the specific opportunities you have?
Yeah. So specific to Gate-All-Around, there's the OCD component. That's obviously the biggest component. In that area, as the structures become more complex, you need more, more metrology, ideally OCD, and so far, we've proven we could. So one of the... There was 3 customers were qualified, and all 3 that are driving the Gate-All-Around node. To give you an idea of what it means for us, one of those customers, it's essentially double the process steps that an OCD is being used in. So from a FinFET to Gate-All-Around, they're doing twice the number of metrology measurements or process steps. The others are more around 20%-25%, but still an increase.
In addition, we just talked about our better sensitivity to the structures and the better modeling capabilities we have, allows us to pull out the data, pull out the measurements, and provide the information they need. So we're gaining share because they're-- we're-- of that capability. So more steps plus share gains. And then the other area where OCD is important for Gate-All-Around or it's important for us, is the number of APC layers increases. So APC means, it's essentially 100% way from metrology because they're gonna use that data to feed it into a process equipment. So to adjust the process based on our data. In FinFET, that was roughly two steps where there was 100% in because of the APC, automatic process control. In the Gate-All-Around, we're seeing between four and six steps.
So that's a significant increase in utilization for the OCD as well. In total, you'd start adding that up and you say, "Whoa, that's, you know, double, triple your opportunity." The problem is, the customers aren't so dumb, and they push us very hard for throughput, so we've also doubled the throughput of the tool. So in essence, we say that Gate-All-Around for OCD is roughly a 30% increased opportunity for us than FinFET, than the most advanced FinFET node. But in addition, we talked about films. In the area of films, we've gotten qualified for the Planar Films in two of the three Gate-All-Around customers. So that's another upside where we didn't have that in the FinFET. So as you start to see that grow, we should see more OCD sales and additional film sales, Iris Planar Films.
Maybe just a quick follow-up, right? I think you talk about APC. It does sound like there's a software component to that. Is there any contribution to your software revenue, I mean, for all the things you talk about in Gate-All-Around? Or maybe this is more of a part of the hardware sale; we won't really see that from your software revenue line?
It's not so much. So the bulk of our software revenue is gonna come from fab-wide sales and-
Yeah
... other sales that'll drive a larger portion of it. A lot of these customers have their own APC systems in place, so we're feeding them.
Feeding.
So that's not gonna really help us. Where we do see a lot of software adoption is in the AI- Diffract. The software that we came out with after the merger to drive much more advanced modeling. There's several new algorithms that we developed as part of the AI- Diffract engine, but that's required to run the tool, so we don't really. We sell portions of that separately, but it's really how we win share for the overall OCD entitlement or.
Got it
... opportunity.
Got it. Got it. Okay, let's stay on the hardware business for now. Sorry, I digress. Moving on to the inspection, and maybe let's throw litho, the discussion on litho-
Mm
... in there as well. I believe this was largely the former Rudolph business, the business you managed for many, many years, right? And so there has been a lot of good news, right, over the last few quarters on the strength of this part of Onto's business. I believe since August last year, right, you announced roughly $230 million of orders that you can say directly tied to generative AI chip demand.
Mm.
One of the questions I received was about the sustainability of this strength beyond first half 2024.
Mm.
Because it does sound like both you and your peers, you, you guys both talk about you get orders, a lot of them were shipped by the end of first half 2024.
Yeah.
So what's beyond that, right? And, are you positive about the strength going to continue, yes or no, and why?
So it is semi. So if I said I'm positive on anything, immediately no one listens. If you asked me in November, I would've said: Look, they're ramping so aggressively, I expect a digestion period. You know, I mean, we went from 0 to double the Dragonfly output in just four or five months. That aggressive ramp generally means there's gonna be a digestion period. However, I visited several of the customers in December, both on the HBM side and on the logic side, and the indications are pretty strong that they're gonna continue to ramp at this level and that they're actually not yet meeting demand. So a lot of questions I get in the meetings are about, "Well, you think it's overbuying right now, and it's just a bubble, and eventually..." No, we're not seeing indications of that.
Which, again, back in November, I might have said, "Yeah, that's a little bit my expectation." So I think there's a lot of positive momentum. I believe TSMC reaffirmed that this morning, announcing that they were gonna double the CoWoS capacity. That's super-duper for us. So. And with that, HBM is sitting around that, so that's also good, so.
Got it. So, I mean, I look at... Because you talk about TSMC, right? I wanna ask you about TSMC.
Look at, I look at your historical-
Did I mention them by name?
Okay. I look at some of your historical filings, right? I believe around 2016, you had, I mean, some really strong orders. I mean the Rudolph filings. I don't mean Onto Innovation.
Mm-hmm.
Around 2016, you had very strong orders coming in from TSMC. Yeah, and now you're getting another wave, right, of orders from them. Of course, that's, we're talking about CoWoS or heterogeneous integration. And what's the difference between that wave of orders in 2016 and this time? Why is this wave of orders seemingly so much stronger than 2016? I mean, maybe let me clarify a little bit, right? The reason why I wanna ask this is that, supposedly, my understanding, the 2016 order strength coming from TSMC was supporting the ramp of Integrated Fan-Out, which is geared towards the smartphones.
Mm-hmm.
Supposedly, that's high volume, right? But now it's kind of CoWoS is, yeah, it's high value, but the volume, yeah, they're doubling it, but from a very low base. Why you are seeing such strong orders from-
Mm
... from CoWoS, given the volume difference between these two packaging technologies?
Yeah. So the process control requirements are significantly different. The InFO started out, I'm going by memory now, but roughly 8 microns lines and spaces, and then they pretty aggressively moved it down to 2 over the last eight years or so. Where CoWoS was starting out more in that 3, 2, 1 kind of range. In addition, because the value of the die is so high, there's a lot more tension placed on chips and cracks, particles on the edge, particles on the backside. So our Dragonfly isn't just doing RDL inspection, particle inspection, we're looking at edge normal, we're looking at backside, we're looking at the thickness and the height of the RDL, the redistribution lines between interconnects, as well as the CD and uniformity of those lines.
In addition, which wasn't available at in during InFO, but-
... some yield issues required us to develop and deliver something we call ClearFind. The yield issues had to do with resist residue on interconnects, so when you're doing 2.5D or even 3D packaging, any kind of resist residue on those interconnects means the worst possible thing. One, you'll bond and electrically test, but when that goes into a smartphone or a car or a server, where you have a lot of heat changing, that bond breaks. Now you've just lost the whole board or the whole device. So ClearFind was the only way for customers to be able to, well, in this case, you know, you already named the customer, for them to be able to detect this source of reliability failure. And I remember, quick, funny story, I think we have time.
We were showing them the ClearFind, and they said, " we don't believe you. There's no way we have all that resist residue." And they kept cleaning the... We were looking at the bottom of a via, so that was one of the sources of issues, the bottom of vias, which is very hard to see. And we kept saying, "No, you have it," and they cleaned it three times, then they were getting mad, and then they escalated to me. I had to go and meet with them. And it turns out, while I was flying there, they did a SEM. So they did a destructive SEM, and they measured and actually showed, "Yeah, you're right.
We did have all that resist residue." And so when I showed up, instead of getting a beating, they said, "Oh, we're going to mandate 100% ClearFind inspection." So it was outstanding. And then that's, that's helped us, so that's also been carried over into the CoWoS world.
Got it. Got it. So yeah, it definitely feels like it was a very light volume production of one kind of advanced packaging, right? By your customer now is actually getting into real volume.
Mm.
Um, and, uh-
True
... that has been very good for Onto, I believe. So we talk a lot about AI, I mean, the revenues you can kind of directly tie to the generative AI chip demand, but how do you feel about your inspection business that's not related to AI?
Hmm.
Yeah, and assuming the demand is kind of at some kind of a bottom level, not right now, but when do you think things will start to recover? When and, how strong do you think?
So, yeah, I'll kind of quantify it or try and answer your question this way.
Yeah.
Onto's strength in process control and packaging is going to be on the advanced side, for sure.
Mm-hmm.
So, the IDMs, actually, with the start of InFO, when TSMC took away the business, the Apple business from Samsung because they could combine the front end and the back end, that woke up a lot of IDMs about the importance of packaging. And InFO was $1 billion, that initial ramp. OSATs were spending $200 million. I mean, that was unheard of at the time. So the amount of money then you saw Intel and TSMC, obviously, Hynix and Samsung pouring into their packaging roadmaps went up dramatically, and that's where we positioned ourselves primarily. From the OSAT perspective, we're still there. They serve a lot of the mobile market.
So if you start seeing IoT devices or mobile devices, watches, things like that, kicking up, we should see some advantage, some opportunities there. But right now, you know, I'd say the strongest position is in IDMs. The other area where there's packaging and people ask questions is around China. China's going to be a lower-end packaging opportunity. There's a huge, I was going to say, about 160 companies are on the entity list. There's actually 160, and I think it's around 100 that are semi-related. I'm asking for that to be clarified. But that's a huge number of companies, mostly under the Huawei umbrella, that we don't have access to.
So for us, China, where there's some packaging going on, it's more low end, and it's not a big focus for us because of the restrictions and the, well, the sentiments that are between the U.S. and Chinese governments.
Got it. Got it. So maybe let me ask you-
Let me clarify-
Yes, please.
Because now everyone's going to say, "Holy crap, they don't have any China business." That's actually not the message I'm trying to deliver. I was trying to answer packaging. But for us, China is really now about specialty. So where we can sell into where our metrology and inspection capabilities have unique value that can't be found outside from other countries, that's where we've seen the most traction. So power semi, and China is actually a great source for power semiconductors with all their EV success and the distribution of the power stations and the grid that they've been working on and developing. So we see a lot of opportunity there. And also, of course, we had the big front-end business that was YMTC, which was put on the entity list.
That, we're also recovering from and looking at some other opportunities there.
Got it.
Wanted to clarify that.
Yeah, I like it. Let me ask you about your lithography business, right? There's a lot of focus, I mean, investor attention on this part of the Onto portfolio, but I know you guys have been saying, well, one year of litho sales is probably on par with Q1 of the sales of Dragonfly.
Yeah.
But still, this represents some very interesting and unique opportunities for Onto, I believe. Last year, however, you seem to have had some supply problems. Right? And that, maybe were holding you back. What exactly were those problems?
... There were many, so we won't go into exactly the problems, but I'll kind of just bundle them into two buckets. One is, we built a—customers asked us to build a tool and bring it, a new tool, we didn't have this tool and capability before, into a brand-new market for us. We had a new tool, new market. Very different market than what we were used to. And in addition, we had to ramp from 1 tool, which we shipped at the, I think, end of 2021, to then, 10 tools or more in 2022, the following year, '22, and then, of course, 80 in the $80 million, so about 20 tools, in 2023.
And that kind of aggressive ramp, while we were making changes to the tool because of what we were learning, while the customer, our customers were qualifying their process to their customers, AMD's, the Samsungs, et cetera. That created an environment where it's very difficult to lock down the manufacturing process and drive such an aggressive ramp. Now, you say: "But jeez, that's only a handful of tools. You ship way more than that in the other businesses," and that's true. So we had, internally, two things. One, this team was used to building highly customized, highly advanced lithography systems once or twice, not volume manufacturing. That's a cultural change we had to address and continue to make. What was the other one, though? Two things. Yeah, forget the other one.
Maybe we can come back.
Yeah.
A related question, right? Applied Materials, they launched a panel litho tool around IEDM last year in December, and how should investors really think about their product versus your product? I hope you're still listening to what I'm asking.
But I remembered my other point, yeah.
Let me finish my question.
Yep, yep, go ahead.
Are you directly competing with that product, or do you kind of serve different segments of the market, so everybody's happy?
No, no, no, everyone won't be happy. No. So I'd say they're trying to enter the market, so directly competing, we haven't seen this product anywhere yet. Actually, that's not true. We've seen it. It's an old product. It was around, I think they bought a company in Taiwan maybe 15 years ago. So we've seen it when we competed against it in the display market, and we won that business when we were going after Gen 6 display business in China before the Trump trade issues. So it's not new to us, and we can talk a lot about the differences, but for us, what it does is it highlights the importance of panel moving forward. It highlights that our main competitor, which was Ushio, the customers were right when they pulled us in.
They said that the roadmaps and the technology platform can't continue to meet their needs, which the JetStep X500 can. So we're already, you know, a couple generations ahead, so we still have that. But the next step for us will be to take our display lens, the optics there, put it on the X500 platform, and be able to deliver sub 2-micron capability, which is what they were talking about in this tool. So that's, for us, actually, that's starting to be built up now in our lab. So that's not any kind of science project. That's just R&D. That's just D. So that's actually positive. Now, question is, okay, but you've had all these problems with manufacturing, and they're gonna just trounce you. So let me give you a kind of a comparison of the tools.
This is what the customers have told us since that press release went out, and it's consistent with our experience competing with it. That tool is around $10-$12 million, so let's say for a sub 2-micron capability, and what we're told is the throughput will be around 20 panels per hour. This is what the customers say, so maybe, you know, there's something a little different, but that's what we've been told. Our equivalent tool, which is already proven, so this isn't demonstrated, ours is proven in production. Our tool would be around—Well, now it'll be about that price, and it would be around 80 panels per hour.
So a significant productivity performance, significant margin accretion, because we were gonna have a good margin tool at much less price, but if they're gonna come out with that, then we have a good opportunity here. So we'll see where things end up. It's one thing to make a press release; it's another thing to negotiate, and AMAT didn't go into this to lose, so it'll be very interesting. But we do like the setup here. For us, it's actually quite positive. So there's another maybe thing I'll highlight about the competitiveness of the two tools. We're a stepper.
Steppers are kind of well understood, and I'll explain the advantage there in a second. That's a laser direct write, so it's a laser and mirrors creating all the patterning. So one of the investors yesterday described it really well, like an Etch A Sketch.
You know, it can only go, you know, X and Y. So diagonal features are very hard for it to print. Round features like vias and holes, which is part of every interconnect, that's gonna be nearly impossible. They have to etch. It's very hard for them, and they have to go back and forth. If you have a deep via, you need to keep imaging down low to get it deep, where imaging with a high depth of focus, you can do it all in one shot. The other thing about the LDI is there's no mask. Now, that's a great advantage in R&D. In R&D, you can just take a design and give it to this tool, and it'll print it out. But in production, in high volume, that's not a good thing. Why?
Because every single panel comes out a little differently. So you're trying to optimize for device performance, and you're using these panels to connect high bandwidth memory with a GPU, let's say, and the interconnects and the thickness, everything has to be precise to get the yield and the performance you want. With a stepper and a mask, once you tune that mask, once you debug the mask, every shot's the same. With this, it won't be. You know, there's a lot of... There's a reason LDI hasn't taken off. It's been around for 20, 30 years, but it hasn't taken off except in PCB. So a lot of people are putting money in it. Obviously, AMAT has put a lot of time in it for the last 15 years. Now they're doing a partnership.
We'll see what that ends up. But, our takeaway is we've got a really nice position with the technology. The market, they sized it $560 million. We had it a little bit smaller, but I'll take their number. And, yeah. So anyway, that's the-- that's our take on that.
Thanks. Thanks.
You want me to go back to what the other piece-
Yeah!
Yeah, if you still remember.
The other complexity is the complexity of the system itself. So our next most complex system would be the Atlas, the OCD, that really advanced system there. That has about 600 custom-designed components that go into making that tool. The JetStep has 3,500, a little more, depending on the configuration. It's almost 10 times the complexity, so that also created some challenges for the manufacturing.
Sorry, wanted to-
Good. Thanks to-
Not look like I have dementia.
Good to clear that up. My last question. Okay, always want to ask you about the long term, how do you feel about business? I know you're going to say everything's good, but let me ask a different way.
That's not usually what comes out of my mouth.
I know, we can basically think about your business as two major piece, right? Inspection versus metrology. We talked about metrology at the beginning, then we move on to the inspection. Between these two, if you have to pick one, which segment do you think will likely outperform the other through the next semiconductor cycle, and why?
So I'm going to answer your question in a different way. So we don't look at it as inspection and metrology. So we would look at the end markets, where are the end markets going to grow? And our job is to try and drive more of our systems on those different waves of growth. So, for instance, we talked about revenue synergies at the start of this, bringing the metrology into the power semi market, both films, films and the OCD. That's a growth opportunity, but of serving an end market. Same with packaging. In the last earnings call, I mentioned we started to see opportunities and a sale of some of the advanced metrology from the front end into some of these emerging packaging applications.
So for us, it's really not about inspection metrology. It's about what's that wave of growth? Right now, AI is a big one, packaging is a big one. And how do we drive a more comprehensive process control solution? Not just inspection, but the metrology that they need as well, stitched together with our software. This is where the software has more growth opportunities, because the fab-wide software. Because one of the unique value propositions we provide, especially for some of these less sophisticated customers, is they've got all these yield challenges. They've got all these data streams, and they don't have the sophisticated IT systems to make sense of all the data streams and drive quick decisions. We can do that with our software.
So when we go into, like, the power semi market, we come in with a portfolio of acoustic metrology, planar films, inspection, and we can stitch that all together in a way that the customers can very quickly ramp, very quickly see which part of the process is out of spec and varying, and drive the adjustments to make that stable. That's something that not a lot of people, maybe nobody, can offer. So the more we can do that, that creates more opportunities for us, versus just a metrology or an inspection play. We look at the end markets and try and get as many surfers on the wave as we can.
Got it. All right. We got 3 minutes, and maybe we can take 1 question from the audience.
I did so well answering all the questions. Hard to believe.
So, let's repeat the question. Can you, I mean, elaborate a little bit more your advanced packaging opportunities?
Yeah, so that's a good point. The advanced packaging is changing very, very quickly, much faster than ever before. And obviously, there's what you see today, the AI and the growth in HBM and the position we have there, also in the 2.5D, the panel we talked about. But what we're seeing is they're ramping, they're driving a denser, smaller interconnects. There's new challenges that they don't have a solution for. One is voids. Where are they gonna find voids? And especially when they're doing hybrid bonding, and is it a good bond? Is there anything in the smaller, denser TSVs? So right now, they have to sink that into a bath, use some acoustic. It's about a panel or two an hour. It's very, very slow. It's unacceptable for them.
So leveraging our MPG, our acoustic, our Echo metrology, we think there's opportunities to solve this in a very big way. And I think, you know, those of you who don't know, we're the only ones with that kind of technology, nondestructive, sound acoustic technology. So that's an example of a new opportunity in packaging as we look forward. And then just the, you know, not so much OCD, but the films. With all this stacking, there's more films being placed, and the uniformity of those films is becoming critical. There's opportunities in 3D height metrology.
The coplanarity is critical, and, you know, that's an area which generally, as they get smaller and denser, the precision requirements become higher, and that's an area we tend to excel at. So I think there's an opportunity to expand our position there as well.
The other challenge in packaging, as we look forward, is the number of interconnects. If you're doing 3D or chiplet packaging, you've got all of these IOs that now have to line up. So the data processing of 200, 300, 500 million bumps is causing a huge slowdown. With our fab-wide software systems that's connected to every tool in the fab, every sensor, collecting data at 1 sec... We know how to deal with high volumes of data and process it very quickly. So we see opportunities there as well to add some additional differentiation. So there's a lot of good stuff happening in the packaging space.
All right. Right on time. That's a wrap.
Good.
And, thanks, everyone. And, Mike, thank you so much.
Thank you!