Welcome to day two of Citi Global TMT Conference. My name is Elizabeth Sun. I am here in Citi's research covering in the Semi and Semicaps team. It is my pleasure to welcome the Onto team here. We have the CEO, Mike Plisinski, and then we have the CFO, Brian Roberts. Welcome both.
Thank you.
Let us start with market overview. Mike, when we were here last year this time, the WFE outlook was more, say, uncertain, maybe a slight growth, and there were some, or a lot investor concerns about some share dynamics at that time. Today, the WFE cycle looks considerably stronger, and Onto, I think, is growing faster than the market. If we start at a high level, what has changed over the past year, both in the breadth and durability of the WFE cycle and also some of the share dynamics for Onto?
I think a few things. First, the markets for sure have been strong. Customers are confident. They are spending money. They are giving us better visibility. They are sharing multi-year expansion plans with us. That is good. Then behind that, we had several new products. The Dragonfly G5 came out and really did an outstanding job in the beginning of the year, gave customers confidence in the ongoing technology portfolio that we have for inspection, which helped them drive or basically give us more share. In addition, we had the Atlas G6 come out also around the same time, and the adoption for new Atlas products typically is relatively long, as customers take a lot of time to make sure that the matching is there between the Atlas V and the process control is in line.
The adoption was very fast with the Atlas G6, and in some cases, we saw customers just go straight to purchase orders and ramp right on the Atlas G6. That tells me that there was some compelling value from the Atlas G6 that our competitors couldn't deliver, and I think that also helped.
Yep. Just to double-click on that piece, I hate to ask this, but I'm curious, have you gained all of the share back?
What share? I don't know what you're talking about.
There's more to go on the share gains part.
It's hard to say, because the particular customer you might be talking about, which I don't recall, didn't spend a ton. They've been factory constrained. They're trying to leverage s to sort of free up some capacity, and we got huge OSAT orders, so that should tell you one thing. But back in, what was it, March or whatever it was, we did say that we were qualified for all. So we'll see where the actual allocations end up when they start spending more aggressively in 2027. But the initial feedback has been super positive. Already the competitor is reacting and trying to fight back. We already have our solution. So I think it's the normal dynamics. We're back to the normal games we play.
That's good.
But I would say as far as did we gain it all back, for sure. We're capable for everything that happened. How much the customer wants to allocate back and forth, we'll see. Generally, that customer cares more about total spend with a particular supplier versus second source on a per-layer basis. So we're optimistic that we should see essentially things coming back.
Sounds good. You talk about backlog more than $1.1 billion, and 60%-70% is scheduled for 2026. Your Q3 guide implies maintaining sequential growth, and Q4, I think, assumes kind of mid-single-digit growth. So does this strong backlog create any upside to either Q3 or Q4? Or is it conversion pace constraint more like supply facility readiness or installation timing?
Yeah. I think our focus is more on driving the additional backlog for 2027. I think the growth we've implied is around 40%, significantly above our peer, nearly double our peers, frankly. And of course, that kind of growth can stress supply chains, so we're more focused on, all right, let's maintain that growth. Let's make sure that our supply chains ramp effectively, quality is maintained or improved, et cetera, and that we're preparing ourselves for an even better 2027. Our focus is a little bit less on trying to drive a lot more upside there versus manage through and prepare for even better 2027.
And what kind of growth are you preparing for 2027? Are you expecting a kind of acceleration or, if we compare to the market, growth better than the continue this kind of outperformance?
Though we're not correlated to WFE. We think the drivers we have in place, for instance, the Dragonfly G5 opening up potentially front-end markets, the Atlas G6, which is in the early stages of adoption and bigger expansions expected next year in logic. The silicon photonics opportunity, which if you add all those up, we would expect we would be able to outperform the market, but again, we're not correlated, and it depends on where the market is.
No, that makes sense. And for that backlog-
Maybe share more at the Analyst Day.
Hoping you can share a little bit more today. For that backlog, how concentrated is that? Is that by customer or by products, or it is very broad-based?
It is fairly broad-based by end markets, but the bulk of it is tied to our core products, the OCD metrology, the inspection. Of course, we have some litho in there, and we have some of the other products all throughout, but the core is there. It is fairly broad-based across the segments we serve.
If you look at growth by segment, what are the biggest growth area you are expecting going into next year?
This year the biggest growth was in packaging, 80% growth. The advanced nodes was about 35% growth. We would expect, not those numbers, but similar dynamics where advanced packaging will still probably grow stronger than advanced nodes. We expect both to grow fairly well.
Understood. In advanced packaging, you are growing 80% this year. Is it more in the HBMs memory side or more on the 2.5D logic side?
It's fairly well split. Every quarter is a little different, but we're generally around this 50/50 mark, 60/40, depending on the quarter, memory or logic being the leader there. I would say, and that makes sense because they're kind of tied. The logic requires the memory around it. As that grows, the whole packaging requirements grow across both. I'd say that's the dynamic we're seeing. Of course, logic is getting more competitive, and there's some constraints in the logic packaging space. We're seeing end customers, the hyper chip designers, the AI device designers, looking at alternatives and expanding the panel-level packaging and leveraging other technologies to try and drive performance, and that's creating new opportunities for us, too. Yeah.
I'll touch base on panel-level packaging a bit later. Just between memory and logic, is there any one area over the other you are seeing more share gain opportunities, or is it also very balanced?
For memory, we've always had very high share, and that we maintained. Share gain is really tied to 3D metrology, so the bump metrology. There, if you hear more about smaller, denser bumps, that creates more opportunities for our 3DI technology. In the logic packaging, there we have some opportunities to gain share because of the very high resolution that the Dragonfly G5 provides.
Understood. You said you received more than $200 million of Dragonfly orders from just one single OSAT, mostly for delivery for next year. Does that signal a broader structural increase in process control intensity in OSATs, or is it specific to the customer capacity expansion specifically?
That's a good question. As the industry is constrained on the heterogeneous integration, there's two things happening. One, end customers are looking for alternative technologies for packaging. EMIB is one everyone talks about, but there's others. The other dynamic is, in this case, TSMC is well-publicized. TSMC is looking at how to release more capacity, the fab constraints. What do you do? You take your lower-end process steps, you outsource those to the OSAT, so you can free up your precious fab capacity for your higher end. That dynamic we saw in play. I'd say the OSATs are, from a capital intensity perspective, it's really volume driving these huge orders, and it's volume, I expect, going to be from both the outsourcing of capacity as well as the capacity that they're going to be selling from their own technology.
Got it. What is Onto's share position on the OSATs versus the foundry players?
Generally, on the higher-end applications, not super cheap stuff like, I don't know, Analog Devices or something like this. Then we have very high share on the 2D side. We've gained some really nice traction on 3D, at least a couple of the OSATs, on 3D bump metrology. So pretty high share on the high end. If it's low end or if it's China, of course, then we have less.
Got it. I want to ask on Dragonfly G5, it was just launched, I think, this year, and you talk about qualified at leading both memory and logic customers, I think. You have said it will be a relatively small portion for 2026 revenue, but how quickly do you expect customers to transit from maybe G3 to Gen 5 in next year?
The key there is that I don't expect so much of a transition. Maybe originally we thought maybe that could happen. But what we're seeing is customers still appreciate the G3 for its lower price point. But if they have applications that don't require the high-end 2D, and it's mainly some of our ClearFind capability or subsurface inspection or something like this, then that platform, the G5 platform, gives them no added benefit, so they can stay on the G3 platform. Or 3DI can also perform the same on both platforms. So if those are the primary applications, G3 continues. If it's really tied to 2D and those applications are extra, then the Dragonfly G5 will be the choice. The real focus for G5, though, was expanding into new markets.
It really wasn't about, all right, G3 and a transition. Everyone's going to migrate to this higher margin tool or whatever. It's really about growing into new markets, because the G3 has been growing phenomenally throughout 2026, so that demand is still high. Our focus with the G5 is high resolution, new applications in packaging, which we've already demonstrated and proven, but now the front end, and that's the focus for 2027.
Maybe-
2027, sorry, because I-
No.
did a long way of not answering your question. I would expect much more growth in 2027 for the G5, but tied to new applications versus a transition from G3 to G5.
Got it. Maybe it's just me not super familiar with your tools. What kind of new applications does G5 open up for you?
High resolution. Okay, so 100 nm type applications inspection in the front end. That is one. But we mentioned there are several others also in specialty devices, again, front end, but different applications than, let us say, advanced nodes. Again, high resolution. We did not really give a lot of detail, but what we said was we had about 10 customers, I think a quarter ago, 10 customers across 15 different applications. We would expect, based on the feedback and how the valuations are going, that the majority of those will convert to orders throughout 2027.
Understood. Maybe a quick one for Brian. G5 is margin accretive?
Yeah, absolutely. It has got a better ASP and a better margin profile for us.
Got it. I do want to ask about silicon photonics. Is that more than $500 million silicon photonics SAM you see by 2030? How should we think about the opportunity split between inspection and metrology?
I think inspection is going to be the largest opportunity. By nature, it's almost always-
Yeah
going to be the largest opportunity. But we are seeing metrology penetration. We are selling metrology tools. We also think there's going to be opportunities for some of our other tools in the space. So yeah, when we come up with that SAM, inspection will be for sure the highest opportunity, and that's reflected in the backlog we have now. The bulk is inspection, but also multiple customers adopting a couple different metrology tools.
Got it. There is a wide range of expectations on silicon photonics or CPO. I am just curious to your $500 million number, what drives the upside or downside to that number?
I think it is how quickly the technology is adopted at the server and then at the rack level, between those two. We are playing in both, so I do not even think it is a question about if it happens. From what I have heard and what I have listened to in other executive forums, from the NVIDIAs, Broadcoms, AMDs, et cetera, they are committed to driving this forward. I think the hyperscalers are committed to it from a data center perspective. It is just a matter of how quickly it gets adopted and ramped. The economics and the performance benefits are, seem everyone is convinced on that.
Yeah.
That would be the upside downside or timing.
Understood. If we stay in this specialty devices, I think silicon photonics is in part of that. But outside that, it has improved throughout the year. You are talking about photonics and also power and silicon carbide demand beginning to show some signs of recovery. How big is data center power for Onto right now relatively to the traditionally, I am assuming it is more about auto and industrial power?
Yeah, I think data center power is relatively small. From the customers, you remember we had a good run with the power semi. Even as EVs were slowing down, we were still growing quite nicely, and that was tied to driving yield improvements for those customers. Then the expansions kind of slowed, and their investment in the yield improvement. But now we're hearing from them that things are picking up and we need to prepare for this because of the data center power. We're not seeing the actual orders. I don't have a good clarity on timing, but if they're already talking to us, I would say 2027 into 2028 would be what they're thinking about.
So it could be kind of another leg of growth driver going into the next couple of years.
Yep. I don't know how big the magnitude of it yet, but it should definitely be. It would be captured under our specialty devices. But it'll be, yeah. We'll see. I mean, silicon power was on a tear for a while there. If the data centers and how big these data centers are that are being announced could take quite a bit of chips.
Yeah. Understood. On the panel-level packaging we're getting here, so it does receive a lot of industry attention, but adoption still appears to be relatively early. Where do you think the industry is right now for panel-level packaging?
I think it's an interesting market. For enterprise servers, the adoption for advanced IC substrates was already very clear, and the two big guys, Intel and AMD, during the COVID explosions, they talked about being constrained by advanced IC substrates. That's an area that the JetStep X500 and our Firefly play in. But then there was overcapacity, everything fell out, you know the whole story there. That has slowly come back and worked itself through. But now we're seeing the constraints in AI packaging or heterogeneous integration drive customers to look at panel packaging as an alternative. We also see as the adoption of heterogeneous integration is kind of clear now. That's happening.
Die are going to get bigger, mean packages are going to get bigger. The economies of scale at the wafer level just aren't going to work. At the 310 by 310, they get better, but compared to 600 by 600 or 510 by 515, still much smaller. So these customers are looking at, all right, where can we drive some of this capacity to these other companies that have a mature process or let's say a proven process, and can ramp right now? That's what we're seeing in our business when we talked about it doubling and then growing again into 2027. That's now the accelerant we see. That part of the adoption, you're right, is early, and I think that's going to continue over the next several years.
Understood. You talk about you are addressing this panel-level packaging market in three ways, like or three lines of products. Could be a little bit early, but, how should we think about the size of the opportunity in the next few years?
I forget what we've said publicly. I think we said it can be around a $1 billion SAM, and our competitors have said even more. For us, that's going to include process control, basically inspection as well as the lithography. We're not going to get 100% of that SAM. There are competitors. Could it be $200 million, $300 million? That would be a reasonable-
Yeah
expectation over the next few years.
Okay. Understood. I am going to take a pause here to see if any questions in the audience.
Hi, thank you for taking my question. Do you see any headwinds at all to process control intensity going forward, or is it pretty much up into the right from here?
Up into the right forever. In the near future, I do not see any headwinds. Demands are high. The industry is just driven by innovation, which drives complexity, which needs more process control. I do not see any headwind other than when do things slow down. That is just a general industry comment. As far as specific to process control, generally, it increases with the increase in complexity, and I do not think anyone believes that the semiconductor industry is going to get simpler in the future. It just always gets more complex. Process control intensity should always continue to increase. Yep. Oh, actually, you are running the show, sorry.
Go ahead.
Say more about-
Sorry, there was one on the mic there. I'm sorry. I think that's you.
Sorry. If you could just elaborate a little bit, you drew some parallels between the data center cycle that we're running right now and the EV cycle as it relates to your business, and maybe other services too. Can you elaborate on that a little bit and kind of where we are in those separate cycles and how they're somewhat different and what the read-through is to your business would be helpful.
Over the last several years, we saw really significant growth in the adoption of our process control, primarily inspection, but also metrology for supporting power semiconductors. That wasn't driven because the market was going down. Capacity overall was coming down. It wasn't driven by expansions from those customers. It was driven by the need to improve yield. Yields were in the 60%-70% range, not in the 90s like you'd typically expect. They really used our tools to drive that yield improvement with the expectation they can drive higher output, right, with higher capacity. That recovery for the EVs hasn't really happened yet. What customers are telling us now, those same customers, is they're seeing demand from the data centers that's going to drive adoption of these power semiconductors. Hopefully that means another growth driver for them.
As they expand, they've already inserted our process control in the process, so we should benefit from that. Yeah.
Thanks for the time. Can you just say more about your partnership with Rigaku? I know you guys initiated a 27% stake. I think transaction's going to close second half of this year. Maybe just talk about that partnership and also the appetite to buy the entire company and your appetite in general.
First it's closed, so that's done deal. We won't talk about appetite for what the future steps could be. Obviously, our history is not to make partial investments in companies, but we're not going to talk about what the next steps would be. That said, the partnership was strategic in that for us looking forward, and actually this is a misunderstanding. Some investors thought this was because we saw OCD eventually running out of steam and X-ray becoming, let's say, the next step or the next thing. That's not what we see. What we're seeing is OCD, at least for us, we can continue to push OCD and serve the market for the foreseeable future.
As the world goes to more 3D transistors or 3D stacking of die or wafers, there are elements of the process between these layers that customers are interested in that optical can't solve. X-ray is a potential solution, a potential way to see through these layers and provide information to the customers. Also, our acoustics, our acoustic metrology, sound wave metrology is a very unique capability for that in the metal space. It has limitations. The partnership with Rigaku, we've started with two different customers. We went up against the leading player in X-ray, and we beat them at both customers. The combination of our AI Diffract, our strong modeling engine, which is used in optical also, and the very strong sensor capability or X-ray capability from Rigaku makes a real compelling value proposition. That's one.
The next is kind of this idea of a hybrid metrology. The idea that the speed of optical and the precision from X-ray, but it's much slower, can combine to provide a more compelling value proposition for our customers. The combination is made much simpler because the AI Diffract, our software, is aggregating all these data streams that we're already modeling. Now we can model the aggregate instead of just one or the other. You see what I'm saying? We get more information in the models. That's another advantage to this partnership.
Mentioned the features of X-ray. I guess to some extent that's doing that partnership. I saw that although I didn't catch that it is conflict sensitivity, I would guess that's the X-ray would co- oexis .
Coexist, right. So far we can push optical metrology to the extent that the roadmaps are that we see with our customers. We're working with them on nodes that you're not aware of. So we're pushing the technology. As long as optical can make the measurement, it's the fastest way and the cheapest way to make the measurement. Customers will always choose optical. That's one. But the limitation of optical is it can't see through thick layers of metal, materials, et cetera. There's gaps as the world goes more 3D. Now, when it was a planar world, didn't matter. Things got smaller and thinner, and optical could see through. As things get stacked and they get thicker, then we need a new technology. We have acoustic metrology for part, but it's not a full solution. The X-ray can be a complementary solution in a world of 3D.
We have a few minutes left. I do want to ask about gross margin. Q2, I think you did 57. You talk about incremental 50 basis points in the next couple of quarters. I guess, going forward, how much further can gross margin go, and is pricing a consideration in that?
Yeah. 2026 has really benefited from the move to the extended factory. All of the margin really that we've driven to date, as well as what we expect to drive here in the second half of 2026, is driven by operational productivity and efficiency. That is the move of manufacturing from the U.S. over to these international factories over the course of the year, and we talked earlier that a year ago this time, we were shipping our first tools from our international factories, and in Q2, we shipped about 70% of our tools. Just a meteoric rise up the curve there.
As we move into 2027, certainly the benefits of the Dragonfly G5, having that highly accretive margin profile, some of these other products Mike mentioned, the Atlas G6 is another, will give us some ASP lift, which will help us to continue alongside productivity drive towards a margin that starts with a six. That's certainly still what we believe is that we can have a gross margin profile that'll start with a six. That then is dropping through at a higher rate to our operating margin. To date, we've driven 250 basis points of gross margin. We've doubled that at the operating margin line. With the growth in the business, it's allowing us to continue to invest in R&D and become more operationally efficient, but also then continue to drive more to the bottom line.
Hopefully we'll continue to see that operating margin expand at a faster rate than what our gross margin's been driving towards.
Is there any possibilities to reprice some of the existing products, except from the new products are getting higher pricing at that point, but for existing product portfolios?
Yeah, certainly, it's not lost on anybody that the two major headwinds in 2026 have really been focused around memory cost and probably freight, fuel shortage. So we've tried to help offset some of the fuel shortage by, again, being in region. That's been a helper for us. But as we go into the 2027 pricing season, if you will, budgeting season, certainly some of those conversations are ongoing with customers around the existing product portfolio too.
Got it. To that point, in this market, people ask about capacities, both on tools perspective and also the supply chain perspective. First of all, are you supply constrained? Second, do you see any supply bottlenecks?
I'd say two parts. One is, again, the move to the extended factories is really fortuitous timing because it allowed us to be able to have partners that can expand alongside of us and give us excess capacity in these new clean rooms without us having to actually put out a lot of CapEx dollars. So that's been a big benefit for us. On the supply chain side, anybody that says it's perfect is kidding you. But the reality is our operations team's done a really good job of diving our way through the supply chain using the incremental visibility that we see from customers and being able to, I think, get ahead of where we see potential bottlenecks. Certainly, we're using our balance sheet a little bit. Our inventory balance has grown. But I think we're doing a good job of mitigating that risk.
Never perfect, but overall, we're not missing customer dates, and our lead times have really not grown too much.
Got it. All right. That's good to know. We are out of time right now. So thanks, Mike and Brian, for coming to our conference.
Thank you very much. Thanks for having us.
Thank you.