Great! Well, I'm Alex Duval. I head up the Europe, Tech, Hardware, and Semis team, based in London at Goldman. Delighted to be here, with Skip Miller, Head of Investor Relations, for ASML. Thank you so much, everyone, for joining. I'd just like to state, before we get started, that this conversation is not intended for the media and is off the record. Great. So let's perhaps start with a high-level question. It'd be great if you could just, help us with the latest demand trends you're seeing on the logic and memory side. And what are your sort of latest growth expectations in these end markets, if you think about this year, if you think about 2025, and what are the key drivers here?... Yeah. Thanks, Alex, and thanks, everyone. So yeah, maybe to start off, and I think this is consistent with what we said in July, is that the recovery so far this year and the growth this year has been primarily AI related. I think that's definitely the case, both the logic and the memory. Come back a bit on that, more of the different market segments. But in terms of the other more traditional, you know, PC, smartphone, auto, industrial, a bit slower to recover. So we're still, you know, waiting on that recovery or watching as that, those industries also those segments recover. I think there's clear, you know, indicators with inventory levels coming down, litho tool utilization is going up, to give you some confidence that you're seeing the recovery. It's just, I think, a discussion now on what's the slope. What's the slope of that recovery as we go through the second half of this year and into next? With respect to what it means for, you know, logic and memory in our space, I think, you know, first off, this year, you know, we said logic would be likely lower this year relative to 2023, as the industry digests a lot of the capacity that was put in place last year. Memory, we saw it being higher this year, and it was primarily a technology-related, we call it technology node transitions. And what we mean by that is that we saw a lot more of the capacity shifting to the higher end, the more advanced nodes, and that was, you know, in support of primarily the HBM-related demand, High B andwidth Memory, in support of some of these AI, advanced applications. You can still use that advanced technology for DDR5, that supports the other, application space as well. But I think that's where we are in terms of what we're seeing, this year in terms of logic and memory. I think if you look forward, obviously there's opportunity there, both in the logic and memory space, with the move to 2 nanometer, as you look forward in terms of more advanced technology, again, for high performance compute, but also the broader, you know, memory recovery, not only in HBM, but the broader, demand for capacity. Utilization is still trending, but we'll just have to see how the slope unfolds in the coming quarters. Super helpful. And you touched on AI there. You know, one of the investor debates we come across is clearly sustainability of demand, what the use cases will be. You know, how excited are you about AI, and what are the ways that ASML benefits? Is it more about logic? Is it more about memory? How should we be thinking about that? Yeah, so I think AI, it's not like it's a new thing that popped up. I think we've had AI in our you know different call it long-term trends and secular trends we've been talking about for a number of years now. So it's not like AI is totally new. However, I would say that we're now seeing this year really AI you know starting to take off. I think that again is exciting on multiple fronts in the sense that what the capability of AI can mean. I think we're right now just looking at AI in the form of Gen AI-type applications and what that will mean. I think, you know, maybe a number of years ago, we thought it was just this thing that happened more in the, the server space or in the cloud, but now you see it moving actually to the different end devices, and that's creating a significant opportunity. I would say also that it's, you know, just starting, and you don't really... likely, you'll see AI touching across all industries, which will be, you know, quite significant demand, and so a great opportunity, going forward. So quite excited about AI. We'll obviously talk a bit more about AI, in November during our Investor Day, but it's clearly one of the key drivers that we've identified for a number of years now. In terms of what it means, back to maybe a bit what I said in the earlier response, is that for logic, it means more advanced nodes, high-performance compute, and on memory, also more of the advanced nodes, but in the form of HBM. And yeah, don't forget all this infrastructure that's required around here. You need a lot of mainstream technology, which means, you know, call it older technology that equals a lot of DUV systems, whereas the advanced too might mean more and more EUV. So it's AI is exciting, one of the key secular trends. The question obviously we get is, do we have it properly scoped when we talked about it in 2022, about future demand, and that's something we'll update you about in November on this front. That's super helpful and one of the most frequently asked questions. Another really frequent question is about Intel, so obviously, we've seen news about adjustments in terms of how they're going to spend CapEx. It's obviously a big spender within the industry, so it'd be great to get your sort of perspective on what that means for the outlook this year, next year. How should we be thinking about that? To the extent that Intel might spend less, what could that mean? Does it matter if there are others who step into that opportunity? Yeah. So yeah, first off, obviously, we don't get into customer-specific revenue when we talk about our revenue and our plans going forward. And I think maybe a more generic comment is that a lot of times we have very routine conversations with our customers many times across the quarter in the form of technology and capacity needs and so on and so forth. So a lot of times, you know, that's obviously critical because the lead times, especially on our systems, so our EUV machines can be 12 months or more, High-NA even more than that. And so you need to have the dialogue with our customers. It's critical we get it right. It's not like there's an alternative elsewhere, where you can go get a different, you know, somewhere else to get an EUV system. So that's critical that we understand their needs and make sure we have the systems, and it's also critical that they communicate what they need so that we don't limit anyone's ability to ramp the technology. Having said all that, you know, that dynamic, obviously, I think if you look at that customer, they're clearly working on their five nodes in four years. A critical component of that is EUV. And so they need to not only demonstrate the ability to get to the five nodes in four years, but also the ability to start ramping some of that technology. I think, you know, shorter term, I think of what we said in July with respect to our guidance is still very much intact. In terms of our longer term, call it 2025 outlook, also, we said in 2025, we will give more color as it relates to where we land within the EUR 30-EUR 40 billion as we get out there in November. And that obviously would include this customer in that story as well. Very helpful perspective. Another frequent question is obviously about China. We saw last week news article about change in terms of regulations pertaining to the Netherlands. Could you please put that in context? How much of a change is this? Is this a fundamental change? Is this just a process change? What does it mean? Yeah, so the comments made last week or the changes made last week were well, there was no fundamental change in the restrictions, meaning that we can still ship 1980s to our Chinese customers. It does require a license. The difference being a technical process change, in that the approval comes from the Dutch government. We welcome that. That's where in terms of the approval process coming from the Dutch government. And in terms of expectations, again, going forward, we expect still shipping 1980s. We won't ship EUVs. We won't be shipping NXT:2000s and higher to China, but we will ship 1980s to our Chinese customers, with the exception of a handful of restricted fabs, which we communicated at the end of last year, the beginning of this year. So that's the summary. No major change in scope. There was a technical process change on the approval and who does the approval, but that's a welcome change from our perspective. Very helpful. And if we stick on China, one other area we had questions was about services. Just wondered if you can put into context how big your exposure would be to sort of China domestic services, in the group context, and any other color would be very helpful. Yeah, I think. So the question is what's the service revenue to our customers in China? I think if you look at domestic China, it's around 2% of our revenue last year. And, you know, that's for all of China. And I think any discussions we've had, which have been, we talked about even earlier in the year, was about, you know, how to service the restricted fabs. And the discussion is more on the sense that we view it strategic to be in those fabs, because if we're not in the fabs, then that increases some of the competitive risk. And so that's the dynamic in the discussion there. I think it maybe got expanded into the all China discussion, but that's the service status. Got it. That's very helpful. And I guess just to round out the discussion, I mean, some people looked at your China revenues in Q2. I think it was around 49% of group, but my understanding is that's sort of fairly unusual, and that will normalize. Could you help us sort of understand how we should sort of think about a normalized China- Yeah exposure, perhaps with reference to your backlog? Yeah. So if you at start of this year, there was a question, okay, with some of the changes that went in place last year, what impact could that have on the China revenue this year? We said that if you look at the 2023 revenue, there could be a 10%-15% impact in terms of the China revenue with respect to, you know, the changes that went into effect in January. However, we also said that demand continues to be quite strong. And so the ultimate outcome for this year was gonna depend a lot, not so much just on the China demand, because we see that continuing to be strong, but more importantly, how will the non-China demand recover? And so if you look at what happened in the first two quarters, we see that the revenue to China was, as Alex mentioned, around 49%, which was a combination of the two things mentioned. One is the fact that we had this strong demand over the past years that we're unable to deliver, so we're able to deliver more of that in the first half of this year. And the second part is that the non-China recovery, as I mentioned, was a bit slower to recover, and so therefore, that inflated the percentage that went to China. On a go-forward basis, if you look at, you know, we don't guide by region, but if you, there was a question of what does that, could that be? I think, you know, the expectation is that as the non-China business recovers, that will more evenly distribute the revenue across the regions. And therefore, that percentage would likely come down to more something aligned with what we have on our backlog, which today is a bit over 20% of our backlog. Super helpful. I guess, you know, we touched on services, but if we look broader than China, so across the group, I think there was commentary for around flattish kind of growth. But obviously, there is an expectation in the second half that there will be a pickup in terms of- Mm-hmm ... market demand, units, and so on. So just curious to what degree that's conservative to talk about flattish services? Yep. Yeah, so maybe first, when we talk services within ASML, we're talking installed base. So installed base for us means a combination of service revenue, you know, actually servicing the machine, parts and repair. And then there's an upgrade component, upgrading the installed base to extend the capability of the machines in the field. And so that combination is what we call installed base. And this year, we said we see the installed base business, you know, flattish relative to last year. The service part continues to grow, but as you install more and more machines, that generates more service revenue. But on the upgrade business, that's a bit more lumpy, and that's you know more dependent on when customers are ready to do the upgrades. And in the past, we've said, "Hey, if you look at the utilization levels back when we were really high, I think it was 2022, 2023 timeframe, we weren't able to do upgrades because the fact that the fab managers did not want to take the machines down for fear of losing output. We were able to do some software upgrades because they were very limited in the scope in terms of machine downtime to do so. And so when utilization levels come down, you think that'd be a great opportunity to do some upgrades, but now the CFO may say, "I'm not quite ready to do the upgrades until I get more confidence in business recovery." And so that's that balance. And so as you're kind of in that transition from lower utilization to higher utilization, that obviously would be the ideal time. But that's obviously something that we have to align with our customers, and so we can get some of that in a much shorter lead time. But software, obviously, is something you know can fill that demand in a much shorter lead time than, say, some of these major hardware upgrades, so I think, you know, upgrades still, we see as an opportunity to go forward. As the industry recovers, that will be a part of the growing business that we see going into next year. That's great. And I guess before the last quarter, the question was about, you know, further orders coming through, from, you know, one big customer, on the EUV side, on Low-NA. And specifically, there was discussion about, you know, framework agreement that would allow that to happen. Then we had the quarter, then we saw, obviously, a good number of orders did actually come through. So one question I had is just into the back half of the year, should we expect sort of further order flow, now that that's the gate has kind of been unlocked? One would hope, it would appear. How should we be thinking about that? Yeah, so orders for us, you know, we've said we don't guide orders, obviously, because why? Because they come in a bit lumpy, so we build off a forecast. We have that aligned with our, you know, customers. That determines what we need to do on a go-forward basis based on what they're planning on their technology, but also on their capacity ramps, and then that's what we work on, and then we finalize that, and then as we get closer to the timing, which we need the orders, then they issue orders for, you know, in this case, EUV systems, which can be twelve to eighteen months out. If you look at last quarter, yeah, there was a higher percentage, almost 3/4 of our backlog, or sorry, bookings in Q2, you know, July quarter, came from Logic, and in there, it's right to assume that could some of that be two? Yes, that's probably a reasonable assumption. I think if you look at what our customers are saying with respect to the demand for these more advanced nodes, in this case, you know, 2 nanometer node, I think it appears to be a very strong node. There's comments that was even, you know, greater than three, which was greater than five. So that clearly gives you, you know, a view that the opportunity for two can be quite strong. And again, I think it's driven back to some of these discussions we've had earlier around high-performance compute as it relates to AI, but also smartphone and PC space. In terms of the ramp of those wafers, obviously, that's something that our customers would have to answer. How do they see the, you know, the wafer ramp? And then you have to back off and say, "Okay, then that means we need to have systems there a quarter or two before those wafers need to be there." So clearly, on a go-forward basis, you know, they will be adding more systems for this, you know, ramp of this new technology, and we'll see orders come along with that. But it will be probably lumpy, like you've seen in the past. There's no reason to see that won't continue to be lumpy quarters on a go-forward basis. Totally appreciate, based on what you've said, that, you know, 2 nanometers should be strong. Clearly, the shape of the ramp, we'll have to see. But one other factor that had been identified in prior quarters was clearly this idea of pre-building. Mm-hmm. So curious to what extent that could also be a factor in play here. Just could that impact customer behavior? Yeah. So we, I, just maybe as a reminder what we said. So we planned coming into this year, we said it's gonna be 2024 is a transition year. We're looking at 2024, 2025 as a combined year and looking, we don't know exactly when the recovery is gonna take off, so we're putting these two years together and saying we'll do some pre-building. And so that's the situation that we're in today. That will allow us, if the demand pulls more aggressive into the second half of this year, or 2025 is stronger than what we can deliver in one year if we just wait to build all machines in that year. So that's the plan with pre-build, and that's what we've, you know, have done to date. In terms of, you know, future pre-builds or how things will go, that will obviously depend on how we see 2025 and beyond shaping up. ... Great, and we've obviously talked a fair amount about Low-NA EUV. I'd like to also talk about High-NA. Sure. So if we look at the press and the investor discussion in the last sort of six months, I guess there have been a couple of things that I'd love to touch on. First is pricing. So talk of some customers saying, "Oh, this is an expensive tool." I'd love to get your perspective on what that means. Does that change the story at all on High-NA? And then secondly, the other question we get is around this Intel cut. So if they're cutting CapEx, does that impact their proclivity to buy these high-end tools, or is that actually unchanged, given their need to win share? Yeah. So first off, a High-NA, just for those that aren't familiar with all the terminology, High-NA stands for High Numerical Aperture, and it's basically scaling the lens. So the lens gets bigger, it goes from a 0.33 NA to a 0.55 NA. What that does, it allows for smaller and smaller feature sizes, so you end up printing smaller feature sizes, and allows the continuation of shrink in time. It also in this case, reduces the process complexity. And so the value that High-NA provides is that instead of having to go to a multi-pattern or double patterning technology, where it takes two exposures to complete one layer, you can do it all in one. And that reduction in process complexity will improve things like yield, better chip, better die, a number of good chips on a wafer, will help with cycle time, you don't have to do as many process steps, and also cost. And that value proposition is what drives our customers to pay for the High-NA systems. Yeah, they're not cheap, but they bring a lot of value. And I think that same story is out there with Low-NA EUV, is that the EUV systems went above 100 million EUR, and so they said, "Oh, well, that'll never be affordable." But in the context of if you can replace not only two litho steps, two immersion steps in this case, but also the non-litho process costs, that allows the cost advantage to go into this EUV Low-NA solution as we go forward. The same is true on High-NA. So I think High-NA is exactly what played out when we went from DUV to EUV. Low-NA is that it's about process simplification and improvements in yield and a reduction in cost. But I would actually add a bit more in that the High-NA today, that we'll be shipping... The ones we're shipping today, but also the ones we're planning to ship in next year, which will be our next generation 5200. The productivity is much higher and much closer to the reference process, which in this case is EUV Low-NA, meaning it's going to be closer to 200 wafers an hour, which will allow that, the cost analysis and that benefit and that value to be even more clear. And the second part is that the ecosystem that's in place within EUV, think of photoresist, think of masks, so on and so forth, these things are already in place, and so it's a more evolutionary change when you make this numerical aperture EUV Low-NA to High-NA change. So for that reason, I think there's a lot of value that we are confident will continue to demonstrate itself as High-NA is rolled out. It's still in the early stages, so we're, you know, our first shipment we started at the end of last year, and those systems will end up at, you know, customer sites, and they will be qualified, and that will allow customers to start learning, but in the meantime, we have a machine in Veldhoven that is a imec- ASML joint lab system that will allow customers to go and run wafers, their own designs, on the machine in Veldhoven, so they can do some early learning as they wait to bring their systems into their fab, and then once they do that, they'll go through development and eventually then move to high volume manufacturing ramp. So I think we're in the early innings of looking at the data off the High-NA machine in Veldhoven, but the data is quite encouraging. Customer feedback is quite positive, so that's all a good sign for the early stages. I think the timeline that we've been talking about for High-NA is still very much intact. Customer plans are there to adopt across not only logic, but also memory, which is also a bit different than we had back in the EUV Low-NA era, where memory was later to come to the game because of productivity. So there's a lot of reasons why we're confident in how EUV will be rolled out, EUV High-NA will be rolled out, but it's still in the early innings where customers have to gather that data. So it's going to take some time, but I think it provides a great opportunity, and again, another value that will allow us for the lowest cost-per-transistor, over time. And yeah, one of these customers in there that you mentioned are part of that story. They'll be building, planning to ramp, High-NA as well as part of that, but there's other customers as well that, are planning to ramp, that technology at time. Brilliant. And perhaps just to round out the High-NA discussion, obviously appreciate that in the long run, it's very important that High-NA rolls out. It's important for cost for customers. That's good for the whole semis industry. Everyone can grow, they can buy more from the semi cap industry. But in the event that there were some delay in the midterm in terms of the ramp of adoption, economically, what does that, what would that mean for ASML? I totally appreciate your saying that the near-term signals are quite good, people are trying this, the roadmaps are there, no bottlenecks. But if there were to be some kind of delay or something like that, what would it mean economically for ASML? Yeah, so I think the timing of High-NA, first off, I'll say what we've said in the past number of not only quarters, but years. I think we're still executing quite well on that program. But your point being is that if you don't have High-NA, then what happens? And I think the point being is they then use two Low-NA systems. So from a financial perspective, it's not a bad thing from our perspective, but we're always looking at what's the most cost-effective and what's the best way to drive, you know, cost-per-transistor down. And we think in the end, you'll find that the solution that offers the least complex process, and the highest yield possible and the lowest cycle time possible, therefore the lowest cost, will be the winner in the end. So even though there's a strategy, they can incorporate double patterning with Low-NA. In the end, we think High-NA will still be the solution that customers go with. Super clear. We've talked about High-NA, we've talked about Low-NA EUV. I'd like to touch a little bit on the trailing edge, and perhaps we can divide that between China and non-China. So if we think about the demand outside China, how are you looking at that? Obviously, a lot of that's driven by some of these areas like, perhaps PC, smartphone, and then, of course, some of the industrial auto demand, where the visibility is a bit lower. So how are you thinking about the outlook there? And then if we think about China, I think one of the other semi cap players was talking about the fact that they believe going forward, there could be some digestion of tools related to sort of China demand. Obviously, different to lithography, but curious how you'd think about that. I know you have the data point you put out there, that in the last few years, China only got half of what it wanted, so just curious to what extent, if we put aside all the discussion about regulations, there can still be demand coming from China. Yeah. So, yeah, to your first question on what is it, in terms of how do we see the demand on, what do you call it, mainstream or mature technology, but 28 nanometer and above type of technology. In our Investor Day 2022, we, you know, took the end markets, and then we converted that into wafer demand, and then we broke that wafer demand out between advanced logic, mainstream logic, and then, you know, memory, DRAM and NAND. And what we found is that if you total all that up, that you basically need to add 780,000 wafer starts a month, every year, of new wafer capacity. Within that, though, almost 380,000 of that is mainstream technology. A big chunk of that is the mature technology. Why is that the case? Because it goes into everything. So whether you talk very advanced applications or more trailing, in this case, you still need mainstream technology across the board. So where will that be built? You know, it's going to be built, obviously, in many locations across the globe, including China. China's a big part of that. They're building a lot of new fabs in China that supports that with their end goal, obviously, to be more self-sufficient as they build the technology, which is the bulk of what's being built in China today, which is the mainstream technology. In terms of the outlook going forward in terms of the utilization levels within these different customers within China, there's a broad base of customers in China. There are, you could say, a handful or so that are, you know, a more mature customer, and then you have a lot that are in the process of not only, you know, establishing themselves as a customer, but establishing a technology and learning how to ramp these fabs. Under those scenarios, those type of customers, they are still working through the qualification of the technology and therefore, the ability to ramp, and therefore, the utilization levels are lower. They'll have to come up, in time to align more with the more mature, you know, five to 10 customers that are already been in China and running for some time. We see it as a case where you need these wafers, a significant demand there. It has to be built somewhere across the globe, if not in China, elsewhere, but someone needs to continue to build these fabs, and today, obviously, China's been doing a lot of that, but, you know, other regions around the globe may also choose to do so, because it's going to be a continued demand because of the broadness of the infrastructure of that technology. Great. So, I think my last question before we've got a few minutes for questions from the floor. You know, we've talked about High-NA. We've talked about the strong demand from advanced logic customers. We talked about China. You know, we've touched on briefly, I guess, the cyclical side of things. You know, we're sitting here in September, so obviously, people are thinking about 2025. So if you just sort of pull together the different dynamics you see, you know, how should we be thinking about things relative to that sort of EUR 30-EUR 40 billion guide, which you obviously have issued in the past for 2025? Yeah. So, our 2025 scenarios that Alex is referring to, we communicated in 2022, between EUR 30 billion and EUR 40 billion. Obviously, the end market being the big driver between how will the markets be in 2025, whether you go the high or the low end. And what we've said is that we still see that the range, but what are the key reasons why we feel that type of range is justified? And we said, well, first off, these secular growth drivers that we've been talking about, AI, for example, we've been talking about for a number of years. Those are still very much intact. Secondly is the, you know, where are you in the cycle? And I think, you know, our view coming into this year is that we would be clearly in the up, going to the upward part of the cycle. But some question on what's the slope. So what's going to be the slope of this recovery? As I mentioned earlier, we're seeing utilization levels and inventories moving in the right direction. So the recoveries, there's clear signs of recovery, but there's a lot of debate on what's the slope from our customers in terms of how these other more traditional end markets, like PC, smartphones, auto industrial, how all these things will recover in the coming quarters. And then lastly, we said the fabs. Will there be fabs available to take our systems? And I think that's very much on track, and still fabs are being built both with the, you know, globally, with the support of government, but also without support and across the globe in many different locations. And there are many pedestals that will be available to take our machines. And so it's, I think, the one secular intact, the fabs will be ready in the middle part, which is the cycle. Where will we be in the recovery of the cycle? I think that's the question mark that we still have to continue to monitor and work with our customers to determine where we actually land between that EUR 30 billion and EUR 40 billion in 2025. We've said we don't see how to get to the low end, but the question is, beyond that, where we land. Looking forward to November. Maybe just one quick follow-up. You know, obviously, one debate in the market is, okay, 2025 should be a rebound off a lower year. If we think about those kind of drivers you've talked to, the cyclical, the secular, the fab buildouts, you know, to what degree do they pertain in 2026? Can there be sustainability of growth qualitatively if we think forward two years? Yeah. We haven't even got to 2025. We're already talking 2026, so maybe it's a bit premature there, Alex. But I think, you know, if you look at what I just said, the three things which are the secular drivers being very much intact, there's no reason to believe that would be a one-year event. I think the cycle, if you look at our history, having a one-year or less cycle would be very rare, meaning it's a multi-cycle event. So, and then lastly, I think the fabs that are being built not only can support what's going on in 2025, but beyond in 2026. So for those reasons, it's hard to see a, you know, scenario without a continuation. But, the exact numbers, obviously, we'll have to wait and talk to more as we get closer to 2026. Fantastic. I think we have time for one question. If somebody could put their hand up if they have a question. Yes, gentleman at the front. I don't have a microphone, so- I can hear you. Just High-NA, back to High-NA. TSMC, I think, are saying that one user with A16, in general, they can get ordered in five years. So who will actually get time? I think the A16, the comment that was made by TSMC, was that they would ramp wafers in 2026. And if you back that up and say, "Okay, we're going to start shipping our EXE:5200 in the 2025 timeframe," that would be a pretty aggressive and risky immediate adoption from first shipment of the model, and they're going to immediately put wafers on it in 2026. I don't think there was ever an intent there to have High-NA in there. I think what we've said all along is that the initial adoption of High-NA in the 2025-2026 timeframe has to do with the fact that that's when the EXE:5200 is available. And the plans from both our customers in the logic space, but also in the memory space, is such that they are planning to bring High-NA in that timeframe. But the actual wafer volumes obviously come later in time. We've said we expect revenue on five High-NA systems in 2025, and we're planning for a capacity of 20 High-NA systems in the 2027, 2028 timeframe. So that's still what we're working towards, but obviously, as customers get more exposure to High-NA and really see the value there, we'll have to continue to assess, is that, you know, what will that exact profile look like beyond 2025 as both logic and memory customers adopt? Very brief follow-up. Very brief follow-up. Just logic and density in WLP- Yeah ... going forward. Do you think the next five years will affect the last five years or just packaging technology? So logic, I think our general view, that was logic specific or just general? Logic. Oh, yeah. So logic, I think if you look at the upcoming Gate-All-Around, that is very good for not only the industry, but good for lithography in the sense that it enables future shrink. So once you complete this transition with Gate-All-Around and prove that it works, that there are no issues there, then you can resume a more aggressive shrink, and that will enable additional EUV layers on a go-forward basis. Both a combination of Low-NA layers, but also High-NA to avoid the double patterning. So there's, you know, again, nice growth going forward. In terms of the relativity, prior five, future five, obviously, that's something we can talk more about in November. Great. Well, on that note, Skip, thank you so much. The very illuminating discussion. Thank you all for joining, and enjoy the rest of the conference. Thank you all. Thanks, Alex.