ASML Holding N.V. (ASML) Earnings Call Transcript & Summary

Yes. So let me make a couple of comments on maybe both the capacity and demand sides. First, on capacity, I think what we have stated now for a number of quarters is that we have a plan to take cycle time, manufacturing cycle time with our EUV systems down from around 30 weeks at the start of this year to around 20 weeks by the end of the year. We're making good progress to that cycle time reduction plan. And by the end of Q2, we were around 24 weeks. So I think it's fair to say we're pretty well on track. And this should set us up for -- to get the capacity to 45 to 50 systems next year. In terms of where we are with the current orders and backlog, the EUV backlog at the end of Q2 was around, I think, it was at 54 systems. And of those, 28 were going towards 2021. So we already have a pretty decent order coverage at this point in the year. And obviously, we're expecting more orders in Q3 and Q4. So in terms of the exact demand in 2021, I think it's best to say it will depend on a number of elements, which is what we commented on in July regarding some of the uncertainties out there as it pertains to next year. One, I think the most important being macro. The economic impact of COVID on the global economy and how this will trickle down to end markets, still, we have to see how this unfolds. And obviously, you have this political, call it, this potential geopolitical environment that continues to provide different components of uncertainty in there. And then if you look at the different market segments as it relates to EUV, I think Memory, as we mentioned, will start in DRAM. So that looks to be pretty stable there. On Logic, as -- clearly, the vast majority of the EUV demand is from Logic. It really will depend on the slope of the different node ramps. I think the timing of the foundry nodes are on track. That's clear. But the question is on the slope of the ramp based on how many customers they have at each of the nodes and how aggressive these ramps will be. Obviously, there's -- finally, we'll have some impact. We'll have to see what -- one of the customers concerning their 7-nanometer delay. It's still early to know at this point as we are currently in discussion with this customer. And in summary, I think the next couple of months will give our customers a bit more time to get better visibility on their ramp. We are in discussions with our customers. And obviously, these discussions will provide more clarity for our demand as it pertains to how we see EUV demand building out in the 2021-2022 time frame.

Got it. And on the famous company that has pushed out 7-nanometer, I think clearly, as I understand it, the orders they have -- you have on hand with them have a certain slot, certain delivery date in mind. If they were to say, look, sorry, guys, we need 9 months, we need 12 months more, does that incur a penalty? Or is it something for negotiation? Is it something you constantly have to deal with and maybe you can pull forward different customers? I'm just wondering, is there a cost for you guys? Or is it just something that you just have to live with?

Yes, I think -- well, right now, we're -- in the short term, again, we're -- we'll need to determine what they really plan, what do they need to make any changes in terms of the demand distribution, the tools and time. But it's too early to say at this point. We're in discussions with them as we speak, so I don't want to comment any further. We'll have to see how those discussions go before we can provide any clarity on what we need to do if we need to do anything with respect to demand.

Got it. Just moving on to High-NA. It seems an incredibly ambitious effort to get this High-NA into production, particularly given that it's the size of a double-decker bus in terms of height. How are the bottlenecks today to hit this production ramp in 2024? Would you say that the bottlenecks are more infrastructure, like pellicle or resist or something like that? Or is it more on your side? And is everything kind of on track as things stand today?

Yes. Maybe I'd say it this way. Yes, there's clearly -- I think if you look at High-NA, there are really 3 major developments as it pertains to the scanner itself, one being the optics. And we continue to make good progress on High-NA as a whole. But obviously, a major development of this technology will be in the optics. And a lot of this work is being done with Carl Zeiss. And so we had -- do a lot of expansion, a lot of building up of the -- not only the building and the metrology systems and equipment inside. Next, obviously, is our stage technology. We have to really accelerate, increase acceleration of stages. That's something that we see in our area of expertise and something we know how to do. And then the last, obviously, is a source in terms of scaling the power. So these are the 3 key components of scanner. And again, we're using the same type of LPP, laser produced plasma source that we have on our 0.33 NA systems. So that obviously means we're not starting from ground 0, but actually more of an evolutionary technology with respect to the source. On all those fronts, again, we have aligned milestones with our customers, and we continue to make good progress towards that. We plan to ship the first system starting in 2022. Obviously, the customer pull for High-NA continues. Obviously, they were -- first off, they put orders upfront. Clearly wanting these systems and the time frame targets and nodes, which are basically will enable their node strengths, but also it [ defines ] the use of EUV double patterning. I think that's the key motivation driver for our customer. In terms of the ecosystem and trying to say, is there one versus the other, I think the way to look at the ecosystem is that it's really a second generation of EUV. And just like you see with prior wavelengths, when we make the lens numerical after changes within the wavelength, those changes are, call it, much less disruptive or difficult. They're -- because you already have the wavelength established at the first generation, so future generations within that wavelength are more evolutionary. And when I talk -- when I say evolutionary, I'm talking things like resist, mask, pellicle inspection. I think if you look at what was stated back at and presented at the conference in February, SPIE, you can see there are a lot of areas where there's work ongoing. But clearly, the industry is fully engaged and driving improvements in multi-areas. So I think there's a lot of brain power, a lot of resources behind this to drive this. And we don't see any major bottlenecks today, but instead, you see this is continued improvements. Obviously, a lot of difficult and challenging engineering, but nothing that we consider a bottleneck, and we still see the overall program on track.

What do customers tell you about this whole reduced field size that comes with High-NA? Because you've gone with this half field concept for -- which allows for a faster stage. But I assume that means that the days of a huge GPU or FPGA are kind of potentially in the past. Does that mean that some customers will move more to chiplets? Do you think that will be a trend that will accelerate with High-NA?

Well, I don't know. I mean, again, it depends on the field size. And as you mentioned, it's half the field size, but obviously, the productivity is still quite high on the High-NA system. So yes, it's going to depend on the design of the chip. So if it's a larger chip, then you get into discussions around stitching. If it's within the half the field sizes, which is still a pretty large field when you're talking 26x16.5, you still have a large die to work with. So those are discussions ongoing right now with customers in terms of what actually the strategy they'll employ based on their design and their devices and do they -- is it outside of that half field or not. But that's something that, again, we continue to work with our customers.

Great. And then moving on to the HMI acquisition and your efforts in e-beam, I understand you've just launched the eScan 100 -- 1000, sorry, earlier this year. You can maybe -- for those who are not so familiar with that business, you could just maybe discuss the competitive situation in e-beam and where you are relative to peers with your multi-beam strategy in terms of really making inroads into that space.

Yes. So we continue to see strong adoption of our e-beam. They're single-beam systems today with the bulk of our revenue primarily driven by voltage contract application. And again, a bit more skewed towards memory space. We are seeing demand expanding in other application spaces as well. And I think the key competitive advantage being resolution and productivity. The demand for e-beam is really driven by the improved sensitivity or resolution that e-beam provides over optical inspection. As the industry moves down to smaller and smaller features, it's becoming too hard to actually see and determine what that defect is with using the optical capability, and therefore, you need higher resolution, which e-beam provides. The challenge, as you know, is that e-beam is around -- is really around productivity. And to try to address the productivity or try to increase the productivity of the e-beam tool, we're increasing the number of beams. So the first system, our first multi-beam system that we shipped, which is the eScan 1000, we shipped that in Q2 this year. And we plan to ship a few more of these eScan 1000s to customers this year, and they'll be primarily used for development. It's a 3x3 9-beam system, and the system will increase throughput up to 600% compared to the single-beam inspection system. Now we are stopping the 3x3 beam. We plan to further scale the number of beams and ship higher-productivity beam systems next year. And these are targeted for volume production as our customers start ramping in 2022. And we will continue to scale. We have a technology road map, and we plan to continue to scale the beams and productivity of these systems over time. As I mentioned, we're not only looking at this from a -- just a strictly a litho perspective or inspection. We actually plan to utilize this not only within lithography, but outside if we can improve process control. Because really, with our metrology system, just like our YieldStar, e-beam is another metrology system that it feeds high-quality data. And the only -- the data is only quality and/or bad data and bad results out. So we need to have high quality, high accuracy, high performance. And we need a lot of volume, lots of data, which is where the productivity comes in, to feed our computational lithography models that they can then in turn utilize this -- the models can in turn make some meaning out of this data and convert that into something that the scanner can use, which has hundreds of control knobs to adjust and optimize the process. And so I'll say our unique capability is provide value for our customers via improved pattern fidelity with this combination of metrology, computational lithography and, obviously, our advanced lithography scanner systems, and we call this holistic lithography.

Great. Moving on to the Installed Base Management/services line. Maybe just if we could just -- if you could just update us in terms of breaking that business down by its different component parts and what the gross margin is on each of those different parts? Obviously, EUV services is one particular item, I think, where there's a lot of upside currently.

Yes. So just as a reminder, our installed base, it's a combination of service, which you could think of is obviously maintaining our machines at a customer site; and upgrades, meaning enhancing the performance of the machines at customer sites, so basically maximize our customers' capital efficiency so they can use these machines in future nodes. The service revenue really grows as a function of increasing installed base on machines. So our service model basically is transitioning from a kind of a break/fix model in deep UV to more of a wafer-based, value-sharing model with EUV. The more wafers, obviously, that we run, the higher value to our customers, obviously, and we would share in that value. And so that's obviously something that's starting to kick off. And if you look at our services business, as we -- the EUV services business, as we entered this year, we are actually operating at a negative gross margin. And by the time we go through this year, we expect to be at a positive gross margin. That's really a combination of 2 things. And that profitability, we expect to continue in time. And it's really a combination of 2 things. One, on the cost side, it's about improving the efficiency at how we support these machines, which is put as man-to-machine ratio improvements, meaning as you ship more machines into customers' locations or into customers' fabs, you don't need to add the same number of people in time. We already incurred all that initial burn, which is why we are operating negative in the past. Secondly, as they run more wafers, obviously, with this new value model, we will see increasing revenue as more EUV wafers are run in time. So this will continue to drive the future profitability on the service side as it relates to EUV. Then the other side of the business is the upgrade business that I mentioned. It's a bit more lumpy because it depends on customers' ability or willingness to take machine down to do an upgrade. Our systems are designed to be upgraded in time. And this provides, obviously, a nice benefit for them, but also an opportunity to grow our business in time. And so this year, if you put those 2 together with combination of growing service and then a strong year of upgrades, we see installed base growing significantly this year. We indicated on the order of 20% over last year. And we will continue to see installed base grow in time and become a larger percentage of sales over time.

I'll just remind everyone on the webcast, if they would like to ask a question, now is a good time. Just one last one from my side as we're coming towards the end, then obviously, please do ask your questions, guys, on the line. You've obviously made a big announcement. I thought about your payment terms on EUV in the last set of results. I think in the past, you've been very generous doing pro bono services and offering -- pay us in 11-month deals. It sounds like you sort of backed down on that. I was just wondering, obviously, that's going to help your cash flow a lot next year, but what's the reaction been from customers so far to that news?

Yes. So maybe just to talk about cash flow a bit. But in general, our capital allocation policy as a whole, and because people asked also about buybacks, so our capital allocation policy really remains unchanged, meaning in the next few years, we'll generate a significant amount of free cash flow. And that cash that will not be required to support the future growth of the business will be returned to our shareholders in a combination of growing dividends and share buybacks. As you know, we're investing in the growth of the company, both CapEx and working capital. We plan to have CapEx at levels previously communicated in support of some of the future technology and required expansions that we're going through. And because the way we are growing and also in addition to some of the COVID situation, we had to take quite a bit of inventory just to be safe. So we are working, to your question on contracts, we're working our way through a transition period on our customer contracts as it relates to EUV. The older contracts being from a few years ago when EUV was a different point on the maturity curve, primarily with respect to -- if you want to talk maturity curve, we probably talk to productivity. So our contracts may also include extended payment terms because of this maturity or lack of maturity at EUV at that time. Today, we're a different point on the EUV maturity curve. Obviously, with -- our productivity is over 170 wafers an hour and moving to even higher, 15% to 20% higher with these systems next year, along with availability improvement. So we're in a different period now, in a transition period as we move to these new contracts to include down payments, and we're working on that with all of our customers. And this obviously is, again, a reflection of long lead times and early supply chain commitments that we need to make. Therefore, you should start seeing improvement in the free cash flow in the second half. And then in 2021, we should really see further improvement in free cash flow. The other consideration on the use of cash for share buybacks was obviously related to COVID and the environment, the risk and uncertainty that it created at the end of Q1. COVID is not behind us, but clearly, we're in a much better position today. And so this includes our judgment on whether we see this, we need to also come to the rescue of our supply chain. So when you take all this into account and consistent with what Roger mentioned on our July call, it is possible that we could start share buyback or resume a share buyback in Q4.

Great. Yes, well, we've run out of time. So thank you for joining, and thanks, everyone, for joining on the web. And have a great conference.

All right. Thank you. Thanks, Rob. Take care.

Thanks. Bye.