ASML Holding N.V. (ASML) Q1 2013 Earnings Report, Transcript and Summary

Ladies and gentlemen, thank you for standing by. Welcome to the ASML 2013 First Quarter Conference Call on April 17, 2013. [Operator Instructions] I would now like to turn the conference over to Mr. Craig DeYoung. Please go ahead, sir.

Thank you, operator, and good afternoon and good morning, ladies and gentlemen. This is Craig DeYoung, Vice President of Investor Relations here at ASML. Joining me today from our ASML headquarters in Veldhoven, the Netherlands is Mr. Eric Meurice, ASML's CEO; and Peter Wennink, ASML's CFO. The subject of today's call is ASML's first quarter 2013 results, and we'll talk about a couple other subjects as well. This call is being broadcast live over the Internet at www.asml.com, and a replay of the call will be available on our website for approximately 90 days. Now before we begin, I'd like to caution listeners that comments made by management during this conference call will include forward-looking statements within the meanings of the federal security laws. These forward-looking statements involve material risks and uncertainties. For a discussion of risk factors, I encourage you to review the Safe Harbor statement contained in today's press release and presentation, both found on our website at www.asml.com and in ASML's annual report on Form 20-F and other documents as filed with the Securities and Exchange Commission. The length of the call will be 60 minutes, and now I'd like to turn the call over to Eric Meurice for a brief introduction. Eric?

Thank you, Craig. As usual, Peter and I would like to provide an overview and some commentary on the first quarter results and our view going forward. Peter will start with a review of Q1 financial performance and read the comments on the near-term outlook. I will complete the introduction with a brief status update of our product lineup and the Cymer acquisition status. So Peter, please.

Thank you, Eric, and welcome to everyone. Our first quarter results are very much in line with the expectations we set at the end of the fourth quarter, with sales slightly above expectations. These systems sales, again, remained largely skewed towards the foundry IDM sectors, which was 70% of total, including noncritical KrF systems supporting capacity addition. Combined memory represented the balance of 30%. As we did last quarter, we are now reporting memory sales, bookings and backlog as one total value due to lack of transparency around system allocations between NAND and DRAM. The average selling price of all systems recognized in the first quarter was EUR 23.3 million, a bit higher but basically the same as in the previous quarter. Service and field option sales were EUR 250 million. First quarter net bookings came in at EUR 715 million or 25 systems, excluding EUV, with booked ASPs at EUR 28.6 million versus EUR 20.9 million in the fourth quarter. The average selling price of the first quarter bookings were impacted by a higher percentage of immersion systems. Next quarter bookings will reflect the secular ASP trend as well as we expect increased demand in the second half of the year for the company's leading-edge systems for logic applications. Our systems order backlog at the end of the first quarter was EUR 1.27 billion, which totaled 42 systems, excluding EUV. The backlog profile at the quarter's end remained very similar to that at the end of the previous quarter. With respect to our cash generation profile, I would like to mention that we expect, over the coming 12 to 18 months, our working capital requirements in support of our EUV 450 millimeter programs to be substantial. Longer factory cycle times for these new products and investments in facilities and…

Thank you, Peter. I will focus my remarks first on the status of our technology development and start by our EUV program. Regarding the scanners themselves, EUV scanners themselves, we have now demonstrated production with the 10-nanometer node compatible imaging and overlay performance. Stability and reproducibility will certainly have to be improved as we prepare for production in 2014, but the specification target has certainly been reached now on the NXE:3300 platform. Regarding the source, significant progress has been made. The EUV light sources are now running at 55 watts with production-worthy dose control. This is equivalent to approximately 43 wafers per hour with a standard resist sensitivity. To put this performance into perspective, one should remember that, of course, we lost for the year in 2012, when the original source concept that we used to call NOMO was converted to a more robust concept called MOPA Pre. Since this concept changed, has been debugged, and it took a bit of time to do so, we have executed cycles of learnings as a standard in this industry, leading to continuously improved performance through optimization of the system software and hardware architecture. Our targeted production throughput for 2014 and commitment to customers since, in fact, last year, since 2012, remains at 69 wafers per hour, upgradable within 2 years to 125 wafers per hour. The current progress and achievement of 43 wafers per hour now, puts us well up to target to achieve our commitment. Indeed, the machines will require another year of continuous improvement to achieve the specification targets and, as already mentioned, production-worthy reproducibility and stability. However, the systems performance data is now confirmed, and the nodes for which EUV can make a significant contribution, namely a 19-nanometer DRAM on about 2 layers and the so-called -- this industry's…

Thank you, Peter and Eric. [Operator Instructions] Now operator, if we could have your instructions and the first caller, please.

[Operator Instructions] Our first question comes from Nick Gaudois.

Nick Gaudois from UBS. First question is on your comment on inquiries from memory customers effectively coming back in addition to shrinks requirements, if I understood you well. Could you give us a bit more color in where do you see that in the timeline? And I'm assuming this is more NAND flash-related, and in that case, whether the feasibility of 3D for [indiscernible] nanometer is actually playing a role in the discussions, effectively moving to a firm commitment and planning for [indiscernible]? And I've got a follow-up.

Nick, the issue is, like we said, we have received the first inquiries, and your question is very specific. We don't have the specifics as you require. In the context of what we're seeing in the industry is that [indiscernible] quite a significant increase in the DRAM price. We have seen a small increase in the NAND prices, so it is not a surprise that those inquiries are yet coming. If we would have to make a guess, it would be more DRAM-related than it is NAND-related. But like I said, it's pretty recent, and you're asking very specific questions on the architectural choices on the NAND side, which, actually, we cannot comment on at this moment in time because we simply don't know. But the general direction is more DRAM-related than it is NAND.

Well, I guess, as a follow-up to that, from a DRAM standpoint, your own model probably doesn't indicate wafer capacity requirements. So assuming if you see anything going through that has to do with technology conversion more than capacity, would that be correct?

DRAM is -- as we all know, PC business is not doing well. So the driver is not PC. The driver is mobile in general. And DRAM mobile has been a fairly good business in 2012, but you could not see it because it had been compensated by the very bad PC business. But now you can start seeing that there is a genuine growth in the DRAM arena due to portable -- to mobile. And therefore, the question that we are raising about scanner availability for the second half is probably for the possibility of moving higher numbers of wafers in DRAM due to mobile.

Okay, that's helpful. And a quick follow-up, you guys talked about YieldStar actually starting to be material for revenues as of this year. How should we think about the ASP addition this will make at all over the next couple of years?

Well, that's what we -- I kind of hinted to you this time a bit more specific than usual. We are going to add about EUR 350 million or significantly more than EUR 350 million this year of this additional non-scanner products, which is not part of the ASP.

Not only the YieldStar.

And it's not only the YieldStar. And we plan to have this number off because I am a departing CEO, and I put that one target to my successor. I would say there is -- we see potential, but I won't tell you when. We see potential towards EUR 800 million to more, but please don't put this on Peter for next year.

Thank you very much, Eric.

There is, indeed, potential above EUR 800 million for these products.

But generally speaking, when you look at the requirements, the production requirements that our customers have on advanced logic, it is -- we think it is very necessary that the base scanner is equipped with those system enhancements, which includes, amongst others, also the YieldStar solution. And that means that when we look at the list price, and I'm not going to talk about net price because that will include discounts related to the volume that customers are buying, but list price, it could be between EUR 10 million up from the base price. So it's quite a significant increase of the average selling price if they would choose to take the full suite of the system enhancement products. And like I said, on the 20 nanometer multipass requirements, we think it is -- we at least advise our customers to include those in their product choices.

Right. That's very useful. And we try to maybe forget that it is part of our targets, but not for Peter.

I would appreciate it. Thank you, Nick.

Our next question comes from Simon Schafer.

Actually, I had a question on memory. It sounds like maybe expressing a little bit more confidence there. I'm just wondering, your magic model on the memory segment, what sort of wafer start additions do you guys assume for NAND in the second half? I think Lam said on their call that it's probably around 100,000 or even north of that wafer start additions in the second half. I'm just wondering whether you guys had any thoughts on that.

Yes, I think it's not an unrealistic number. It could be around 100,000, yes.

Got it, okay. And I think at the beginning of the year, you guys talked to us about the potential for your memory revenue run rate to be down 30% to 40% this year. That seems to be quite a lot better now, and that's offsetting or mitigating the downsides in your EUV revenue. Is that the way to look at it?

Absolutely.

Yes.

At this moment, what we have put into this guidance, where we still say about the same level as 2012, is a bit less EUV compared to our last guidance, compensated by a bit more, in fact, memory business. But in addition to that, there is still upside to the memory business.

Understood. That's very clear. My follow-up question is just on litho intensity. You guys have been very clear about the type of layer intensity increase that you see on the 2x node on 20. I just wondered whether you can outline a similar sort of maps as foundry migrates to 16-nanometer FinFET. What sort of increase does that entail? I'd just be interested on your thoughts on how that looks then.

For lithography at this moment, we see no difference on litho intensity between what is called in the industry 14 nanometers, 16 or 22, FinFET or not. On paper, it's just same amount of litho. So the issue will probably be if there is a yield issue. Therefore, the machines will not be as usable. The utilization may become effective, but we don't have any effects which would tell us at this moment that there is a difference.

Our next question comes from Satya Kumar. Satya Kumar - Crédit Suisse AG, Research Division: I was wondering if you could comment on what type of order mix you'd start to see over the next 6 months? If I look at your second half guidance, it looks like you will need to see orders increase from about EUR 715 million in Q1 to maybe EUR 900 million to EUR 950 million, excluding EUV and service. And foundry orders are fairly large in Q1. So I was wondering if all that increase is going to come from memory? Are you going to start to also see increases in IDM as you move into the second quarter?

With respect to the trends, as we said last quarter, this year will be largely driven by logic with a, let's say, much lower share of our memory customers, which is still the case. So in your assumption that we should see an order increase for the second quarter, we would agree. Although the exact number, we will discuss with you within 3 months. But indeed, the order mix will be as we planned it, largely skewed towards the logic customers, the foundry customers with what we did earlier, with some upside on the memory customers. But the year will be a logic-driven year, and we will see those orders that are needed to support the sales in the second half. They will come in. Just to remind you, the ordering process with some of those customers where we have very strong relationships and we know that we are the technology of choice that we've said in earlier quarters, that the ordering process has become more, I would say, administrative in that sense. And it is more important for us that we have an agreement with the customers on the shipment dates that they need our tools to introduce their next-generation technology. So that is why we are more confidently guiding you on sales, and we're pretty confident that the orders backing up the sales will follow in due course. Satya Kumar - Crédit Suisse AG, Research Division: Got it. And then a quick follow-up on the 3 comments you answered earlier on that. You think it's not unrealistic that you could see 100,000 wafer starts or wafer capacity expansion in the second half of this year for memory? Could you talk a little bit about what you're baking into your guidance in terms of the wafer start expansion for memory for your EUR 2.7 billion or the 100k going to present some sort of upside to what the current forecast is for you?

For the DRAM business, what you have in the current guidance is a reduction of wafers by about 100k also and replaced by about additional 100k of NAND. That's what you have in our basics. And we believe these numbers are conservative at this very moment. On the DRAM side, which is visible now because we are negotiating what we call availability, which is another way of saying we're getting orders. And on the NAND side, we haven't yet negotiated orders, but when you talk to the customers, their statistics of bids out is planned, is higher than what we think, we have planned for them. So although this is not yet to a point of discussing offers in NAND, there is a discrepancy about the conservative backlog and what we have in our guidance and what was said in terms of bids.

Our next question comes from Sandeep Deshpande. Sandeep S. Deshpande - JP Morgan Chase & Co, Research Division: JPMorgan. Just a couple of questions. Firstly, regarding your laser roadmap, I mean, you've talked about that you have a 55-watt laser, which is now working. Can you talk us through how the 55 is going to transition towards 105 through the year? Do you have a roadmap? Is it 55 to 70 or 80? So how we should be monitoring this through the next, say, 12 months as you transition? And then I have a follow-up as well.

Yes, so, of course, we have a roadmap, and of course, I won't tell you, Sandeep. Because if not, you are going to ask me every minute whether I am on top of it. How can I answer nicely your question, though? Yes, we are seeing -- we are going to be forced to present to you, on a quarterly basis, improvements. And we are significantly certain that there is improvement to be shown to you towards this, I would say, minimum performance of 69 wafer, which is 105 watts, by mid of next year, when we really need to be having this done in production. So in other terms, I think we are now in the conservative curve. We'll see improvement every day, but we don't really want to be measured on whether we have one wafer less or more because, indeed, things are not that obvious. You have multiple architectures being developed at the same time, in parallel, so it would make no real value or understanding for you to know whether we are on track or not depending on this internal tracking method that, of course, we have.

I think like Eric said, we are going to inform you quarterly where we are. Also, given the fact that we are confident about reaching our, like Eric said, conservative target of a 5-watt, i.e. 69 wafers per hour. We will inform you every quarter where we are, and you can judge the progress. But I don't think the forum start discussing the, let's say, program timetable. That's not that relevant. Sandeep S. Deshpande - JP Morgan Chase & Co, Research Division: Okay. And if I ask a question following up on one of your comments in the last conference call, Eric, you talked about that the EUV tool went in high-volume production, could sell along with all the options for as much as EUR 100 million. You just talked about this potential EUR 800 million of additional revenue from YieldStar. So is that included in that ASP calculation of EUR 100 million that you're talking about for, say, 2015 or 2016? And secondly, do you have a potential size that, in 2015, you can do this many EUV tools and you think that there could be potential demand for that?

Yes, I can be free now, as you know. So I think the answer is yes and yes. The EUR 100 million potential EUV ASP will be justified by its content, but also by its spec. And remember, this EUV machine is going to have a multiple NA -- or not NA, but illuminator settings with part of the ASP of EUV machine, and that is not part of my EUR 800 million. So I make a difference between a machine with, I will say, attach options which are 100%, and that could go, okay, between EUR 90 million and EUR 100 million a machine. And then a set of options which are not 100% attached, which are attached only for a set of process issues. And that is where I say that, at some, point we could see EUR 800 million. But I am not saying that it's going to come. So it is, in fact, duplicative as you can add them up, right?

And also, Sandeep, as we've said in earlier calls also, when we talk about the sales number of Holistic Litho products, some of them will be shown in the ASP of the tools, and others will be shown as system enhancement sales in field options. So it's clearly impossible for us to give you the exact impact of that number that Eric quoted, which, of course, is not a commitment for us. But if you say how much of that will translate into the ASP and how much of that will be the field sales.

And regarding your question about EUV ramp, we absolutely, and it is proof of the stabilization of EUV program, we committed a number of units, about 25, 30-ish in 2015 and about 60 in 2016, and we confirm that this is the proper simulated number. So of course, it's a simulated number based on layers and numbers of wafers, but it is still there.

Next question comes from Francois Meunier.

It's Francois from Morgan Stanley. I got a question about the EUV shipments for this year. I think last time, you were talking about shipping up to 11. This year, now it's gone down to 6. Obviously, they will be shipped in 2014, I guess, but what is the difference? Just maybe customers being a bit cautious on the tools and preferring to wait? Or just what -- could you explain what happened basically?

Yes, it's [indiscernible] customer side. It's -- we have 11 scanners on the production floor that people who recently were in our factory actually count them in a very advanced state of production. The thing that is currently missing is the speed with which we can produce the EUV sources. And there is a delay, not so much because of the technical issues that we have with the source, but it is the supply chain issues. So we are also ramping up the supply chain for key components of the EUV source. And that is just falling behind a bit, which is only 1 or 2 months, but it just pushes the shipments into 2014. That is the only issue. That's more logistical issue than it is any customer-related or fundamental technical reason.

Okay, that's very clear. If I may, I got a follow-up about the 7 EUV tools, which are -- it seems like it's -- where do you use this commitment for 2014? What is the difference between a commitment and order? What does it take to move from one to the other?

Okay. So I tried in the script to be subtle and that didn't work. So I'm going to be [indiscernible]. At this moment, there is indeed an acceleration of interest of EUV because, in fact, we are seeing to make our duty and have enough data that's say it's possible. Then the customers are starting to see that it's a huge opportunity, but a huge problem. Because if they choose to execute EUV with less enthusiasm than their competition, they can get significantly behind, okay? So it's very clear that you have a risk level, which is not yet proven. We still have to do some effort every quarter, as we said today. And on the other hand, you still need to take the right risk to avoid being late in developing a chip, which will be significantly worse if you don't use EUV to reach its maximum extent. But if you plan it and EUV is not exactly performing, you did the wrong thing. So that is an obvious question. At this moment, we see a huge momentum in the discussion of how many units and layers will need to be reserved for whom in due time. This discussion is much more important than the administrative PO question. So I hope I made myself clear that at this moment, the paperwork is past a second priority to how many layers? When? What's to expect? What can you do or can you guarantee? What are the exact timing? Can I go with a single node? Can I -- Am I going to do what they call a mid-node or something that -- the back up? Or so the back up, et cetera. So all that discussion has intensified. So I think that's a positive statement to say to you, it's been commitment and we were discussing more. And we have not pushed for paperwork.

And you could say, if you would want to compare this with our normal PO process, normally, our POs are non-cancellable and -- because there maybe is a big fine. These are commitments, you could say, POs with a possibility of a customer to actually say that tool needs to perform to certain specs, which you could call it a cancelable PO, if you want to make that particular understanding. That is what we call a commitment.

Depending on the spec...

Yes, depending on the specs and on the performance.

The specs being discussed depending on the numbers of layers. So it's a fairly complicated activity that's being done at this moment. It's a huge positive.

Our next question comes from Timothy Arcuri.

Cowen. A couple of things. First of all, on the ASP for EUV. You said last quarter that you were going to do 450 to 500 this year, which was typically revenue recognition on about 7 tools. You were going to ship 11 and you were going to rev rec 7, which implied an ASP of about EUR 70 million. Now you're going to rev rec 280. So is the ASP still EUR 70 million, such that now you're talking about shipping 6 but recognizing revenue on 4?

Yes, that is absolutely correct. So you could argue that 223, we had some supply chain issues with the source. The source comes in later, so that we cannot book in the revenue this year. Benedict Pang - Caris & Company, Inc., Research Division: Perfect. Okay, great. Secondly, on NAND. Both you and Lam are talking about this 100k worth of wafers added this year. If you just write a simple supply-demand, that's going to add a lot of NAND supply this year, such that it would really pretty dramatically change the NAND supply-demand balance, and it's hard to believe that Sendo's stock would be trading at a peak, if that was really true. So I'm just wondering, whether you think that the industry can physically add 100k worth of NAND wafers and still maintain supply-demand balance?

It's only 10%...

It's less than 10%.

It's less than 9% and 10% wafers. This is the only way to do 36% big growth. So at this moment, you will hear that they want to do even more than 36%. So technically, it makes sense. So you can imagine that, of course, if PC continues in the wrong trend. But on the other hand, look at the new PCs, their content in NAND is much bigger. So now this thing is going to be driven not so much unit of PC, but it will be driven by content per unit. And the content per unit is clearly going to double easily with the number of -- I mean, the need for solid-state drive minimum size. So it can be less conservative than us and take 10% less. But again, by the way, 10% less is one thing. So yes, I'm okay, with one machine less, plus or minus.

All right. Okay, okay. And just 2 more quick ones. The tax rate this quarter was lower, and I was thinking something like 13% for the rest of the year. That's where I thought tax would come in. So should I still think 13% per quarter for the rest of the year?

No, it's not. It's more like 8% to 9%.

8% to 9%, okay. And then on R&D, you had previously said something like EUR 750 million to EUR 800 million for the year, but you're tracking a little below that. Should we still think EUR 750 million to EUR 800 million for the year?

You can do EUR 750 million to EUR 800 million is okay, but it's logical that it tracks a bit lower than the average because we are ramping up. So it's normal that in the R&D profile, you would see it going up over the quarters. Between EUR 750 million and EUR 800 million is what we said, and that's still where we are.

Our next question comes from Didier Scemama.

It's Merrill Lynch. First, I'd like to say, congratulations to Peter for his promotion. And also many thanks to Eric for the quite amazing work you've done in the last 9 years or so. So we'll miss you, most particularly on the conference calls. And I just like to...

Thank you. Why amazing? Amazing like you're surprised?

Not good enough. No, I'd just like to -- just take a step back here. So if I take basically your comments so far on your quite punchy [indiscernible] punchiness and confidence on EUV, what you just said. How should we think about 2015 -- EUV deliveries in 2016? Are we still talking 30, 35 tools and 60? Or can we think about a higher number because of, what you said, the sort of various options and plans? Numbers of layers? How should we think about that? And I've got a quick follow-up.

Well, as I said to Sandeep, 25 to 30 in 2015, 60 in 2016 is the nominal simulation. Does not include any NAND. Only includes 2 layers, I think, of DRAM, and only includes 7 layers of logic, 5 to 7 layers of logic, with an aggressive uptime. So in other terms, if we miss 7 layers, 5 to 7 layers, we may not have the right uptime anyway, because technically, that means the customers will not want to use that many machines because they don't feel comfortable. But then, that means the machine will be economically not as good, so we'll ship more of them. So that's why the simulation tools come around about the same amount.

And you have to also remember, Didier, that these numbers are bouncing against the ceiling of our capacity, of our production capacity. So like I said in my introductory statements, we'll have to start spending money on some production facilities to be able to meet those numbers. And they're currently planned to do what Eric just -- especially output those numbers that Eric just mentioned. So there's not much upside to that.

Got it. And then my other question is regards to some sort of the DSA technology as a potential threat to EUV. What are your thoughts on that?

No. DSA is like double -- sorry, how do you call this, double patterning of space. It has an opportunity to steal 1 layer or so from us, but this is not -- and it's only on the NAND arena, could be 1 layer in logic for very, very specific designs. It's within the noise, say, we will prefer that if it's not working, then we have 5% more sales, 7% more sales, if they are not working. If they work, you won't probably see the difference anyway.

And then, I guess, I think what we understand from DSA that it's another slam dunk either. I mean, there's quite significant specificity issues that still need to be mastered, so this is probably, you could say, is a promise to our customers, but it's not really production-worthy at this moment.

Fine. And then maybe final point, I think you sort of touched on that, Peter, on your prepared statements. But in terms of foundry bookings, it looks like it came down a bit in Q1. But obviously, to meet your revenue target the second half and given the sort of 16-nanometer of FinFET and 14-nanometer of FinFET that the foundry customers have got in mind for next year, do you think we should see acceleration that could -- proper acceleration of bookings in Q2 and Q3? Or as you mentioned before, because basically, the customer base is so concentrated and you know what they want and they know what you want, basically there is no orders, and therefore, you're more confident on revenues than on bookings?

I think you get the answer in your last sentence. We are guiding on sales because the sales is driven by the need of our customers to be extremely specific on when they need what for their next technology ramp. The order process is almost an administrative process, and sometimes you have customers placing orders 6 months before shipments, sometimes 3 months, and it's an application, it's 2 weeks for whatever the reason. And we need that order because we need to have the shipping documents that go with, too, when it pass customs. So it's almost an administrative thing. But we are highly confident that the order profile will mimic our sales in our profile that we are guiding you.

Our next question comes from Stephane Houri.

This is Stephane Houri from Natixis. I'm sorry to come back on the EUV delivery schedule, but I heard you said that 25, 30 for 2015 and 60 something for 2016 the right number. You also gave some elements for 2014. Could you be a bit more specific on the delivery number that you have for EUV in 2014?

As we said regularly, once a month -- 1 a month. So for the people who follow me closely and remember the notes, I used to say 1 a month in 2013 and '14. That's 24. So now I say 1 a month as of now. This is a bit less because we have lost basically 6 months of this run rate. But the interesting bit is that we've lost, I would say, an R&D comfort at the customers, so that is bad news. This R&D comfort, we lost it. But then the customers created FinFET as an opportunity, so there are less pressure on them. And so they can build the node without EUV with FinFET called 14. There's 22 nanometer with a better architecture. But now comes back the specter of the new -- the real node beyond 14, and that's one having to prepare 25-ish tools in 2015. That is the one which we have to commit if they really go into production with that node.

Okay. And if I may, I have a follow-up. You have repeated that you expect the Cymer acquisition to be completed by the end of next month. If it was not the case, would it imply mechanically a delay on EUV?

No. Of course, not. No. We've been working with Cymer closely on this sub projects. And at this moment, exactly for the law, without infringing on the rights as a separate company, we've been acting like a supplier and the customer having partitioned the job correctly, so that we can execute. But this has been the case since -- even before the merger.

Yes, I think the closing will have a significant effect over time on the industrialization of the EUV source, which, of course, is very important to support the ramp in later years. That's where you really would see the benefits. So that would enable us much better to reach those 25 to 30 goals for 2015 and 60 in 2016. But it will not have an impact on our current shipment schedule.

Our next question comes from Weston Twigg.

Pacific Crest. Just a couple of quick questions. Just on EUV for new timing, first. For this year, it sounds like all the revenues should be lumped into the second half, those 4 tools. And then in 2014, if you're shipping 1 per month, can we actually expect revenue to happen at 1 per month? In orders words, we have the supply chain issues worked out on the source side?

It could, but to say -- we could also have -- because we have, you could say, a block of shipments, if we can accelerate some of the supply chain issues, you could see more -- you could see 2 per month, so -- in one particular month. So -- but I think, too, for your planning purposes, it's safe to just do 1 per month.

Okay. And then for this year?

Exactly as you said.

As you said, it is H2.

Okay. And then as a follow-on, just looking a little bit further down the road. ESMT keeps saying they need 100 wafers per hour on EUV, and you're talking about getting and picking it up in 2015, 2016 from logic and foundry. But it doesn't sound like you'll necessarily or you've -- that you've yet committed to have 100 wafer per hour EUV tool available then, that you're still targeting the 69 wafers for about midyear 2014. So I'm just wondering, can you meet that kind of a roadmap, 100 wafers in hour in 2015 to match the timeline for a company like ESMT?

If you remember what I said, our commitment is 69 wafer per hour in 2014, upgradable to 125 wafer per hour 2 years later, so 2016. So if you make an average, you'll come up with 100 wafers. About -- in fact, more than 100 wafers per hour during the period of time. And remember this node, the famous node, [indiscernible] whatever FinFET and things, is really ramping in 2015. So the customers usually do not need the speed on the first day -- on the first year. In fact, that's also when they optimize their process. So by saying 125 wafer per hour, we are a bit above the 100 wafer per hour business.

Okay. Would it be safe to assume that the customer would have a contractual commitment from you to give them the confidence in that kind of a ramp to meet these types of milestones, even if it's not applicable early on?

Absolutely. The PO we look like the price of the machine is X for 69 and this Y for 125, we take the risk of X and Y, if you know what I mean. And both X and Y, X will not have the same margin as Y, but X will be manageable and Y will be good.

Our next question comes from Andrew Gardiner.

It's Barclays. Another sort of quick follow-up on EUV. The initial 4 commitments that you've mentioned last year were from DRAM, if I remember correctly. The additional 3 that you've now signed, is that for logic? And can you also talk about, of course, the breadth of the customers in terms of number of vendors or manufacturers that have now committed to this?

Memory is on the 7. I think it's -- memory is difficult to know because as you can imagine, one customer is doing logic and memory at the same time. So at this point, I could imagine 3 to 4 is memory and 3 to 4 is logic, and the one in between, as you can imagine, is the customer who can do both.

Got it. Okay. Also just in terms of sort of the increase in demand or increase in inquiries that you've been seeing recently. I know you've been keeping a buffer stock on hand of tools to sort of decrease your lead time or decrease the time from actual firm order to shipment. Is the amount that you've got -- so that flexibility, is that enough to get you to full year guidance? Or is the potential for upside to that, depending on just how quickly demand to -- just sort of wondering about your flexibility there in terms of the lead time?

The production capacity we have is perfectly sufficient to do the full year guidance, and it also poses upside. So if it would get more, we would be able to ship within, let's say, a standard lead time, because at the beginning of the year, we're confronted with a lot of uncertainty as to where these markets were going. And we probably mentioned previously on the call, we decided to add some buffer tools during the period, we will be able to meet, I would say, a very reasonable part of the potential additional amount.

Okay, that's clear. And just finally, one quick one. I mean, you've been talking about expanding the capacity for EUV as we come to the coming years. How should we be thinking about capital spending for this year now and then next year given that rate, a material ramp that we should be forecasting?

Yes, we'll be spending about that 18-month period for the facilities and it's the -- it will be about EUR 250 million to EUR 280 million, of which maximum of EUR 50 million will be spent this year. The rest will be spent next year.

Our next question comes from Amit Harchandani.

Amit Harchandani from Citigroup. Before I go ahead, congratulations to Peter on his new role, and thank you, Eric, for your work that you've done at ASML over the last so many years.

Thank you. 8.5.

My first one is probably digging a bit deeper on the foundry side. Clearly, the customer in Taiwan has ordered strongly and continues to order strongly. But besides that customer, how do you see foundry demand from some of the other customers, let's say, in Korea because there was some hesitation at the start of the year? And as a follow-up on the EUV side, clearly, imaging and overlay performance, as you reveal, has been quite good, and productivity or throughput is based on the source. But how do you see in terms of defect, particularly mass effects and the rest of the EUV ecosystem? Is it shaping up and progressing in the manner that you expected it to do so? And do you foresee the need to do anymore M&A to achieve your target for EUV?

Very good. Thank you. it's usually difficult for us to talk about customers, and now it's so granular. I mean, you know there are 3 large foundries. So I don't know what to tell you. The only thing to tell you is one of them, as you know, is ahead of the others in a huge buildup of 20-nanometer because of a number of potential customer lineup that we have, so they have to build it up, which forces the other competitors in this business to buildup. Indeed, the orders haven't been received from the other 2, but we know that they have to. So at this moment, what's happening is a bit good timing, where you have one big customer ahead, and then the second one comes in and will allow us 3 to 6 months of delay, which is good. And then the third one will come with 3 to 6 months delay, and that would probably help us extremely well in 2014. So that's why Peter's job is going to be pretty easy in 2014. On the subject of the EUV infrastructure, it's no -- it's not going to be a walk in the park because in this business like we had in immersion, you are -- the customers are going to discover issues. Resisted is going to be good as expected. Mass defects will -- blank defect will be the problem. Defect in writing mass will be a problem. Defect in processing will be a problem. Inspection is a problem, but they now have a solution entering for a while. But it is absolutely standard difficulties. Today, as you heard, 9 months to 12 months ago, everybody said FinFETs will not work. It's very difficult. Even Intel had difficulty to build up this technology, et cetera. But then 1 year later, they have mastered it. So we're probably going to be in the same mode here of huge amount of work. In fact, this is why -- what the problem of EUV is. Addition -- the problem of EUV is not that people question whether the machine we work in 2015 when, in fact, it is due. I guess, at this moment, everybody's getting [indiscernible] for that. But you would prefer to have a working machine now, so that you have 2 years to create the recipes. So it's an accumulation of recipe improvement that, that is going to happen to resolve the issue of infrastructure. So indeed, we don't plan to buy anybody or acquire anybody. They won't have that chance.

Okay. And then just maybe quickly, if you could share with me, you said 20-nanometer might turn out to be a big node. Any numbers that you could maybe share with us? What your plan -- thinking about how do you see that particular node shaping up in terms of wafer capacity?

That an impossible question, because what you call 20-nanometer, I call 20, 16, 14 because the same node in lithography will include different transistors, FinFET transistors or whether a tri-gate or whatever they call it. And it all will have a little tri-set to make this. So this thing would -- if it is normal node, then it is driven by so many detectors that we see now in the portable world where everybody has a tapeout plan for 2014, we will expect a minimum of 300,000 wafers. If this is above that, then this is, again, great for Peter, because the lithography intensity is such on that node that the multipliers for it, similarly, is huge. So if you go above 300,000 wafers, we don't need EUV. It's a big one. So -- but at this moment, we only simulate this to be a normal node, 300,000 wafers. And then getting into the node, which has all the goodies I mentioned, particular the scaling factor.

Gentlemen, frankly we've run out of time. We -- IR already has a queue of calls lined up here. But I would invite you, if you didn't get to your question asked and you think we can help you, to feel free to text us or give us a call, and we'll do our very best to get back to you. But please be patient as we have quite a few stacking up now. So with that, on behalf of ASML's Board of Management, I would like to thank you for joining us today. And Kristin, thank you for managing the call for us. And if you could formally conclude the call, we would appreciate it.

Thank you. Ladies and gentlemen, this concludes the ASML 2013 First Quarter Results Conference Call. Thank you for participating. You may now disconnect.