Ladies and gentlemen, thank you for standing by. Welcome to the ASML 2018 third quarter financial results conference call on October 17, 2018. Throughout today's introduction, all participants will be in a listen-only mode. After ASML's introduction, there will be an opportunity to ask questions. I would now like to open the question and answer queue. If you would like to ask a question, please press star one to register. If you would like to withdraw a question, please press star two at any time during the call. Your questions will be answered in the order that they are received. If any participant has difficulty hearing the conference, please press star zero for operator assistance. I would now like to turn the conference call over to Mr. Skip Miller. Go ahead, please, sir.
Thank you, operator. Good afternoon, good morning, ladies and gentlemen. This is Skip Miller, vice president of investor relations at ASML. Joining me today from ASML headquarters in Veldhoven, the Netherlands, is ASML CEO, Peter Wennink, and our CFO, Roger Dassen. The subject of today's call is ASML's 2018 third quarter results. The length of this call will be 60 minutes, and questions will be taken in the order they are received. This call is also being broadcast live over the internet at asml.com. A transcript of management's opening remarks and a replay of the call will be available on our website shortly following the conclusion of this call. 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 meaning of the federal securities laws.
For a discussion of risk factors, I encourage you to review the safe harbor statement contained in today's press release and presentation on our website at asml.com, and ASML's annual report on Form 20-F and other documents that is filed with the Securities and Exchange Commission. With that, I'd like to turn the call over to Peter Wennink for a brief introduction.
Thank you, Skip. Good morning, good afternoon, ladies and gentlemen, and thank you for joining us for our Q3 2018 results conference call. Before we begin the question and answer session, Roger and I would like to provide an overview and some commentary on the third quarter, as well as provide our view of the coming quarters. Roger will start with a review of the third quarter financial performance and with some added comments on our short-term outlook. Then I will complete the introduction with some additional comments on the current business environment and our future business outlook. Roger, if you will.
Thank you, Peter, and welcome everyone. I will first highlight some of the third quarter accomplishments and then provide our expectations for the fourth quarter of 2018. Q3 net sales came in at EUR 2.78 billion, which was towards the higher end of our expectation. Net system sales of EUR 2.08 billion was a bit more weighted towards memory at 58%, with the remaining 42% from logic. EUV revenue of EUR 513 million was from five shipments. Installed base management sales for the quarter came in at EUR 695 million. Gross margin for the quarter was 48.1%, just above our expectations, reflecting the strength of our DUV and applications business, as well as the progress in EUV profitability. Overall, R&D and SG&A expenses basically came in as expected, with R&D expenses at EUR 397 million and SG&A expenses at EUR 122 million. Turning to the balance sheet, EUR 362 million worth of shares were repurchased in Q3.
This leaves around EUR 1.7 billion of the 2018-19 share buyback program remaining. We ended last quarter with cash equivalents, and short-term investments at a level of EUR 2.95 billion. Moving to the order book, Q3 system bookings came in at EUR 2.20 billion. Memory order intake continued to be strong, 64% of total value. Logic made up the remaining 36% of the bookings. We took five new EUV orders in the quarter, which contained a mix of both logic and memory. With that, I would like to turn to our expectations for the fourth quarter of 2018. Expect Q4 total net sales of about EUR 3 billion, leading us to expect another record year with close to EUR 11 billion of revenue. Our total net sales forecast for the quarter includes around EUR 500 million of EUV system revenue from five EUV systems.
We currently expect to ship six systems in Q4, including one EUV system to a collaborative research center, imec, which will not be recorded in revenue but will be used to settle R&D services from imec. Q4 will be our highest EUV shipment quarter to date, bringing the total to 18 systems in 2018. Due to a combination of end-of-year production challenges and customer readiness, we now expect a couple of the originally planned 2018 systems to ship in early 2019. We expect the EUV order flow to continue next quarter, such that we will basically have our 30 systems planned for 2019 covered by purchase orders by the end of this year. We expect our Q4 installed base management revenue to be similar to last quarter, at around EUR 700 million. Gross margin for Q4 is expected to be around 48%.
Taking Q4 guidance into account, gross margin for the full year would be around 47%, which is a step up from last year's 45% gross margin. This reflects the strength of our DUV and applications business, as well as continued progress in EUV profitability. The higher R&D expense for Q4 of about EUR 420 million are due to an acceleration of the NXE 3400 roadmap and the High-NA EUV program. SG&A is expected to come in at about EUR 135 million. We remain excited about 2018, as the customer demand for our products continues to be strong. We look forward to delivering another record year with continued strong growth in both sales and profitability. With that, I'd like to turn the call back over to Peter.
Thank you, Roger. As Roger highlighted, we had another good quarter. We expect the fourth quarter to be even stronger. With the current guidance, we expect that our sales for the year will be close to EUR 11 billion, and that our profitability will improve over last year. We continue to see strong demand for our products in both Logic and memory, as witnessed by our strong order book. Logic customers continue to ramp the 10 nanometer node and are also starting to ramp 7 nanometer. As customers prepare to ramp up the 7 nanometer node, it not only drives DUV demand but also drives significant demand increase for EUV. In DRAM, customers are continuing with technology migrations, as well as adding wafer capacity additions to meet bit demand growth, as evidenced by our strong third quarter order intake for memory.
We believe that the limited number of wafer capacity additions by a limited number of customers, combined with a healthy demand for DRAM bits, should not lead to a structural overcapacity in this industry segment. In NAND, significant 2D to 3D conversions have taken place next to investments in several greenfield fabs. This is likely creating a period of some digestion, as we mentioned in prior quarters. With regards to China, we continue to see strong demand for a broad suite of our products. The China region has delivered around 20% of our sales this year and is on track to set another record revenue number. This is driven by both multinational customers as well as domestic China customers. All five domestic customers that we discussed in prior quarters have at least some pilot capacity in place now and are looking to begin ramping next year.
We believe this region presents a significant growth opportunity under the assumption that these ramps of the domestic customers are successful and that the non-domestic customers will follow through with their investment plans. On the ASML product side, let me start with an update on our EUV business. In EUV, we continue to make good progress. We have multiple NXE:3300 systems at customer sites that are running at 125 wafers per hour or higher and are ready for high-volume manufacturing. Availability is progressing in support of customer volume ramp, with a clear focus on machine consistency. The overall progress has led to the decision to accelerate our EUV roadmap. We are, as a result of this, now planning the introduction of our next generation 0.33 NA EUV system, called the NXE:3400C, in the second half of 2019. This system will deliver a productivity of over 155 wafers per hour.
We will talk more about the performance specifications and the roadmap during our investor day next month. As Roger mentioned, we continue to increase our shipments per quarter and plan to ship six systems in Q4, bringing the total to 18 systems in 2018. As we mentioned in earlier calls, this year, our production output is heavily back-up loaded, which has led to some production output challenges, combined with customer fab readiness logistics. We now plan to ship a couple of systems originally planned in 2018, now in early 2019. Our shipment plan for 2019 remains at 30 systems, as we now have an increased mix of the NXE:3400C systems in the second half of 2019, which will enable a significantly higher wafer output capability than the earlier specified 125 wafers per hour.
With this higher productivity, we expect to be able to meet our customers' current EUV capacity plans in 2019. As Roger mentioned, we expect order flow to continue next quarter and expect that our 2019 amount for EUV covered by orders by the end of the year. In DUV, the introduction of the NXT :2000 system into the market is making significant progress and will be used in volume manufacturing for both memory and logic. We're also seeing significant demand for our dry products in support of a number of greenfield fab ramps in China and other regions. In our application business, we continue to see growth across our full portfolio of software and metrology products, notably related to the adoption of our YieldStar 375 system, expanding from logic and DRAM now also into 3D NAND manufacturing.
To summarize 2018, we expect the growth to continue from Q3 to Q4, set us up for another record year in both sales and profitability. Regarding 2019, it's a bit too early to provide detailed guidance, but I will provide some qualitative comments regarding our initial views. We continue to see strong demand for our products in both memory and logic, as supported by our bookings. DUV demand continues to be healthy in memory, as discussed earlier, and we expect DUV demand in logic to further strengthen in 2019, driven by the 10- and 7-nanometer ramps. Furthermore, we expect continued growth for our applications business with expansion of both metrology as well as software products. EUV demand continues to be driven by logic, but also with a clear opportunity in DRAM where we meet our availability and productivity targets.
EUV revenue growth is expected from both a significant increase in unit shipments as well as a higher ASP of the NXE:3400C, which shipments are planned starting, as we said earlier, the second half of 2019. Our install base will continue to grow, driving increased service revenue. Furthermore, we expect customers to take advantage of system performance upgrades on their install base to maximize capital efficiency. Our current view of the overall business next year remains positive. We expect the first half to be somewhat similar to the second half of this year, with business strengthening in the second half of 2019. Putting this all together, we expect another year with good growth opportunities.
I think we're well on track to achieve our 2020 targets with significant growth potential beyond 2020, and we plan to communicate the size and the extent of this growth opportunity through 2025 at our Investor Day, which we'll hold on November 8th this year. With that, we would be happy to take your questions.
Ladies and gentlemen, the operator will instruct you momentarily on the protocol for the Q&A session. Beforehand, I'd like to ask that you kindly limit yourself to one question with one short follow-up if necessary. This will allow us to get as many callers as possible. Operator, may we have your final instructions and then the first question, please?
Of course, sir. Ladies and gentlemen, at this time, we will begin the question and answer session. Again, if you have a question, please press star one to register for a question and star two to withdraw a question from the queue. If you're using speaker equipment today, please lift the handset before making your selections. One moment please for the first question. The first question comes from Mr. Krish Sankar. Please state your company name followed by your question.
Go, Krish.
Hello. Can you hear me?
Yes, we can. Loud and clear.
All right. Thanks for taking my question, two of them. First one, Peter, it looks like your demand from your memory customers has been very strong so far, both in terms of bookings and sales, and you're also guiding to a strength into 2019. Just sort of help, can you reconcile what's been going on in the memory industry with NAND pricing weakness and potential CapEx cuts in NAND and DRAM, and how do you reconcile that with your numbers, and where do you see the strength in 2019? Is it going to be DRAM or NAND in the first half? I had a follow-up.
Okay. Well, I think you're all throwing memory on the one big heap, whereas, of course, you need to separate, as you indicated, between NAND and DRAM. Let me talk about those separately. On DRAM, we haven't seen CapEx cuts. We have seen, in this year, some push-outs but also pull-ins from different customers. You could argue it's customer specific, but we haven't really seen a change in the DRAM memory demand for our products this year. We don't see it also in 2019. Now, we always seem to forget recent history, let me talk about DRAM. You have to put it all into perspective.
Up to and including 2016, there's been a significant conversion of DRAM into 3D NAND, which resulted in reduced DRAM capacity, which actually also led to a reduced DRAM memory spend of about 30% in litho and also a 10%-15% reduction in the wafer capacity at our customers. Now, in 2017 and 2018, customers have been working to recover this wafer capacity and to increase the bit supply. The bit demand also was higher than anticipated. This required, indeed, a much higher litho spend per unit bit growth. That is due to a combination of increasing litho intensity at these new nodes due to, say, an increasing number of critical layers, which includes double patterning now, and the slowing of the shrink roadmap, which actually means you don't get the same number of bits through shrink.
You actually get less, both leading to higher wafer capacity additions to meet this 20%-25% demand bit growth. This is what we're seeing. In that context, the high investments in DRAM from our customers is not a surprise, and it's also what we are seeing in 2019. Now, on 3D NAND, over the last couple of years, almost all the 2D to 3D conversions have taken place. They've actually happened, and those were very significant. Now, we have not participated in that as a litho supplier because we basically use the same litho. Next to that, there were investments in greenfield fabs. If you add those two together, it's the 3D conversions and then the greenfield fabs, and that has indeed led to a level of capacity.
Don't forget, these are big fabs, so there's step-ups in capacity to a temporary weakness of the 3D NAND pricing, which we've all witnessed. That is the digestion that we're going through as we speak. That's logical, especially if you look at the number of greenfield fabs that have been opened and the wafer capacity that have been added to the industry. It's quite normal. Don't forget that the growth rates of 3D NAND are particularly good at 40%+. I think this is how we look at the market, and this is why I also think that it's not a big surprise that our customers are still significantly spending on increase in wafer capacity both for DRAM and to a lesser extent today, we need them.
Got it, Peter. That's very helpful. Just as a follow-up, if I look at your commentary on calendar first half 2019 similar to second half of this year and centering the second half of next year, the fact that DUV should be strong in memory and further in logic, is it fair to assume that DUV units next year should be higher than this year?
Well, I think the DUV units will be at least at the same level as this year, whereby I think the mix which was this year will be skewed towards memory, will probably skew a bit more towards logic, although it's too early to say which part of the industry segment is going to be the largest. In DUV, I think we will see at least the same number of DUV shipments and sales in 2019 as in 2018.
Thanks, Peter. Thank you very much.
The next question comes from Mr. David Mulholland. Please state your company name, followed by your question.
Hi, it's David from UBS. Just coming to the comments you made on EUV and good to see the progress on the 3400C. I just wanted to clarify a couple of things. Firstly, of the bookings that you've seen in the quarter for EUV, are those still the 3400B or are you now booking the 3400C? Then as we look into 2020, I wonder if you comment just on what impact in the way that this is potentially slightly dampening the number of tools needed in H1 2019, what that might mean on 2020. Obviously, we've assumed some of that's made back on pricing. Where do you end up in kind of revenue expectations for 2020 from EUV as you kind of net those two, higher productive systems, but potentially higher value?
Okay, good. Well, I think everything we're booking is 3400C. We're not taking any orders for Bs, because the only thing we will sell as of the middle of next year is Cs. Yes, there will be a higher potential productivity coming out of these systems, which of course will have an impact on the number of systems that customers potentially want if they look at their wafer capacity that they are planning for. Now, having said that, there's also a flip side to higher productivity and higher uptime, and which is cost. The cost is actually going down with these higher productivity tools, which actually means that it opens also a possibility to add, one, in logic more layers, and two, in memory DRAM, we start using EUV in DRAM.
Generally, you could say if you have more than 2,000 wafers per day productivity on a DRAM system, it becomes attractive to basically start using EUV for several layers in DRAM. That will drive the 2020 number. I think what is important for us is that we execute. That's why we pulled the R&D in, because we want the 3400C ASAP, because it will, one, as you indicated, provide us with a higher value. Two, it will also provide a higher value, i.e., lower cost to our customers, which will drive the demand for EUV. Which means that we still stick to our production capacity of 40 units in 2020. I think the final 2020 number will be a function of our successful introduction of the productivity and the availability metrics that we have currently in our targets.
Maybe one just quick follow-up. You haven't commented as directly in numbers in terms of the progress on availability for EUV. Obviously, it seems like you're saying it's at the level you need for insertion with customers. Last quarter you were saying you had to get to over 90%.
Yeah.
Can you quantify where you are and where you're heading?
Yeah. I think we've said we want to have a target of 88% availability by the end of the year. This is where we're heading to. I think with the 3400C, where we go over 90%, I think we have a target of 92%. I think what I said earlier, we need that 90% threshold. That's what we said in earlier calls. That it's our opinion today that with the current availability targets, customers will use EUV in HVM. Logic is very simple because without EUV, and I just refer to comments that was made by some of our customers, without EUV simply won't work. There is so much seven nanometer demand or if you want, or seven-plus, or five nanometer demand, that you cannot escape using EUV. They will use EUV at 88%.
We would love it to be higher, and it will be higher, but that is not a make or break number.
That's great. Thanks very much.
Our next question comes from Mr. C.J. Muse. Please state your company name, followed by your question.
Yeah, good afternoon. Thank you for taking my question. I guess the question, if I could go back to your 2019 outlook for DUV. It sounds like you're now saying kind of first half similar to second half in explorative growth into the second half of 2019. Just curious, is that a changed statement from your views three, six months ago? If so, what has changed, I guess vis-a-vis DRAM contribution, advanced logic in China?
I think it has not changed. We haven't qualitatively guided any trends for 2019 until today. I think this is the first time. In truly, of course, we have this outlook. I don't think it has changed that much. Absolutely not. In China, nothing changed in the sense that what Chinese customers were planning, let's say, this time a year ago On 2019, they're actually executing on. You could argue that their execution of their first lines and the pilot lines have actually gone well. I think there's no change. I think there was no significant change, nor in memory, nor in logic.
Very helpful. I guess, as my follow-up, can you talk to how you're expecting linearity of shipments for the 30 EUV tools in 2019, and how we should think about the progression of gross margins in that same timeframe? Thank you.
Linearity is what you would expect with a ramp. This year, if you look at the quarter, we do three, four, five, six. I think this is the kind of linearity that you would also expect next year, whereby 3400 C, of course, is the model that customers would really like. You would clearly see also the second level of the year, the math for that product going up. On Roger, on your
On gross margin, I think we've articulated the target of 40% there for 2020. I think we're on track to get there. I think as we've mentioned before, four levers to get there. The first lever obviously is higher ASP, which is to a very large extent, correlated with the productivity and the throughput of the machine. That's a major driver of the gross margin. Second volume. Fixed cost coverage obviously increases to the extended volume for the ramps. Third, learning curve. We're already experiencing that, and we continue to experience that into the next couple of years. Fourth, service. Service revenue and service margin will go up as well. The combination of those four levers, we believe, gets us to the 40% target that we've articulated before for 2020.
Okay. Thank you.
The next question comes from Mr. Mitch Steves. Please state your company name, followed by your question.
Hey, thanks. Mitch Steves from RBC. I just had a quick question on EUV. Basically, because you guys are all up by about two units here in 2020, or sorry, 2018, you guys are still re-guiding to 30. Is the 40 units still the right number for 2020? Second, I guess, why doesn't the 2019 number go up by two units? Thanks.
Yeah. To answer your last question, we're introducing by the middle of the year, the 3400 C, which has a productivity which is over 155 wafers per hour, which is a significant improvement in terms of productivity, which means that customers are not planning systems, they're planning wafers. When you get more wafers out of a machine, then you might potentially use less machines. That's why the 30 unit is still good, and it's not more than that. You could argue that the two units are then cannibalized by the higher productivity of the 3400 C. Okay. That's good because the 3400 C is also a higher value tool. We'll just price it higher. From a sales point of view, I think it's a good progression. On the 40 units, I said it earlier, in a previous answer.
I think the 40 units is the capacity that we have. I think whether we will sell it all is really a function of the success with which we're going to introduce the 3400C, and we're able to start running up the availability of the machine over the 90%. That will drive down cost for our customers significantly. Cost is the main driver for our customers to buy tools. I think the opportunity here is in the memory space, in the DRAM space, and also somewhat in the logic space, because then you can add a few more layers to EUV because the cost is just better. In DRAM, like I said, if you're over 2,000 wafers per day, we come into a realm where customers are really seeing the economic benefits of EUV application in DRAM. 40 is the capacity.
Let's go after it by executing on our 3400C program.
All right. Thank you.
The next question comes from Mr. Andrew Gardiner. Please state your company name, followed by your question.
Hi. Good afternoon. It's Andrew from Barclays. Thanks for taking the question. I just had a few more quick ones on the EUV program. Firstly, Peter, you just mentioned it briefly there again, the question of layer count within Logic. If I go back to this time last year, we were talking about 10 layers at the seven nanometer node. To your point, the improved productivity in specs on the 3400C suggests it's going to be higher than that. Can you give us any initial indication from your customers as to how much higher the layer count may be relative to that initial number of 10?
Yeah, I think we hear different numbers. Again, like I said earlier, I assume a successful execution of our 3400C specification targets, and you could look at anywhere between 12 and perhaps 14.
Okay, thanks. Just quick follow-ups on EUV. Can you give us an idea of mix between Logic and Memory in the 30 tool shipments next year? Clearly, again, if I go back a couple of quarters, Logic was going to dominate, but you're sounding a bit more optimistic about DRAM demand. Also perhaps one for Roger. Is there going to be any EUV deferred revenue left to recognize in 2019? Or is the rev rec next year purely on the 30 tool shipments? Thank you.
I think the mix is predominantly logic. Like I said, there is an opportunity there. Of course, throughout 2019, when we see the first module results, the test results of the critical modules of the 3400C, we can probably engage with our customers at that time, the memory customers, to see whether they would like an increased number of memory tools. That is an opportunity, I would say, but it's predominantly logic.
In terms of revenue recognition, as you know, at this stage, the systems revenue gets recognized upon shipment, and that will obviously continue for this model into 2019. The introduction of 3400C at this stage, again, we believe that we will recognize revenue at shipment at this stage.
Okay. Thank you very much.
The next question comes from Mr. Alexander Duval. Please state your company name, followed by your question.
Yes. Hi, Alexander Duval from Goldman Sachs. Many thanks for the question. Just a quick one on logic spending in 2019. You obviously talked about both memory and logic spending remaining on high levels in 2019, but talked about DUV logic actually being up even though revenues are already on a high level. I wondered if you could just talk about what the key swing factors are that are driving that. As a brief follow-up, you talked about a 2H weighted year for your overall group revenues in 2019, and you've just talked about a flattish half-on-half growth rate in the first half. How should we be thinking about the step-up into the second half? What is the key reason for that step-up, and are we talking low single digits growth half-on-half or something of greater magnitude? Many thanks.
Okay. The logic spending in 2019 is up, but don't forget, the majority of the spend this year was in memory. Logic will be ramping 10 nanometer in microprocessors and 7 nanometer in the foundry space. That's happening because when we listen to the customers, the tape outs are there, the customer orders are there. That will happen, and that will increase. Like I said earlier, that's why I think that the DUV business for 2019 will be at least as good as in 2018, but by a little bit more skewed towards logic. It's driven by 7 and 10 nanometer. On the half-on-half, I said in an earlier answer that our view as to 2019 and, let's say, the shipment levels first half, second half, haven't changed that much from where we were one or two quarters ago.
What effectively means is our customer plans, which is a result also of when their fabs are ready, when can they take the tools, that hasn't changed that much. I would say the half-on-half is more a function of when the customers need the tools. When do they ramp what? That's the main reason. There's nothing magical behind it. There's no other reason why there would be this particular sacred cow, if you want to call it this way. I think it's just the way our customers plan. This means that the first half of 2019 will be somewhat the same as the second half of 2018, which was a pretty good half. The acceleration you will see in the second half. Also, the EUV numbers will go up. The 35C will be there. That's all the second half skewed.
That's probably the only answer I can give.
Great. Many thanks.
The next question comes from Mr. Stéphane Houri. Please state your company name, followed by your question.
Yes, hello. This is Stéphane Houri from ODDO BHF. Actually, I have a question about the OpEx lines because we saw really an increase in R&D and SG&A. As you said, it's to ramp your new EUV tool, the pace is accelerating throughout the year. We are now up 26% year-on-year on R&D. Where do we go, and how do we model it for 2019? Thank you.
I think we've said before, also on the Q2 call, I think we've mentioned that we believe in the short term, there will be an uptick in R&D, and that uptick is to a very large extent or is uniquely related to two things. It's the acceleration of the point 33 EUV roadmap, as we mentioned before, the 3400C, and also the High-NA program acceleration. That's why we said midterm, we expect that. We said short term, that will lead to an uptick of the numbers. We also said that medium term, we expect that to go back to the model that you've seen before and that we've given to you for 2020, which is 13%.
Sales.
Yes, sales.
The follow-up is about metrology and inspection. You had a very good quarter this quarter. Is it a trend that we should push forward, or was there anything special this quarter?
No, I think it's just a trend that metrology and inspection will become more important. There's a couple of drivers there. I think the introduction of the YieldStar 375 is a metrology system that is now not only being used in Logic and in DRAM, but now also is introduced into 3D NAND with very clear advantages for our customers. On top of that, we see good growth, very clear growth in HMI, in the EUV business. We're planning to ship the first 3x3 multi-beam tool in 2019. That will also help the top line. There's a whole suite of software products that we're helping our customers to deal with the complexities and the intricacies of seven nanometer, and the five nanometer development nodes.
There's a whole suite of products that are actually helping our customers to basically deal with the increased cost of the next nodes. That's particularly helpful, when you look at our metrology and inspection business. It's a trend.
Okay. Thank you very much.
Next question comes from Mr. Mehdi Hosseini. Please state your company name, followed by your question.
Yes, sir. Thanks for taking my question, Mehdi Hosseini from Susquehanna. Peter, I just want to go back to your comment about 2019, first half of 2019 versus second half. I appreciate the details. It's still the same view as a couple of quarters ago. I'm just wondering, does that reflect the finalized CapEx plans by your key customers? Or if there's a change those CapEx plans later this year or early next year that could have an impact on your view that has not yet materialized.
Mehdi, what you're asking me is, looking at my crystal ball, what customer CapEx plans are going to be going forward. I don't know.
Yes.
If they are going to-
Right.
If they're going to change, they're going to change. There's nothing today that leads us to believe that they're going to do that. Just when you look at their plans, it's about technology transition in logic. Trust me, it's going to happen. If you then look at the DRAM expansion plans, there's a limited number of customers, only three, and two of them have some capacity expansion plans where the fabs are being built. You're long enough into this industry to understand that once you have the DRAM structure there, you're going to fill it up because the only way to cover your fixed cost is to crank out as many DRAM bits as you can in this new fab. These are all plans that are pretty cast in stone. Whether they are going to cut or to slow down that plan, I don't know.
The current plans are what they are, which means that the shipment that we're seeing in H1 and H2, that's been planned for some time now, they're still valid. What changes in future? I don't know.
Sure. Thanks for the sincerity. The fact that your customer mix has increasingly consolidated, does make it more challenging to forecast.
It makes it easier.
Yes. Well, it's easier for you because your crystal ball is better than mine.
That's absolutely true.
Right. One thing with EUV, I'm just very intrigued. We started the year with the commentary that you could ship 22. You see now it's down to 18. I appreciate the improved throughput with the NXE:3400C coming out second half of next year. On the flip side, your customer mix has also consolidated. One of the key foundries is no longer pursuing leading edge, and the leading foundry is now the leading semiconductor manufacturer, and they're well ahead of others. Perhaps the DRAM industry is waiting for NXE:3400C before they finalize their plans. I'm just trying to better understand when we dial in a 30-unit system into our expectation and 40 into 2020, what are the key wafer capacity targets that you're looking at? In the past, you've talked about certain foundry capacity for leading edge.
Is there any metric that you could provide us so that we could have a more realistic set of expectation? If there is a change, we know what are the key parameters that have changed?
Okay. Well, I think, the more realistic expectation is the expectation that I gave you. I think it is realistic. It has to do with the fact that, yes, our customer base is consolidating, which in itself generally leads to a better capital efficiency in the industry. Every customer plans for winning business, and if they have multiple customers all planning for the same business, then that might be a reason, or that might be a very good reason why you ship a few more systems. In this particular case, it's not the case because, like you said, there's one customer has stepped out of 7 nanometers, GlobalFoundries, and that business now goes to Taiwan. The real question is, if that's a 7 plus or a 5-nanometer type business, what is the size of that 7 plus and 5-nanometer business for that customer?
That is significant. I can only repeat what the CEO of that company said a couple of times. 7 nanometer, 7 plus and 5 is going to be big. That's based on what their customers are telling them, what they need in terms of wafer capacity. Since they're the only one really in that space, I'm not going to tell you anything about the planned wafer capacity. You should ask them. That's not my role. I can tell you that this is a big driver for the 2019 EUV shipments. That's only for the foundry business. On top of that, you have your micro processes, and you have the first start of some pilot production on DRAM. If you add it all up, you see those plans and you see the roadmaps, the 30 number we think is a realistic number.
Is there some upside? Hey, if the third version C turns out to be a very good tool and we will figure it out in the course of 2019, even on module testing level, could we output one or two or three more? Potentially. Let the customers decide, and I would not think it's going to be in logic. That upside would probably be in memory and in DRAM. That's the situation today, and we're giving you clear guidance on the 30 units. That's really based on a realistic scenario as presented to us by our customers.
Got it. Thanks for the detail, and I look forward to seeing you in a few weeks.
Certainly.
The next question comes from Mr. Sandeep Deshpande. Please state your company name, followed by your question.
Yeah. Hi, JP Morgan. Thanks for letting me on. Peter, my question is, I'm trying to understand what you've been saying about 2019. Clearly, EUV is up to 30 tools, is what you're guiding. Your metrology business is growing into next year. Your install base management business is growing into next year. I don't think there are questions about that. The question is about the DUV business. From what I'm hearing you're saying in response to earlier questions, that you're looking for a flattish trend, and that is dominated by the growth in logic and memory, not that strong. If you have a flattish trend in DUV next year, I think everybody's estimates on revenue for ASML are wrong.
Well, it's not my responsibility to come up with that estimate. What I said, at least means I see a bottom for our DUV business to be at least the same. Yeah? Now, DUV business has a lead time that is a bit shorter, there are changes from time to time. Customers could still change for the second level of 2019 to go up. I just called a bottom, which actually means that there could be upside. Yes, I would not be surprised if there would be. How big that upside would be, I don't know. Yeah? That is a bit where I have to stay qualitative and cannot be quantitative. Yeah.
Because I'm just looking at the consensus ahead of today. The market is looking at about 7% revenue growth for ASML. Are you suggesting, because we know approximately from the other three line items where your growth would be for 2019. If you have flat DUV, we are looking at well into double-digit growth. Would you say that you could potentially grow well into the double digits into 2019?
I think you've done the math for us.
Okay. Thank you.
The next question comes from Mr. John Pitzer. Please state your company name, followed by your question.
Peter, good afternoon. Thanks for the question. Peter, you mentioned in your prepared comments that China's going to end up being about 20% of business in calendar year 2018. What's domestic China going to be this year? As you look out to your 2019 forecast, is China domestic a breakout year in 2019, or is it more in line with trend line growth?
I think what I said in my prepared comments, that five of those domestic customers are now planning to ramp in 2019, which actually means that we see our business in 2019 from China also growing. Beyond that, I think everything, and I said it also in the prepared comments, how big the growth will be also depends on how successful all those ramps are going to be. The first pilot lines have been installed, and they're actually executing on their 2019 ramp plans. As we all know, as some of these companies are greenfield companies, are they all going to be as successful? We don't know. If they would be, and they're executing on their plans as we currently see it, our business in China will be up next year.
Peter, my second question is just managing through the transition on EUV as you bring out these higher NA, more productive tools. You mentioned in an earlier question that you thought about two tools next year got cannibalized on productivity. Of the 30 EUV tools you have, what's potentially at risk for further cannibalization? Can customers future-proof? Can you upgrade an EUV tool to a higher NA once you've installed it, or is that not an option? To the extent that that 30 number doesn't get cannibalized, should that just upside our 2020 numbers for you guys?
No, I think you cannot upgrade to a higher NA. The higher NA EUV tool is a completely different tool, different dimensions. It doesn't work. You can upgrade from, let's say, a NXE:3350B to a 3400, if you would like to do that. That is a big open heart surgery in the field. That could happen in 2020. Yeah? We see some of those upgrades. I generally would say that there's not much downside to further cannibalization than what we just said. I think there's an upside, if the third version C turns out to be quicker meeting the performance targets, then we could have to lure in one or two or three more systems in 2019 going into 2020, as a start for higher adoption in the DRAM market. Still, too early to speculate any further beyond the 30 units. Would certainly not speculate down.
I wouldn't speculate up yet. If there is a chance for a change, I would say it depends on the performance of the C. Especially opens up the possibility in the DRAM space.
Perfect. Thank you. Very helpful.
The next question comes from Mr. Amit Daryanani. Please state your company name, followed by your question.
Good morning and good afternoon all. Amit Daryanani from Citigroup, thanks for taking my questions. I really wanted to just circle back on a broader topic, Peter, if I could, on just this whole talk about trade wars that's going on right now and potential implications for supply chain. Could you maybe give us a sense of if you have done any assessment whether you are likely to be impacted by the second tranche of tariffs, and if you see any need within your own supply chain to make any changes based on what's already been made public today? I have a follow-up. Thank you.
Yeah. No, we can be pretty short on this. We don't see any significant impact. Not for our business, not for our supply chain either.
Okay. Secondly, in terms of, just without trying to belabor too much on the point for the 30 tools. You said that the capacity would be 30 tools, and then of course it depends on the output for the 3400-C. Would it be fair to assume that in terms of the production output or the demand that your customers are seeing out there right now, it's as strong as it was three months ago? If anything, it has gotten even stronger, which is why you're saying there's more likelihood of numbers being up than down. Would that-
Yeah.
-be a fair assessment to make?
Yeah. First, there's a little correction. Amit said the capacity is indeed around 30 systems, but if you add the tools or you could do 32. What we're saying, we are shipping 30, which includes the two that they are shipping from 2018, because the wafer capacity that customers are needing, because we have a higher productivity tool, they need 30 units, 28 plus two. Yeah. It's driven by the higher wafer capacity output coming out of the 3000-C. That actually means that it could be, from a manufacturing point of view, it's a few, one or two or three upside, but that would only materialize if we get our customers convinced that the productivity of our EUV 3000-C number is good enough and is reliable enough to put them into earlier production for memory, for DRAM in this case.
I said it a couple of times, I hope it's clear now. This is why I said, I don't think from a demand point of view, there is a big change. What we said before is that when we looked at the overall demand for EUV, we did include, for instance, customers like GlobalFoundries, which of course have fallen off. That could have driven the demand over 30 units. They're not there anymore. That is consolidated into one other customer. That's the only thing that probably changed, is the consolidation in the industry. It doesn't have an effect on our shipment plans.
Just if I could very quickly ask, because you talked about the productivity of the 3400-C, have you decided what level of markup in price would you price the 3400-C over the 3400-B? Or is that still to be fixed?
That's still to be fixed. We're talking to a few customers on that final pricing. Let's not do the price negotiation over this conference call. Leave that in the private rooms of the customers.
Peter. Thank you.
The next question comes from Ms. Tammy Qiu. Please state your company name, followed by your question.
Hi, Tammy Qiu from Berenberg. Thanks for taking my question. The first one is, Peter, you mentioned that next year, DUV spending is mainly skewed towards the logic foundry side. I'm just wondering because when logic foundry maker like TSMC move into new generation, their reuse percentage can be as high as 95%. I'm just wondering to what extent you are actually reflecting high reusage in your estimation. At the same time, would you say in your backlog of EUV shipment, has anyone already got full allocation of tools for ramping up next generation, seven nanometer plus equivalent? They are still ordering for that generation? I have a short follow-up.
I think for anything beyond N7, let's say seven nanometer or N7, there's still orders to be taken. That's no easy. Reuse, you have to define reuse. What customers are mentioning when they talk about reuse is that the existing installed base can be reused for the next node. Yeah. The next node needs more capacity. What we're looking at for next year for logic is true capacity additions as extra wafers out. When I said 2019, I didn't say it was mainly skewed towards logic. I said logic is going to increase in terms of its share in the DUV shipment as compared to 2018. Yeah. Memory is still going to be strong, logic is also going to be a higher component of the DUV shipments than it was in 2018.
Okay, thank you. You mentioned last quarter that you are accelerating your R&D process for High-NA. I'm just wondering, has your accelerated R&D been impacting the number of layers EUV can be used by the time most High-NA is available?
Sorry, could you repeat?
Basically last quarter, you have been accelerating your R&D process for High-NA EUV tool, right? I'm just wondering, with your accelerated R&D process for High-NA, has chip makers been making decision about introducing EUV for more layers when High-NA is available? Because it's available shortly.
Yeah. I think what you will see going forward is that High-NA will be introduced in high volume manufacturing, kind of the middle of next decade. You will see a very clear mix of 0.33, you could say low-NA layers and the use of High-NA layers. They're going to be used next to each other. High-NA is not going to cannibalize that much of the low-NA, but they're going to address the additional critical layers of the N3 and N2 node. This is how it actually works. Yes, on EUV, if you think about EUV in total, of course, there will be more layers allocated to EUV in a combination of low-NA and High-NA.
Okay. Thank you.
The next question comes from Mr. Adithya Metuku. Please state your company name, followed by your question.
Good afternoon, guys. It's Adi Metuku from Bank of America. I have two questions. First, the clarification on the OpEx. Obviously your OpEx is ramping up a lot into 4Q. When we look at the run rate, quarterly run rate for 2019, should we assume that the 4Q run rate would be a reasonable number, or do you think that'll start to trickle down as we go through 2019? Secondly, just looking at five nanometer demand and how the ecosystem is developing, I wondered if, Peter, if you could comment a bit on how the ecosystem is progressing, especially from a pellicle and inspection tool viewpoint. Thank you.
I can take the last question, Roger can take the OpEx question. On the five nanometer development, if anything, it's accelerating. I think the issue with pellicle is a function of defectivity. The defectivity numbers and defectivity control is increasing significantly. We have made a lot of progress this year together with our customers on defectivity control. I think on the five nanometer node, the current use or the use of the current pellicles and the defectivity measures are sufficient to support five nanometer. That's what we believe. At three nanometer, which is a couple of years beyond that, we might want to look into whether we need additional inspection tools. That is really depending on how successful we are in defectivity control at the seven plus and the five nanometer node, which looks to be very good.
Whether we need that inspection tool going forward is still a question that needs to be answered. Perhaps it's negative that we don't need it. It depends on the progress that we will make on defectivity with the current generations.
As relates to CapEx, it's in essence the same logic as we had for R&D, which is, over time, particularly when you talk about acceleration, you can see a bit of an uptick, you won't be surprised that CapEx, to a certain extent, correlates with R&D. With R&D going up, there is logic that certain CapEx goes up as well. That will go hand-in-hand, that's something that we see in the short term. Medium-term, long term, you will once again see that it models back to what we presented to you in our 2020 model, which is 4% of CapEx. 4% of sales would be assumed in CapEx.
Sorry, Roger, apologies if I said CapEx. I meant OpEx. When I look at Q4 2018 OpEx as a proportion of revenues, obviously when I take your guidance, you are seeing a pretty strong uptick, and the OpEx run rate is significantly higher than what consensus modeling for 2019. I just wondered, you made a comment earlier on OpEx taking up short term but coming down medium term. As we go through 2019, when should we expect that uptick to come down? When should we expect that downtick?
You can expect that downtick and that going back to the model that we presented to you in the course of 2019, early in 2019. There are reasons for SG&A in particular, because I think we need to distinguish here between SG&A and R&D. I mentioned to you R&D as it relates to SG&A. You see a bit of an uptick in what we expect for Q4. You will see that come down to the 4% model that we guided for in the course of 2019.
On R&D, I think you already answered it.
Yes.
We are driving the High-NA introduction at the frequency, which will mean that we see an elevated level of R&D spending in 2019. Which actually, we give a very clear indication in Q4 of what the levels could be. Medium-term, that will come down again. Where will that be? I think somewhere in the 2020, 2021 timeframe.
Yeah
you will see that, because that's when the peak of the High-NA program will have happened. Because we're planning to ship High-NA starting 2022. End of 2021, beginning of 2022. That peak will be for the next two years, and then it will level off.
Understood. Very clear. Thank you.
All right. Thank you. Before we sign off, we'd like to remind you that we'll be hosting our investor day here at our headquarters in Veldhoven on the afternoon of November 8th. As the event is currently fully booked, we ask those that have not already confirmed to please join us via webcast. We will provide the webcast details in advance of the event. You can contact investor relations with any questions. On behalf of the ASML board of management, I'd like to thank you all for joining us today. Operator, if you could formally conclude the call, I would appreciate it. Thank you.
Of course, sir. Ladies and gentlemen, this concludes the ASML 2018 third quarter financial results conference call. Thank you for participating. You may now disconnect.