ASML Holding N.V. (NASDAQ:ASML)
UBS European Technology Conference Call
March 8, 2012 03:00 am ET
Franki D'Hoore - Director, IR
Gareth Jenkins - UBS
Gareth Jenkins - UBS
Well thanks for joining Day 2 and it’s my great pleasure to introduce Franki from ASML. Most of you probably know Franki already and I just will say that this is the smartest that you'll ever see me. I'm Gareth Jenkins from the European Hardware Analysts and Franki, over to you, thanks for coming.
Thank you and good morning to all. Safe Harbor statement, in the first two points of the presentation, just a quick where are we statement. Business summary, you've seen these with the results. Actually we had the several quarters, last quarter of 2010 and the first quarters of 2011where we were running close to or over €1.5 million.
In Q4, we went to about €1.2 billion and we guided and you will see in the outlook, the first half of this year at €2.4 billion and that means also a runrate that averages €1.2 billion. You could summarize that as revenues did come off 20% if you compare it with the €1.5 billion.
And that is for ASML, but it’s probably also indicative for what our industry does. Margins around these levels are still quite nice in terms of operating margin close to 24% et cetera. Backlog we have shown of course it came down a bit, but we also have guided going forward or just the cash, we returned a lot of cash to shareholders.
So if you, we announced another one for this year that has increased a bit over what was still outstanding and in combination with dividends this year, it will close to €700 million and last year it was about €900 million that we returned and it all results in us having returned over €3 billion in the meantime and we have since 2006 actually the buyback of the shares represents about 30%.
The financial policy of the company is to continue doing these two things, that is dividend and share buybacks. And as you may realize relative to this kind of P&L and cash generation, the breakeven level of the company is around €2 billion. So as a company and in the past, we used to operate in peaks and troughs, around breakeven, we are now so far with that range, above breakeven that it no longer seems to be a big concern if no disasters like 2009 happen, is what you have to add now.
But in the normal cyclicality of semiconductors where big troughs, significant troughs would be 40% in terms of overcapacity correction and where mild troughs would be 20%, it’s not a concern. So this being said what I really wanted and yeah outlook, this is what we said on Q1 and the booking is expected to be a bit above Q4 and then margins are fine, R&D we took down a bit, $5 million and SG&A is about stable.
What I really wanted to talk about today has to do with the fact that there is EUV coming and we’ve started to mention that several quarters ago and now that the EUV is coming, it becomes a bit of, everybody is very focused on EUV as if in the world only EUV exists.
So from an ASML point of view, we perceive it a bit like you know, it is -- the conversation is pretty unbalanced and I think we are eager enough to bring a bit more balance back in the conversation because every time you see somebody and it could be two weeks ago, they ask you how is EUV going, what is the program since last two weeks on the source? Is the power up a bit and et cetera, et cetera? So we think that is not necessarily a very meaningful conversation and I would like to have a bit of a more meaningful conversation.
So reality is that we have a dual product development strategy, that says it is very much focused on an evolution of NXT which is the body that does immersion, the machine that does immersion and then also on EUV, that is the technology for the next 10 to 12, 15 years.
It is very important to have both. On immersion, last year we did a bit over 100 machines. We’ve done a lot of improvements on NXT and we have several NXT machines that are highly, highly productive in the space of customers that do over 4000 wafers a day, that’s a large number. Probably, it doesn’t tell you a lot, but that is a large number. These are machines that do over a million wafers a year.
If you will compare that to numbers that customers would give you on our competitor, they would not get close by any meaningful amount. So NXT is there and it’s there for a long time. EUV is there and it’s there in the sense that we shipped last year, the 3100. It is used with every customer to do recipe development, so it’s not for testing the machine. The machine has been tested. It is for really developing processes and process layers on an EUV ecosystem.
Yes, it is true that a lot of work still needs to be done on productivity for that technology to become economically viable, very true. Now this is a picture that shows you that indicates a bit on high abstraction level, but with a bit of detail, on how the customer roadmaps go in Logic, DRAM and in Flash. And you have the timelines and underneath the timelines, you have the geometries, going down to below 10 nanometers and you also have a bit of an indication on the device architectures that are being investigated and built to get there.
So not only lithography is important to get there, it’s also the device architecture that needs to allow you to go there and needs to be ironed out. So yes there is a roadmap for all of them to, down to in fact 8 nanometers to probably 7 nanometers. Yes there is a dual supplier, a dual technology strategy from ASML which is at one hand, the immersion strategy.
It is immersion, but it is very complex immersion because all the acronyms that you see here are, this is for example a single exposure that is a litho (h) double-patterning sequence. That is a space of double patterning sequence, this is a double exposure sequence. It’s one where you do not do an H in between. So there are several double and triple and double-double exposures and the space of double-patterning scenarios that customers are currently engaged in with immersion and they are certainly for a number of layers of that have to print the critical feature-size down to these levels that can do with immersion.
The bad news is you will see it in one of the following slides, that is at the cost for, the more than one imaging step, also involves a lot of other steps and therefore gets very costly. You will also see in the next slides that EUV to do these feature sizes is a quite simple process. It’s a difficult machine, but it’s a very straightforward process, it’s one exposure and in fact it takes costs down significantly.
So the conversation that if we want to have a balanced conversation, the question is not how many EUV machines are you going to do in 2013, the question is really, is EUV going to be ready in a few years from now to take over several of the critical layers and do them at a lower cost such that it no longer have to be done at immersion and yes, it is also true that at some points, there is a bit of a brick wall that says some layers can no longer be done with immersion.
It doesn’t happen, but it happens in 2015 and 2016, it doesn’t happen in 2013. Now this being said, in fact what you will see is that if the productivity of EUV comes along to 40 wafers per hour or 60 wafers per hour, on the course of this year, if that were to be the case, then the two x nodes would already adopt some EUV layers in the critical layer exposure. It would take a bit more time, then it would happen in the one x node, around 15 nanometer.
Now also it is some people will think that EUV will replace all immersion. Also that is not true because in the mid-critical layers and actually this is a timeline that shows the progression that is currently in the development plan on the immersion and it extends beyond [this]. So there is more detail beyond this, but in terms of wafers per hour, we are at 190. We go to 230 and we go to 250. We go to better overlays, see the uniformity so that the line uniformity, the width of the line uniformity. So there is a lot of, a very significant progress to be made and it is going to be made on all the upgrades of the NXT machine. There is an NXT-B coming, there is an NXT-C coming and it all implements these improvements. In fact the way you should look at the world and that’s why I gave only one example of, it’s a Logic environment. So this is the current 28 nanometer process that all of the foundries are rolling out as we speak, the three major foundries.
This is a 22-20 nanometer process. Now, if you zoom in a bit on what happens; this process has 37 total layers that it images to build up the chip and of those 37, 14 are critical and 23 are non-critical; critical means minimum feature size. Minimum feature size meaning 28 nanometer. So there are a bunch of layers, actually the majority that is not printed at minimum feature size that can be done with these machines and a critical feature size needs to be done with immersion. So 13 are single exposure, only one is a [little actually to etch] layer and that’s a double exposure, that’s a double patterning layer.
When you go to the 20-22 nanometer nodes, now there's an increase to 40 layers total, that’s an increase. Now 19 are critical, minimum feature size and 21 non-critical features, that can be done with these machines and with these. Of the 19 critical now only eight are single exposure, whereas the other was 13. But now we have to do 10 layers with a little [actually to etch] and even one layer is triple exposure. This is triple patterning, plus the 21 non-critical. That in fact results in 30% more exposure than on the former one. This is what happens as we speak for the next 12 to 18 months, you will see as a result of this nearly doubling of immersion tools per node. It has to do with multiple, patterning, process steps that can no longer be done in single exposure.
What it does to cost is this. Because not only do need to expose multiple times, you also need several depositions etch that; that you have to do with other equipment. This is a cost per layer. This is a cost per layer. So the cost per layer is doubling and tripling and in quadruple patterning, that also exists, some people “might use?” It would even get worse. Now with EUV you get back here in terms of cost per layer. And yes. I have to admit this is with a 25, 125 wafers per hour EUV machine. But even if it is only 60, it is already economically viable for all minimum feature sizes.
So what we will be seeing, back to the earlier conversation, in the years to come. That is, you will see, for example here that might be one or two layers that if you get to a stable machine EUV, at 30 or 40 wafers per hour there that may be one layer, that becomes economically beneficial to do with EUV rather than to do it with immersion. So there will be a gradual handover and it might happen at 22, where customers may decide to no longer use immersion machines to do one or two layers, but to use a few EUV machines. Because the overall cost for them is significantly better. This is a roadmap I wanted to show this, before the roadmap. There are layers like; there is a compactful layer at 16 nanometer. In DRAM there is one layer that is also very, very complex to do. Let me first clarify. This is a work that customers have done on the six machines that are on the fields, EUV machines. So the way you should look at EUV is from an imaging and an overlay perspective, the machines are gorgeous.
I wont say they are finished because they have a whole roadmap of improvement but for this specification in terms of imaging and overlay, the little graphic imaging related side of what the EUV needs to do is please give me plenty yesterday. But yes, it is true that to become economically viable, the productivity has to triple. This where we are today. Also to put that a bit in perspective, when you look at the productivity of EUV four years ago, we improved it by a factor of 100.
So there is a road map on NXT that extends well in to 2015 for critical layers but then also, you have to note that this immersion that is now doing critical layers will also gradually, when this becomes EUV, when this becomes 15 and 12 and 10, the mid-critical layers will also become more demanding in feature size and those will, you will need to do those in with immersion.
So immersion is never going to go away because if you look at the single patterning environment, what always happens is when this did not exist these were doing the critical layers. So you always have the new machines are doing the critical layers, the former machines are moving in here and in fact the number of machines could be larger than the critical ones. So immersion is going to be here for ever and EUV is going to be inside it. At some point in time according and it will be related to productivity because productivity is related to the cost per layer, that's the picture.
EUV roadmap; we have several improvements lens wise and it’s improving the NA, the Numerical Aperture. That's an old story, that's been done with 248, that's been done with 193. So this is what you do in lithography, that is you improve the Numerical Aperture of the lens and you get smaller features. Its physical it is in fact the Raleigh formula. Now, you can later on also get to double-patterning and you can get to half the wavelength, that allows you to go to seven nanometer.
What this tells you for non-litho guys is once you have a working and productivity-enough EUV machine it has an evolutionary life for over 10 years. You have a body that is ironed out in terms of stability; reliability; the cost of goods; and improve the margin; the cycle times in the factory, that usually takes two years, by the middle of 2014 this will be done in terms of getting all those things right. And then as of then it is an evolutionary, this is [coherence] improvement. This is of axis elimination. So this is in fact all techniques that we have done with 193 and 248 starting in the mid-90s.
So with EUV, you are starting back, as a point where you have printing lines that are larger than the wave length of the light, which isn’t easy, that’s is the lithography of the early 90s. And then you do in terms of techniques on how you get a smaller feature sizes, you do the same thing over again of what you have done before. So the stability and the productivity of the machine getting that to 30 wafers per hour is what needs to be done and there is nobody who doubts in ASML, and not with customers that will not be done in the next two, three years. In fact we all wanted it to be much sooner for all the obvious reasons and the obvious reason is cost.
So this all done by customers and please I wanted them yesterday. Yes, the machines are done in the field. They are used and they have done and in fact to do the work that I have to do today, they are more than good enough. They don’t need 30 wafers an hour. What they are doing in terms of productivity and it’s a lower number of productivity, yes. But you can do enough wafers a day to do the qualification of layers and to do the assessment and the process development of the chip layers. It is nothing to do anymore with the machine, the chip layers that you have to do.
Now anybody who has ever visited our site will may be recognized this, this is where I live. This is the old 200 millimeter factory. This is a 300 millimeter factory. This is the NXE factory. This is the existing EUV factory, in green. That’s where we are building new EUV machines today. This one is nearly finished. So that is actually, these are offices, this is a factory. This is an additional parking, because our biggest problem is parking lots space. And so, this is all development. These are all the white development buildings. This is development factory. And these are the cranes, building these extension and these are the two first clean rooms that are ready. Already in the extension and in total, we would have 15 if all this is concluded and that would also be enough to address all production until 2015.
So in terms of CapEx, we’re working capital intensive company. CapEx this is finished now, more or less. So that spend will go down and a summary on conclusion of EUV, the imaging and overlay that we have seen, the little related performance is fantastic. The successor of the 3100, which is the 3300, is scheduled to ship this year. You get single exposure, fewer designs, less complexity and overall it is less cost. It is extendable for many, many nodes. And yes us and customers and also the ecosystem around it has made a commitment of making that industrialization happen.. Thanks a lot.
Gareth Jenkins - UBS
Thank you Franki and let’s now open to the…
Well, there was a girl going to bring the microphone in, thank you very much.
Gareth Jenkins - UBS
So I’ll try to open to questions, I mean I can start overall; Just wondered, if you could talk about some of the challenges that you face with your partners in taking EUV from a kind of lab environment in your own, you know [your work] and you’re actually taking it to the real world environment with customers it sounds like it’s been I guess a bit of disappointment in that transfer process of actually getting the tools running in a real world environment. What are some of the challenges are in terms of operating in a vacuum or whether it’s the clean environments or just the cost burden that it takes?
See the six 2’s that we shipped last year they are all in a customer fab environment. They are no longer in labs. They are doing devices and yes the fabs that were used EUV in terms of floor space, in terms of the path where it goes, it’s a more complex fab, it’s a more expensive fab. But there is not in terms of how you have to integrate in the world, there is not, there are no show stoppers any longer, there are in terms of us having the vacuum and everything right in the machine is fine.
Yes, there are still a lot of interactions going between the source and the machine to move productivity up. We have addressed 15 different working packages on different aspects in the productivity to be improved and that is all being worked on today. That integration of all these improvement, all these work packages is going to happen later this year and then we will know that the result is and then we will know whether or not we have a stable machine at 60 wafers per hour or stable machine at 40 wafers per hour or are there still a few of the instabilities that needs to be worked on. So that is from our point of view what we still need to do.
From the point of view ecosystem around EUV, that has to do with the replicas, the masks that you need to image. The metrology to do mask inspection to do overall inspection on the process, the metrology, the wafer tracks, the photo resist, they are all in the shape that is useable. And you will not hear me say that it does not need to be improved, because when you introduce a new wavelength in lithography, as we have, if you look at the first days of 193 which was around 2000, yes the photo resist with hindsight was awkward.
But that is in fact very natural because you then get an evolution into consumables that are using these processes that the manufacturers of those consumables are only incentivized by volume. And that is a very natural phenomenon. But I can guarantee you that we do have customers today where we have a conversation, where we try to make them cautious and where we tell them life is very difficult and they tell us that I hear you, but I ignore.
And to me, this is a strongest way of telling me how much I need EUV. And it’s not because the integration and their fab and the ecosystem that goes with the machine and around the machine will be a problem and it’s not because it cannot -- in fact it can be implemented in production today where the throughput high enough.
Just to ask on that, what do you mean by work packages that need to get the product to….?
It means that if you have a machine that 10 wafers per hour you analyze what aspects in the machine needs improvements and there are physical phenomena that are going on in the machine that are at this moment limiting productivity. Now all these physical phenomena you isolate in small work packages and improve as single aspects or two aspects of improvement because you cannot do them all together. So you have to chop it in pieces; you chop it into 15 pieces; and one might be optical and one might have to do with getting rid of the eliminating the infrared, the heat in the system and one may have to do with debris mitigation and one may have to do with power on resource. So there are many, many aspects on how you can improve productivity, you chop them into pieces, you have 15 major sub-packages that need to be worked on then there are projects that manage these 15; one of each.
As were you discussing before the presentation, I saw a couple of your customers recently and they were all complaining about increasing capital intensity and it all sort of boils down to -- boils down to lithography throughput. And I mean you talked about the plan to get immersion throughput from 190 to 230 to 250; what's the mechanism of improving the throughput on the existing immersion technology? They are just more volts?
Without going into well, number one, first I think I am a bit challenged by the statement that capital intensity, true; yes, capital intensity and in lithography is important, but if you look at a cost per bit improvement that we have been giving the world over the last 10 years and 20 years, we did from an economic point of view stay Moore’s Law. So you look at a cost of 1 gigabyte NANDs in 2000, it did cost $2,300. Last year it did cost $1.
So yes it is true that the customers always complain about increasing prices, but the price model that we are using to price our machines and in fact we price them higher always, so we did with market share with the machine that cost more than Nikon’s; it’s a bit odd to believe, but its true and its what we did and it has to do with the productivity. So it has to do with you deliver the same imaging capability at a much better productivity. So otherwise it would never have chosen our machines.
So the fact that it all came our way, it has to do with the fact that we are the best supplier you can find in terms of cost per bit improvement with any generation. And it’s highly, highly profitable for them. I mean, look at our giant customers. They will show you gross margins of 60%. We have 45%. We are not abusing anybody.
Gareth Jenkins - UBS
I guess, what you have achieved in the past is in dispute. It’s sort of concerned about keeping on that roadmap in the future?
Yeah definitely and if you plot EUV and the productivity on EUV, it stays on that roadmap and if you have to do immersion double-triple patterning in the meantime it’s a challenge. And it’s a big challenge. And the roadmap of NXT and the improvements on NXT getting to better overlays and to higher productivities is taking away most of that pain. But I will not deny that if you will have to stay on a continuing immersion double-triple patterning that that is ultimately for Moore’s Law not sustainable; you’ll have to get an EUV. That is definitely true.
Now what is also very true is that sometimes customers -- see the world is always like this. Initially they want EUV soon and everybody wanted EUV soon and then the whole world realizes that EUV cannot be as soon as you wanted it, because there are – it’s really hard; there are really several very difficult fundamental things that need to be resolved.
And then they look at alternatives; and then they think wow, fantastic. Double exposure and double patterning is a great solution and then everybody washes EUV off the table and says, EUV double patterning is my future. And there is not to do double patterning and there is not to do with a bit more of an industrial environment and then they find out how complex our patterning is and then they find out how ugly all these things are and in fact, they’ve done a bit to themselves as well, because if we would have known sooner, we could have anticipated sooner with productivity roadmaps. But I think for the time being, it’s pretty well in balance, but customers always want more.
If I could ask a follow-up question, also on specifically on NAND, talking to some NAND producers and you know obviously there is this problem with endurance on NAND reducing as you move to nodes. And there was some commentary around, perhaps they can do one or two more nodes and then they have to go to 3D and I was just wondering, what were the implication for ASML be going 3D rather than going down another node, because I guess what they were saying is…..
Yeah, the ability to go down to those levels; the endurance on NAND just is unusual?
Well, its not even to do with the endurance. It has to with the fact that today the architecture they’re using as a floating gate and a floating gate means it’s a polysilicon gate on off sides. You inject electrons in it and when you are shrinking further, the amount of electrons that you inject into that gate is becoming so low in number -- you’re talking about 40 to 60 electrons. Now this gets to a level where you can no longer distinguish a one from a zero with enough statistical evidence. So yes the major reason why they have to look into non-floating gate architecture, has to do with the fact that, that sooner or later comes to an end. Yes there are several alternatives investigated at this moment and as we have said before, what it does to us is when it goes to a 3D structure, it is three times more machines. It’s not three times more advanced machines, but it’s three times the number of immersion machines.
If it goes to, not a 3D but a 2D structure, so there are several 2D structures that are also under investigation at this moment. If it goes there, it does more of the critical layers at a lower feature size, every modeling that we have done there gives you in fact a similar revenue number. So it’s either three times more machines of the former generation or it is what we have seen in the past, a continuation of what we have seen in the past, but then with higher ASPs of a critical machine. Then that for us is, it doesn’t matter.
Is it still economic, that three times the machines (inaudible)?
We think it is economic if you are resolving the right architecture. The answer to your question has been the same, you could have asked the same 10 years ago and 20 years ago and all of us would be standing here and not knowing the answer. It is the same challenges what we had ten years ago. There is no more, no less.
(inaudible) I guess with EUV taking some time, people naturally start to talk about alternatives again and we have seen this various times in the last 10, 15 years, but one that does seem to be getting a bit more traction, I think I mentioned this last year is direct self assembling on the chemical side, I just wonder whether internally you are working on 450mm directed self assembly or any of the other kind of may be non-traditional approaches to the problem?
See people talk a lot about DSA, and this is especially because AZ Electronics talks a lot about it. If you look at what it does in reality today, so one of our partners IMEC is of course doing a lot of work on DSA. There was also, that published at SPIE on it, it is at this moment far from an industrial solution and major issues with DSA is that it still has a very, very high level of defects.
Many people think of DSA as it is could be a total replacement of lithography, but nobody does know or seems to realize or looked into enough detail to understand that you need a pattern to use DSA. So you cannot, it does not replace lithography. You first need a lithography step to then within one set of lines to multiple, to do a multiple of that page.
They also don’t realize, it works in one dimension. So it’s highly restricted. So if it ever were to be defect free, it would still be very, very restricted because it is only one dimensional. There is not one chip that is one dimensional. There are, if you define the first layers of a NAND, of a flash device today, that is a one-dimensional pattern.
In the events where all the process and defect issues would be resolved, it might potentially be used at one critical layer, of the many critical layers potentially.
Gareth Jenkins - UBS
I guess, the question is just, are you exploring some of these things internally for the 450 mm day size some of the kind of alternatives. I know, you historically looked at adjacent markets, I was just wondering whether you’re looking at some of these alternatives, potential complements that may serve as more as a complement than a substitute.
The DSA is a compliment. We are looking at alternative markets. Yes, it’s too early to come to a conclusion there. The main reason is that the end markets and end users are a bit hesitating but yes we are doing all of the above. We are still convinced because we put a lot of work in it in the past.
If you for example, look at electron beam as an alternative, that has many more fundamental show stoppers than EUV to ever get to a production were economically viable. You can do feature sizes that are small enough and overlays that are good enough if you have the right machine, but to get to productivity levels that makes EUV an economical solution is several orders of magnitude more complex than EUV.
In the past you see there have also been people looking at nano imprints, which is a stamp that can reproduce these features. The defect problem that that technology has is also a major, major showstopper to ever come to an industrial solution.
So at this moment frankly spoken we put all our effort in immersion and all of our effort in EUV to sustain this roadmap and to do all the necessary things that I would towards bringing to the table, because that is what is really important for customers to happen and at this moment no further dilution of what we do is needed.
Gareth Jenkins - UBS
Does it have, what kind of impact?
Gareth Jenkins - UBS
Negative impact on the cost of the machine, on the pricing of the immersion machine.
Gareth Jenkins - UBS
With EUV or…?
No. The impact of the pricing on immersion is that when you come up with better immersion machines they become more expensive, because they create more value. We are actually easing. At this moment the pain that customers have to make double patterning and triple patterning processes work, the big difficulty there is managing yields.
Now next to the machine, we have now a whole set of holistic lithography solutions that is helping the customer managing that yield and that is where the pain is. It’s a yield pain because you have to have enough good wafers and good dyes on the wafer that work and that is a challenge, but if you provide that value and if you have the customer getting there sooner and to ramp up to yield levels that he did target and you help them get there sooner and you can, then you create a little value. And that’s a long answer. The short answer is it does not anything to prices, but what it did in the past that it’s prices get better.
The industry has decided that they are willing and able to swallow a higher cost of layer in double patterning and triple patterning and implementing it today because, see, NAND today is running for, I think for three years on space of double patterning. It’s implemented in NAND, it’s implemented in Logic, it’s implemented on a few layers of DRAM; it is being used. It’s there. And the reason why it is there has to do with the fact that the cost of bit improvement that they get is so compelling that, yes that’s why they are managing a much more complex situation.
If they cannot price wise offer an attractive device to their customers who are selling iPads etcetera, if that were not to be the case and those iPads will not be able to sell at 400-ish dollars or Euros or pounds per piece, it wouldn’t happen. So the whole economic chain has to be balanced and it’s the case.
Franki, just one question with the NXE: 3300B sort I’ll be quick, did you accelerate that on the roadmap given the issues with EUV….?
I missed your first….
The NXE: 3300B tool, has that come forward on your technology roadmap because of EUV and because customers are needing you to address double patterning and triple patterning and also is that the tool that will get you to the 250 wafers per hour in 2013 that you referred to in your presentation?
NXE; the NXE B and C have always been on the roadmap. They have been on the roadmap ever since we understood the complexity, yield complexity that customers have in double and triple patterning. They have always been on the roadmap; to get to that productivity level we have to introduce again when you say, it’s there, it’s nice, it’s a significant improvement over the first stage of NXE and so that is one of the major improvements and it delivers the capability and of a better overlay and of a higher productivity.
And double patterning especially needs better overlay as well, because when you double pattern all of a sudden overlay becomes part of the control of the dimension equation. You have to make a drawing to understand this, but normally it’s only the wavelength of the lights and the characteristics of your optical path that determine how well you can control what you print. In double patterning the positioning accuracy is even involved in that equation. So overlay, if you can't get overlay down to one nanometer you cannot support customers in the next two years. There has always been a roadmap.
Gareth Jenkins - UBS
To wrap up there Franki’s around break, grab for him a coffee or something in the break.
I have two days compressed in one and during lunch grab me if you can, if you have more questions.
Gareth Jenkins - UBS
Thanks very much Franki.