Everyone has coffee, water, cookies? Because the industry can't offer you more. Welcome to today's session, analyst presentation. We hold that every year. Many of you have attended many times. Slide, please, Cor. I don't know who's pushing the button. Ruud, it's your job. That's a nice one.
You can self-serve.
Agenda for today, strategic overview, end markets and drivers. Ruud will update you on the big picture, also more focused picture, what does it mean for Besi? That's the product development strategy, huh? I'll update you on the strategic initiatives, our capital allocation, acquisition strategy. The last point where you all came for is outlook and summary. Anyway, let's start with the strategic overview. All the way here. 2019, those of you who haven't woken up yet, it's become very clear that it is a transition year with an anticipated turn somewhere either end of this year, next year. Depends on many things. We'll tell you more what are the drivers for that next upturn. In those years of downturn, our focus is always on increased development spending for next generation systems.
It won't surprise you that the investment is increasing simply because the specifications of the end products are ever more complex, smaller designs, which require more accurate systems, more expensive components, and more uncertainty. Because also our customers never know exactly what the final design of their end products will be. That's a very interesting time. Exploring new opportunities to exploit our technology leadership, huh? New opportunities means different processes, which are also complex. Number 3, a hobby of Besi is always cost. Cost, cost. How can we build these machines at lower cost? Enormous focus on the supply chain, but that starts with designer. Why do we have so many components in machines? Why can't we build that more simple? All focused at reducing cost. For the next upturn, hopefully getting lower costs with higher margins.
Disciplined capital allocation to maximize return on investment. All of you know, number one is our shareholders. Whatever we earn on top of a safety level, which is about 20% of our revenue in net cash, we distribute to shareholders, either dividends or share buyback, both. We'll spend a little bit of time on that. Some of you who don't know yet, we are focused on profitability and not in the first on market share, because in technology land, every next cycle is depending on the system you have to offer. Market share won't help you. Profitability will help you. Our business model versus our peers. We've also added some more background simply for you to help you in explaining what is different and some elements you may like better or not like. We're happy to hear your comments.
Of course, we're well-positioned for the next upturn. We've showed that, in the past, I would say decades. We ramp faster than our peers. We also correct faster and better than our peers. Anyways, some background why and how we can further improve that. There are certainly favorable long-term drivers in place to generate strong advanced packaging growth in the future. Remember, all of the major, the winners in this technology environment, whether it's Apple or Intel or Samsung or TSMC, they all say in very clear words that assembly packaging, advanced packaging, has become more and more critical for their end product success. That has been already very advantageous to us, but we expect that to be even more in the future. Next slide. Some perspective. You all know this slide back from 2006, huh?
We show here a clear trend, first in revenue, always peak years, and then absorption years, usually three, sometimes four, sometimes two. That's based on new technology rounds implemented in end products in end markets. Also very nice to see, we've done a bit better than the average, and also our profitability, here explained by gross margin, has increased significantly. That trend is certainly what we are working hard on to continue. The right button finally. Who stands spending that money? Just to give you, again, a clear view, who are the potential customers and where are we, and who is doing more or less? One of the big disappointments this year is clearly the top one, Samsung, who spent EUR 21 billion in 2018 and is forecast to spend only EUR 15 billion, nearly half in 2019. Intel is a bit better. Good for us.
Slightly more. Hopefully, that stays. Hynix. All the memory guys are falling off a cliff, but that's not unusual. We see TSMC, but we can go further down the ranking. In total, and certainly won't be right, VLSI has a picture, but that picture today is certainly not the same picture how it will be at the end of the year. Forecasting now -16% could very well be -20%, -25%. It certainly won't be -10%, statistically. On the right side, IDM and foundry CapEx, also self-explanatory, those numbers. It took VLSI some time, and the whole industry I think, to realize that we are in a cyclical industry. A year ago, I think same time, May, June, we saw the first clear signs of a peak of a cycle reached and basically started to prepare for this next downturn.
Many didn't understand why, because then the opinion was shared there is no more cyclicality in this industry. Please accept and remember forever that this is a cyclical industry, and it gives great opportunities. Logic and cloud application are one of the few bright spots, take Intel, for example, partly TSMC. That is in some way helping us. Again, now the right button here. Here you see the cyclical behavior, assembly equipment spending, total 2017, 2018, EUR 4.4, EUR 4.5, then reducing by -17% to EUR 3.7, which is a forecast today, which could certainly be worse. If you look historically, -17%, -16%, in 2015 we had the same picture. Basically over that same period, you see in the dark green, has done certainly better in upswings, sometimes also in downturns.
As I mentioned before, this is a nice picture, but key is to focus on profitability and not on market share. Current environment in some bullet points. Downturn of four quarters continues post extended super cycle. Super was eight quarters, so four quarters down is still, let's say, early. Q1 2019 results indicated some stabilization, maybe some seasonal patterns were included. Certainly, the U.S.-China debate currently has a clear impact. People have become hesitant because historically, the semiconductor industry has been highly sensitive to any trade or economic GDP issues. The end effect is that our customers stand on the brakes. GDP growth still favorable, although excess chip inventory persists. What does this sentence mean? We still have a positive GDP, however we look at it. If that also changes, you could certainly see an extension of the correction we currently have in our industry.
Looking forward, timing of the rebound uncertain. Is it a U? Probably not. The best one is a V shape and a quarterly correction. Some people a year ago were of that opinion. Then a U. Today they are more canoe. Canoe can stretch over more than four quarters. Anyway, however you want to call it, a big question mark. Via CSI anticipates recovery 2021. Excellent. 5G could become a game changer. We're very well prepared for that. Mobile makers advance 3D and facial recognition. A lot is happening there. Data mining, analytical, artificial intelligence needs keep expanding. Those are in a summary, what we are reading every day in every type of message, whether that is directly from customers or analysts. Tariff battle is reshaping semi supply chains.
What certainly is happening because of this current situation is that you see some customers hedging their bets and investing less in capacity in China, starting to invest in capacity outside of China. Many of the U.S. customers in particular. If you look at Besi, our exposure to China is very limited compared to our peers who have revenues of over 50% out of China. That's not the case with us. Of that EUR 24, most of that, certainly 80% of that is outside Chinese customers. Customers like in the automotive, in the mobile space, who organize production, assembly of parts in China. Both foreign and IDMs and Chinese subcontractors, limited exposure to Huawei. Far, revenue impact you could say is neutral. There's some softness on the U.S. side, offset by increased demand on the Chinese side, who want to be more independent from outside supply.
Still the whole picture is an uncertain picture. You see that here. Our peak revenue is quarter two 2017, EUR 170 million. Q1 2019 was less than half of that, EUR 80 million. Q2 guidance, slightly better. Margins creeping slowly further up. Actually, this slide is amazing if you see that. No impact or hardly any impact on gross margin related to our revenue. That's unique. It also gives us the ramping capability. Remember, ramping Q1 2017 to Q2. If you go further back, Q1 2016 to Q2 2016. If the tide turns, we can accelerate very fast, and the tide will turn at some point. Gross margins here compared to our competitors and peers, just for you as a reference. Besi variance Q1 2016, Q1 2019, 6.7% better. ASMPT less. K&S also slightly better. Operating margin, us 6%, 5.9% better, and the others minus.
Anyway, just to give you a reference. You're always telling us our peers and competitors are doing better. Anyway, these are public numbers. OpEx, very important. You can see from the peak in Q1 2017, Q2, Q1 2018, we have managed to bring that back very nicely. Also by taking out cost all the time. The basic OpEx over the past three years has roughly stayed around EUR 26.5 million. That means significant operating leverage, which we all hope and are sure will come. VLSI finally came with some market share numbers just yesterday before this meeting. We told them, otherwise we have to postpone this meeting. Again, market share is interesting, but you have to understand that all the data they collect is partly public and the rest is based on some data collected by sources which you cannot verify.
Still in the end, our market share in 2018 still somewhere 32.4% in the addressable market. Die attach has declined a little bit. Packaging has increased significantly. Plating has increased significantly. Besi hasn't done too bad. Anyway, these are numbers. R&D spending. As I mentioned in the very beginning, there's a clear trend up, not because we like to spend more money, but simply because the complexity is increasing and our share of the pie at our key customers is gradually growing, which is wonderful, but also has a big risk whether we succeed in those developments. Still, clearly, die attach 2017, 24.3%, 2019 expected 28%. Packaging, plating, all a bit more, adding to EUR 40 million expected. We have, of course, our capitalization and amortization, and that leads to an end number.
Important is that every cent we spend is focused on identified and customer orders underlying these development programs. Our model, mentioned that also last year. Clearly, in the next up cycle, we are aiming to reach a higher level of revenue than the last cycle. Based on the success proven, but also many of the new R&D programs on the way, and also successfully on the way, that should give us the opportunity to have a larger share of the pie. With margins, similar level, what we currently have. Net margins also demonstrated. Headcount split 70/30 today, 80/20 in the future. That is our model in the next up cycle. A few words about the end markets. This becomes too complicated. Ruud, please help me.
Yeah. First of all, I think I'm going to move also forward because otherwise you sit so blue after the table. Let me go a little bit over where we're going. This picture I had last year also, basically our digital society. This has not changed with the phone market, internet, all the data centers, automotive, medical environment, all of that requires more and more, basically, semiconductor content. If we look at what are our key markets, first of all, we have the mobile world. This one became bigger. Automotive, computer, PCs, industrial, others, and also spares and services already a good part of our business. Steady growth in automotive. Increased electronic content. Computing, of course. Mobile still a very important part.
If we look at all the doom saying, we discussed this with Richard mid last year, saying super cycle, everybody telling us, "The world will continue. It will be different." We said, "Nah, not sure." I was sitting in Seoul Airport, beside me, I heard people talking about semiconductors. My ears, of course, they become big. It was one of the founders of Raspberry Pi. This is a little computer out of England, a very nice little thing, and the head of Samsung U.K. I started talking with them and how they receive it. The Samsung said -- I said, and this was around this time of the year last year. He said, since three weeks, they see all the orders going down, the orders cancellations, pricing going down, and they had expected it already earlier. It didn't come.
They said, "Now it's there." Then I said, "Richard, now we have to be careful." When we said that, everybody said, "You're too pessimistic." We think we have a good feel for that, yeah. Despite all this doom and gloom at the moment, if you look at, this is from Ericsson, the data over the mobile networks. It's still growing with 30%, 40%, 50% per quarter. It's a real high growth numbers. Per quarter, 70%, but on a yearly base, enormous. This data traffic, this is growing. The fundamentals of our industry are not shattered at all. They are clearly there. It's also funny if you see, this is mobile networks, voice is just this here, and it's all data. If you analyze the data more precisely, 60% is video.
With increased video content, not only more movies but more whatever we have, but also with 4K, 8K coming up, the data is tremendously increasing in the networks. 5G should help us in that, and 5G is starting to come now. If you see this, these are the expected subscriptions in 5G. You see it starts here about 2020. This is roughly the same curve as you had here, if you go back here, as the 4G network started. We are in the beginning phase of 5G, and that will give us a very long time, positive momentum, I think. Yeah. Where is this happening? North America is a big growth market for 5G. Western Europe to some extent. Northeast Asia, that includes in also China, is also growing very strongly. Here are the growth areas for 5G.
This is for all of you who want to read more background on this. Ericsson always have these mobility reports, the next one comes out just a couple of days after this meeting, so I had to stand with the older one. It's always very interesting to go through that. There's a lot of data in that for understanding this, yeah. What's also clear is that 5G rollout seems to be faster than the 4G rollout. If you take the comparable timeframe, 4G, there were four operators and three OEMs. At the same time, already 20 have announced and 20 OEMs have announced. There's a lot of momentum in this. The other part, you have seen this before in 5G presentations, 5G is much more than only a faster telephone.
It will be around mobile, of course, it will be industrial, it will be automotive, it will be, of course, in the computing area, smart cities, retail. It will spread much deeper than the 4G networks. What is expected in this network is, first of all, with all this data going, about 1,000 to 2,000 times traffic capacity. More connections, better usage of the spectrum, decreased latency. That means that the reaction time between the network and your phone or an IoT device or a car will become much faster. Of course, yeah, more data. The question is, how? It's all nice that you say, "Oh, I need to bring more data over this network," how do you do that? Also a 4G network is already a very good network.
The whole area here is if you go to higher frequencies, one gigahertz, three gigahertz, 24 gigahertz, you can have work with a higher bandwidth. That means the data channels, you can put much more data in the data channels. You see here that in a 24, 28 gigahertz network, you can have much higher peak rates, 10 to 20 gigabytes per second. That is simply because you have more room, you have more space available. That is what you see happening. This is what is a real new part in the 5G network. You get higher data transfer. It means also you have to have new transmission technology in there and more advanced networks. What you will also see, you will get a lot of small cells closely by.
If you take an area like Amsterdam, here, this area, you need a lot of data everywhere. Outside Amsterdam, it may be a little bit slower. You get a lot of new hubs, so to speak. If I go one picture further, this is current technology in 4G. You have a base station, that radiates around and you pick up data, it sends data up and down. In 5G, this is in the millimeter wave, something new, this is for some people a little scary. You will get a new type of base station, this will actually direct a beam to where you need the data. In this beam, you can send over much more data than in a normal broad range. These are active antennas, active phased antennas, as it's called.
These base stations that you see everywhere, if you open a base station like this, you will see an enormous amount of small antennas. These are like 24 by 24 or 10 by 10. By phasing these antennas, that is why it's called active phasing, if you send the pulse a little bit later than here, you can drive this beam in a certain direction. You can steer this beam. This beam will follow, in this case, the little train here, or it will follow maybe a car. That's it. The disadvantage of this technology, it's a nice one here from Google, is that the wave propagation, the distance you can cover with this technology is much shorter. Where you have in 4G, in a, say, sub-6 gigahertz, you have 1.5, 2 kilometers here, or maybe 0.5 kilometer or less.
means you need much more of these base stations, you'll see much more small cells. Will this be everywhere? I think this will be mainly in the bigger agglomerations, where you really need a lot of data. The second step then, of course, is if you send all this data and you have this beam, it's all very nice. The question is, how do you receive on your mobile phone this beam? How does that work? What you see is you need new antennas for that. 5G, IoT, 1 gigahertz, 6 gigahertz can be done with the classical antenna packages. That is still really doable. You will see little or no change in how you do the packaging. Here, a little bit change, but here you need what is called Antenna in Package. You have to have the antennas very close to it.
You will see multiple antennas, new modems to handle all this data. We are actively involved in these things. If you look at this here, this is an example from Qualcomm. This is the first new, I would say, antenna module. Of course, there's a modem connected also to it. You also need, per telephone, more of these antenna systems. You will have two or three or four because these millimeter waves, if you have your phone in your hand like this, or you have it like this, it will work differently. With the older antennas, they are much more random, much less sensitive there. Yes, you need new technology and you need more of it. This is an interesting area, we are clearly involved.
A lot of our customers in this field, like the Qorvo, the Skyworks, the Avagos, the Murata, they're all working on this integrating this new technology, getting these antennas to the right format. In the mobile space, it is not only 5G. There's a lot of other things happening also. You see people start to come with foldable phones. If you think about this, if you fold the phone one time, two times, if you have to do this many times, you have connectors in there, these connectors have to be very, how shall I say it? Bendable. We have some projects running in plating where we actually make these connectors with a specific plating technology. Another thing you see is more and more cameras.
I think this is one where it has about five cameras in there, which is basically good for us because we do a lot of camera mounting also. Richard already mentioned face recognition. We are very heavily involved in this, especially in this 3D face recognition, where you have a VCSEL sending out basically very small laser beams and then reconstructing the picture. What you also see is new fingerprint technology, that you have a fingerprint sensor somewhere here, that you don't have a specific spot anymore. You can just push on the screen. This world is ongoing. This is more what we do with our classical machines, actually. Another thing you see in the and I call this also mobile space, is Who of you has an Apple Watch here? Okay, at least now, yeah, five points this time. Yeah. You also. At least two.
Gentlemen, buy all of a watch because you help me. Yeah. What you also see now, you get more sensors again in the fourth generation. If you look in the internals, it's unbelievable what is all cramped in there. We are involved, of course, in this type of technology. Also here, you can get an ECG already out of your phone. I'm not sure if it's good or bad because you, "Oh, shit. I'm dying." Not a good idea here. Sorry, this is a little too much maybe. Another part is, you see also in the and I call this also mobile space. You have these new Googles where you can look through, and you get, at the same time, a projection of what you're doing. This is initially just started with gaming, but it's becoming a very big market for education and also for instruction.
We have done for ourselves, even a few cases where we have a service engineer with his glasses on, and you can see what you do, and you have a picture of what you have to do. Works very nicely. If you open one of these things, there's an enormous amount of electronics in it. Not only electronics to get data in and out, but also maybe a MicroLED display to get a very good display. You have a lot of technology in these things also. This mobile world, for us, is still developing very well. Yes, the number of smartphones at this moment is a little bit slowing down, but we see more than enough opportunity to continue in here. That is a little bit the big picture for the mobile world.
If we go to the automotive world, this I found special because both BMW and Mercedes Daimler joining together in a new mobility service provider instead of buying to consuming. Nevertheless, you still will need a lot of cars. I like this idea that you can take a car everywhere. I'm a big Mercedes fan, so I thought, "Hey, with BMW, this is Let's see." There's a lot happening. In this car industry, the number of vehicles, 94 million. This is my normal car. Oh, this is an Audi. It's not good. 105. Yeah. The growth is not very big in, say, units in the automotive industry, but the content with electrification, you see a big growth.
Automation. I think on the last, maybe 2 meetings ago, I had a picture of a Mercedes with this big red button that you could switch out the automatic, it works very well. Connectivity, of course, is growing, comfort and more things in the car also clearly there. Here is still very good growth. Long term, we clearly see this. The number of packages in cars is unbelievable. This is from Amkor Technology, just an example. You have wafer level fan-out in the radar part. You may have in the info systems, you may have SiP. Here in the electronics, in the lighting. You still have a lot of wirebond packages because of anything with power typically has or wirebond or clip bonding, but you have a lot of that in there also. QFN is really coming in.
Car is for us a very good thing. We see this. That's also why our share in automotive was bigger than last year. It was the other way around. You see a lot of technology fan-out coming in, wafer level chip scale packaging coming in, new packaging in here for the sensors. Power, copper wire going to copper clip, for better cooling. We also see flip chip QFN coming in. In the automotive industry, you always have to take care. Testing is enormous because you have to guarantee this typically for 10, 20 years. Before this thing, it's a slower process, but once you are in, it is running very long. Another part we see in not only automotive, but generally in the power, is you start to see new materials coming in. Silicon carbide, gallium nitride.
These materials, they are a little bit a different semiconductor material, you can use higher switching frequencies. That means a lot of DC/AC, DC/DC converters. They are based on switching technology, and the higher speed you can switch, the more efficient these modules become. Also, silicon carbide, for example, has a lower loss when it's on. That means when the current flows through, there's less resistance in the device, and that gives a higher efficiency, and they can work at higher temperatures. A normal silicon module, at the junction, you may come to temperatures of about 100, 120 degrees. In a silicon carbide, you may go to 150, 180 degrees, but that still means you're bonding to the surface area, where if you bond it on copper, that bonding has to be also able to carry these higher temperatures.
You see new bonding technology, like diffusion and compression coming in there. One of the things here also, you can work with a much smaller form factor that is a factor 2 or 3 better. Our classical area, where everything started with the computing market. That market is really exploding because it is artificial intelligence, high-power computing, analytics, especially genomics, for example, the networks, the memory security. Enormous increase again in what is needed in computing. It's data and computing, both long-term are in a very positive trend. This was from Intel in May, their investor meeting. This high-end, this computing is everywhere now. In the past, it was maybe a high-end computer in a data network, but it's in gaming now, it's in medical. There's some enormous fine pictures you can get. Autonomous machines need a lot of intelligence to function.
Automotive. Of course, in the mobile space, these are very advanced processors if you really look in them. This high-end computing is basically penetrating the almost all devices now. The question is, it has always been based on Moore's Law, I had something on Moore's Law in, we skipped it because we said that we'll break the time. The question is always, will this shrinking, will it go on? ASML has been the master in this field, they have been able, with so-called single patterning, with shorter wavelengths, to always work in this direction and get the cost per pixel down. They stopped around here around 2008, 2010, so-called, there was no way to bring the wavelength any more shorter, multi-patterning started.
Instead of doing one mask, you can make a pattern four, five, eight times, you still can go to smaller dimensions, it's very expensive. This blocked the industry somewhat. Now with the new EUV, this trend cost down should start again, you can work again with one or two masks only. Basically, ASML has also the next generation. Here you went, at a certain moment, you went to so-called wet systems, where you have a higher numerical aperture. They're already working on the next generation for this also. This will be installed in imec. My personal belief, having been in the industry now since 1984, that's why I like this plot so much. Richard, you were also in 1984, correct?
We wrote it. We wrote this. imec has this ITF conference always in Antwerp. It's a nice conference, Maarten van den Brink got his lifetime award for his work at ASML, after that, I said, "Very nice, and congratulations." I said, "Please remember that who sold your first machine." That were we, it was not even a PAS. It was not even a PAS 2500. It was a PAS 2000. Maarten van den Brink, he remembered. He remembered. "Yeah, that was there." Anyway, the message here is this will continue, I think this will continue certainly till 2030. On the other side, it will not only be shrinking. We will need much more complex structures. You need to get data and the computing much closer to each other. In the past, you would have it here. You will stack this together.
You will need things for cooling. What you also will see is that to get enough data from A to B. I had a nice picture showing a very congested road, three lanes, fully congested. Five lanes, this road is very nice, yeah. We have this between Amsterdam and Utrecht, you see this effect to some extent. In that way, you can get more data to the system. That also means you have more contacts. Every lane is a contact. What you see is to do that, you need to get to smaller pitches between your contacts, so to speak. You see that today, 40 micrometers is what is mostly used in, say, the higher-end area, but you see this going to 20 micrometers, maybe five micrometers, and even below.
If you go below five micrometers, maybe in the one micrometer area, you need also a new technology probably, which is called direct bonding, where now you do it with solder. There are solder material in between the bonds here. You will have a direct bond, where you just put two copper stacks on top of each other, and they will bond directly. Yeah. These are some examples where you have the memory connected with a bridge here, solder here, and you need more and more lanes in here. What is also important, that is something that Intel recently published, is if you want to do this building block ideas like here, where you have all these nice devices all coupled. You may even split this up, if you do that, you need a standard interface between these things.
Otherwise, you can never mix and match. Intel recently published a standard interface for a kind of mismatch of devices. That was an important step because there was no interface definition so far that was universal. That did Intel well. Of course, at the end of the day, if these dimensions become smaller, we have to more accurately place these devices. Here you see these pitches. If this is 40 micrometers, I think two or three years ago, I explained this once. If you go 40 micrometers, it means you need an accuracy of about four micrometers. If you now go back to this picture where you start to see pitches of maybe three micrometers and one-micrometer pitches here. If you take one-tenth of that, you need to come into the accuracies of about 300 nanometers for all types of machines.
Of course, this all has to be molded in a good way. We are also involved in that. Another plot, I took this also from ASML, because ASML always puts the focus on here, process technology, 30%. If you look in the top picture, 3D stacking, multi-chip architecture, memory integration, that's already a bigger portion than this year. My comment simply is, this is the Bailey world, and we are very well-positioned for that. You see this really changing, that this part becomes a more important part of solving our next-generation problems. All the big players are working on this in different ways with interposers, CoWoS, Fanout. Intel introduced this recently, stacked memories, like we've shown this before. Sometimes embedding in substrates. Here we see something. Substrates. People try to get rid of substrates because it's, again, cost, also a bandwidth issue.
TSMC working hard on it. In all of these projects, if I look at it, in almost all of them, we are clearly involved. Which will be the winner? I think there will be a couple of them. Company A will use this one, company B will use this one, but in all these projects we are. What does that mean at the end of the day for our machines? You have all seen this picture before. We have three key areas. One is die attach, where we have to place the die very precisely. In packaging, once the die is placed, we have to mold it, and then if it is a classical lead frame, we have to bend the feed with trim and form, and we have to singulate it, and we have to plate it. These are our three main areas.
We have shown this picture many, many times. Richard explained this is about 80%, 15%, 5%, roughly in numbers. If we start with die attach, in general, we're working on next generation TCB, nano accurate direct die bonding systems, high speed, high accuracy flip chip, large panel bonding, and we are working on a lot of enhancements of our existing portfolio, because very often it is not revolution, it's evolution. Step by step we move forward. One of the things I mentioned is the TCB world. At this moment, in about 40 micrometers, 20 micrometers, we think this will stay till about 10 micrometers. There is still a lot of life in this, and we are working on a next generation TCB system. We see two main issues. One is higher accuracy. We already have a very high-speed machine for TCB.
The other thing is, that I didn't mention, all these complex devices also get bigger. We also need bigger bonding heads where we can do bigger dies, and that's what we're doing in our next generation. Some examples of work we've been doing. This is die-to-wafer stacking, where 16 on top of each other with very good results. The bond head we have, and this is confirmed by more people now, is the most precise available. We expect in 2019, 2020 that this will be a next round of TCB for us. Next step, this is the most advanced. I mentioned that as you go to five micrometers and smaller spacing with maybe bumps of about one micrometer, we have to go at this moment to about 200 nanometer accuracy.
There are tools in the market that can do about 500 nanometer accuracy, but they're all very slow. We are on our way to develop a much faster machine, a machine that can do 2,000 to 3,000 dies per hour, which is a factor 10 better than the current technology. What we do, and this is something we patented, is we have on the substrate, there is a fiducial. We have a fiducial on the die here that we have to place. We have a second window here. We measure this distance between that, and then we can look at this fiducial and this fiducial at the same time. This is the most technical, complex thing I have in this presentation for you guys.
When we then go down, we can continuously track, are we still compared to this fiducial and that, are we still on the same spot, on the right spot? What you do is you go then slowly down with piezo elements to actively steer this to the position. Compared to all our competitor systems, they have a measurement system in between here. They measure this position, that position, then that is a system that is about this big. You pull this out, then you go down, but then you do not know what you are doing anymore. We can actively steer the die to the right spot. We are reaching about 300 nanometer accuracy at this moment. We do this work in our facility, near Innsbruck in Radfeld. What we see now is that facility is close by a highway.
Actually, the highway is just behind the building. We can see the trucks driving on the highway when we do displacement. If there is a lorry passing by, we see it in the vibrations in the system. We do most of the measurements now on the weekends, but we will probably build a better room for that. Yeah, you are getting a level of accuracies that, say, 10 years ago, you could not even think about. At 200 nanometer, be aware that ASML is in the 5-nanometer area, and we are getting very close to the same issue. That means also not only you have to build a very precise machine, also elements like particles, like how you have to build these machines in clean rooms, and they are also going to be located in a clean room environment. It is the next level for Besi.
Will this be massive volume immediately? This will go step by step because only the most advanced companies work on these technologies. So far, results very promising. Yeah. Another part is flip chip and fan-out is still growing. One of the questions we got there from a lot of our customers is, "Can you make a faster machine?" Our machine is already the fastest in the world. We are typically at 8,000 to 9,000 dies per hour. I showed you a movie last year on this presentation where we showed a 4-head machine with 4 individual bonding heads moving here. That machine is now starting to be released in the field with results up to 20,000 we are getting. Factor 2 better than current technology. First evaluations positive can be used both in the fan-out world and also in the normal flip chip world.
We also see a good application for this in the memory space, where we will replace some of the wire bond now with flip chip. Another element we are working on is a big panel machine. I think I also showed you one sheet of this last year, then we were in the development phase. We can now proudly say that the first commercial shipments are happening. Actually, this month we will ship the first one. This machine can handle 720 by 650. There are 2 elements in why you do this. One is you want to go to lower cost by simply being able to put much more dies. This is the big picture. There is a second thing, is also we focus very much on very high accuracy.
That means for the high-end products, but also for MicroLED applications, where you have to go to a very, very high accuracy. This machine has been very well received. At this moment, what we see from the industry, we are roughly a factor 2 better in accuracy at a good speed. Well, these machines are also in a completely different price range than a normal die bonder. A normal die bonder is between EUR 200,000 to EUR 400,000, EUR 500,000. The TCB is a little bit more. These machines, they go in the EUR 1.5 million to EUR 3 million range. Yeah. Let me see here. Next is, these were some of the most advanced parts, but also our bread and butter machines, especially like the Esec multi-module system. We have more than 6,000 of these machines now installed.
This is one of the workhorses of the industry, it's used for things like 3D face recognition, camera, six-axis alignment, high-end chips, high-end and large interposers, high-end LED power sensors. A very broad application range. In the telephone business, for example, people have a very short range. Every year, you want something new, that means these machines can be adapted also very quickly. You can do almost anything you can do on these machines. What we did over the last year, better cameras in there, a new motion control system, operating system. We are now reaching 3-micrometer accuracy with this system. That is very interesting for photonics, for example, and high-end products. We still can use same tooling on there. Yeah, this is one of our highlights. We simply produce a lot of these machines every year.
Almost every mobile phone in the world somewhere is connected to this machine. Another machine that is a Oh, something dropped out here. This is our epoxy machine. The Esec 2100 is basically the fastest epoxy machine in the world at this moment. General ICs, automotive sensors, stack memory. More than 3,000 of these machines installed. New things are dual dispensing, a better bond line thickness. What you start to see if you mount a sensor, you have to mount it very, very precisely, we can measure this with a 3D measurement, where we can do a fillet height control. We developed a better and even more precise bond head, active jet compensation. We are also starting to add self-learning and artificial intelligence in these machines. For some reason my Oh, there they are. Yeah. This is fillet control, where we measure with a 3D camera.
We measure to the outside whether this is correct. There's also something in the industry, it's called poka-yoke, it's not a game. Basically, it means that your machine has to be so smart that it can work with very dumb operators. It's more and more difficult to get operators, the machine has to learn self, analyze itself, we start to see more of this mistake-proof states here. That's an element, that means more computing power also in the machine here. We're receiving orders for these machines. Another area, we do not speak much about this, we have over 1,000 soft solder machines. The big companies in these fields, whether it's the Infineons, the STs, they are basically 100% relying on these machines. This is for automotive. We are adding every year some new features.
This plasma cleaning came in our dual writing expansion. High-pressure bond test, because for diffusion bonding or the silver sintering, you need more force to make this process. We've added that. Improved the illumination. This is an example, and this is just recently published. If you look to one of the big ones that I just mentioned, this is an embedded die here that we put into a substrate with a diffusion bonding process. The net effect of this is better heat conduction here, for 48 volt and for automotive power. This machine, yeah, has been the standard in the industry for soft soldering, and we keep ahead of competition here. We go to molding. Our molding lines are. We have about 300 of these installed, the LM series, for the large areas.
It's a standard, a thin die molding, exposed die molding, mobile, advanced SiP application, automotive memory. What we are introducing this year is more in double-sided molding with an exposed die on one step and also 120 ton molding capacity, because we start to see that the substrates and the lead frames become bigger and bigger. If you have this bigger, you simply need more force. We have now 90 tons, and we're going to 120 tons. The first 120 ton will have to ship, I think, in about a month from now. If you look at this exposed die mold, you see this a lot with sensors, with fan-out technology, where you have to clean certain areas, have them clean, have them in contact with the outside world.
This is for an antenna device, 5G, but I cannot say who this, but you could combine this. We have about 90% of this, what is called this exposed die market. Everything where you have sensors, everything where you have these antennas now coming in. We do this with a very precise control of the clamping. I have another picture later. This machine has been highly regarded in the industry. Some examples where you see you get devices on this side, you get devices on that side, that means you have to double-side mold this. We do that. This isn't even essentially what it is here. We are involved in these steps here. On wafer molding. We have about 10 of these systems installed now. That's the beginning.
One of the key advantages of our approach, which is the no die swim. We can fill very good and narrow gaps. Wafer placement into the mold is very precise, and we can do this very planar. Better I show this here. In our case, we have what is called transfer molding. Our molding material comes in from the side. You have to imagine you have your dies clamped between two plates, then we bring in the molding material. This molding material, it's almost like water when you start with it. You see here in the first 30 seconds, you flow in this whole area, then you increase the pressure to do the final molding. Because we come in here with very low forces here, it almost flows in, then you do the pressurizing.
The pressure is everywhere the same around the dies. That means the dies don't move. If you do this with compression molding, the molding material has to come from the top here. Sorry. It has to come in down. It's much more difficult. There you always have what is called die swim. We are able to do this work on, first of all, on the whole area. We have very straight lines here. That means the dies are staying very well in where they should be. Also we have absolutely no flash. We have no covers on these covers. We think we have a. This is in the beginning phase. This is for RDL layers, for example. We're starting building up a very good position.
This is something that, with the next upturn, should bring us much more machines. Some of the last areas, Besi trim and form. Richard had some elements on, say, supply chain, on modularity. This is a good example of a machine that we brought to a level where it's fully modular. You have a lot of elements you can couple to it. It's a building blocks. Here it's very often used in the automotive. We can configure very fast. We introduced full tracking and tracing in this machine. Nowadays, in an automotive device, it is not Six Sigma anymore. It is basically zero failure. When we see a mistake in, say, a strip in here, we punch this separate element out, and what comes out is 100% good. This has had a real high interest, sorting and offloading.
We had to improve the vision system for that and add selective punching to remove bad dies. These systems are installed in about 1,000 by now. Very popular system in the automotive industry. Last part, Besi plating systems. We never talk much about it. We have more than 800 of these massive lines in the world, and about 80% of the market for lead frames is plated with Besi plating lines. This is a solar line. We have a bunch in solar also. What is key advantage of this machine is the very high plating uniformity. You have to imagine if you have a device and you have to plate it, you always have an element of uniformity, but especially of the minimum thickness.
If you have a more uniform layer, you can normally go to a better or a lower minimum thickness. That means immediately saving in material. The second element is the way we transport the material in this machine leads to a very low chemical waste. It's a very clean machine. You have very little material that gets lost. The only disadvantage for us, it's an extremely robust machine. These machines live 20, 30 years. I've already said we should develop a bacteria that eats these machines because they should have a shorter lifetime, then we can sell more of them. Newish wettable flank, flexible connector, foldable phones are already mentioned, chemical deflash. In this flexible, this is a machine we delivered a couple of years ago. We developed this machine for RF identification.
You have these little antenna strips that you have on a product, and you can do RF identification. This machine had such a high capacity that with two or three machines, you could cover all the RF need. Now we're seeing high interest for the flexible connectors, for connectors you have here. You have in this foldable phone, it has to go out, and that's the new area. Second element is better pre-cleaning. We're developing a machine that does a much better pre-clean. The third element is in automotive. This is a QFN, and this is your copper contact bonded to the strip. You get solder in here, but you always have an open area here, and this can lead to corrosion and to problems in this area. We're now developing a method where we can fully do it this way.
That is very highly regarded by a number of key automotive players. Be aware, Europe is still a very good automotive country. There's about 50% of all the car industry is still Europe. We get a lot of interest for this also. I think that was the last sheet I have. In summary, all these assembly steps, process steps, they become more and more important everywhere. 5G is for us important because we see development at this moment everywhere, and this will lead in the next upswing. It will lead to substantial orders. Continuous improvements in the computing world with the 3D part. Advanced packaging, we've said this many times, will become more and more. I really like this ASML plot because that's the best advertisement for us. We're working with the leading customers on this.
Fan out, TCB, high speed, high accuracy, die bonding, camera sensor system development, exposed die, trim and form systems, ultra-fine lead frame plating. That's about it for today in a snapshot what we are doing.
Excellent, Ruud. One more thing to add. After all these development programs. Rule of thumb, 50% of the spending, as we showed in the slide, EUR 40 million this year, 50% is focused on extending current platforms. Ruud showed many applications being developed, whether it's in die attach or in packaging or in plating. Sustaining, we call that, and roughly 50% is spent on brand-new process development. The key message is you can't live on one or on the other. If you spend your life only on sustaining, you die. If you spend your life only on the new world, which is highly unpredictable, you will also die. Anyway, we don't die. Next, strategic initiatives. As always, maintain best-in-class technical leadership. Ruud showed you a summary of where we are working on, but the key is, of course, pick the winners, as we always say that.
Which customers, which end applications do we expect to be the winners in the next round? If we do our homework right, that also can lead to best-in-class leadership. We are highly dependent on our key customers for that. As I mentioned in the earlier part of the presentation, get more share of the pie. The only objective of spending this EUR 40 million is to get the highest return on that investment. You get that by being smarter, having better machines, allowing higher margins, but at the same time, get more share out of that generation. Increase scalability and cost reduction. I will spend a few more words in the next slide. That's why they are equal size.
If technology can be beautiful, but if you can't build the machines and you can't build them better than your competitors, then it's all nice, but it leads to nothing. Grow via timely acquisitions. Although we haven't done any in the past 10 years, we still maintain a key focus on are there any additional processes available which can fit into our portfolio, which supports our key strategy in achieving higher returns. Revenue initiatives, increased market presence and share. As I already said, where are those things? Again, pick the winners. Who are going to be the most successful in the next technology round? Many we know for certain who have those leadership positions, but some we don't know. We also mention here, expand our presence and share of wallet in China. Certainly, China will be the largest growth market for the next 10 years.
A population of 1.4 billion with an enormous momentum in development. You can't neglect that, and there are certainly huge opportunities. The key is, of course, you have to be very careful to watch out those areas which are not available to us. Our support in Asia, only one line. We have since many years our support center out of Singapore. Those of you who visit Singapore are most welcome to see that. We have application labs for all of our products now, supporting the ever-growing installed base, and that is of key importance to have a larger share of that market going forward. Grow in excess of the addressable market. Yeah, we have done that. That's nice. The key is, of course, to demonstrate that going forward. How do we achieve that?
By, number one, picking the right development programs at the winners, and at the same time constantly improve our operating model. Now you see those revenue opportunities, one area -- well, we go over them one by one. Like Ruud explained in significant detail, 5G. If you simply look Besi's position in that, supplying not only precise placement equipment, but also packaging, plating for that matter, we are heavily involved in that already, and that can offer us significant more growth than we have had in the past. You see here the expand of share in the Korean, Chinese Android mobile customers. That also remains as it is. Increased share of wallet in memory. Ruud mentioned the flip chip opportunity in next generation memory devices. We've not only been qualified, but we've also received machines already, orders for that. Expand the Japanese automotive supply chain.
Automotive key to Besi. You've seen last year more than 20% of revenue as opposed to in the previous years, about the mid-teens. That's growing. Also Japanese, Rohm and Renesas for many years. The Chinese semiconductor infrastructure, JCET, TFME, Huatian, key customers of us. Then, of course, our complete portfolio in the Chinese market. We have built over the past four or five years, step-by-step, certain systems used by customers in China, manufactured in our own facility in Lucerne, and over time, we are organizing to offer the complete portfolio locally. Then Ruud mentioned it already, renewed customer interest. TCB is going to be hot. Wafer level is growing, slowly but surely. Panel, as you've also seen. Anyway, the drivers for that, digital society, big data, smart everything, and cloud expansion. Spares.
Anyone who has any doubt about the future, maybe this year will be uncertain, but until 2030, growth. Anyway, we always see those beautiful hockey sticks. Importance, spares and service. Growing gradually, now around 15% of revenue, and imagine EUR 80 million business. It's a lot. 10 years ago, that was maybe EUR 10 million, EUR 12 million. Why is that the case? Not because our systems are less good, less performing. Those of you who analyze our P&L in detail have certainly observed that our warranty has decreased year-over-year. Systems are in much better performance. That's also bottom line, et cetera. Why then growth in this area, spares and service? It's because customers want 24 hours production out of those machines. By organizing that as a separate business, we have been able to support our customers ever better year-on-year.
Beautiful margins and also in downturns, a support because spares and service are always required. Cost initiatives. To say that again, first of all, our customers need to know, are we able to grow beyond our last peak? We have certainly prepared for that. In any case, our customers should not be worried if the next round comes, whether Besi is able to supply. Then, ever better supply chain, European, Asian, outsourcing higher value systems and modules, and reduce the cycle times. Gain further supply chain and materials efficiencies. Although our margins look beautiful, we are convinced we can still improve significantly. If you look at that very carefully, we still are unique system builders. We are not like the automotive industry. Under pressure and in downturn, that always helps. We can still achieve significant improvements.
Optimize production between Malaysia and China to reduce costs and better local market needs. What does that mean? Those who deliver the fastest get the best orders, because our customers are never certain. Lead time, delivery time is key. Not only delivery time, but also when are the machines ready for production. There is the biggest opportunity to further increase our market shares. Then a bold statement again. We are certainly EUR 15 million per year off a perfect organization, and you can bet your life that in the next years we will achieve that. 40% is in the supply chain. Still, many companies supplying to us, we have over 1,000 suppliers, which can be improved. Headcount cost. We are still moving certain activities from west to east, more and more in the support. Then last but not least, common platforms, common modules.
If you look at the nice slide Ruud presented of all of our machines, 18, maybe, they all are different. They have hardly any lookalikes. In the next generation of systems development, more and more shall be common. Like electronic cabinets, tracks, name any part which is similar on neighbor machines. That is simply copying the intelligence from the automotive world. They are highly commonized in all their models, and that is one of the key goals for Besi in the years to come. That also has a significant impact on our supply chain. Simply imagine all those variations. You need different suppliers, different inventories, and whatever you can commonize is a significant advantage. Working capital, to name another one. We review that every week. Every Friday is supply chain day. Every Monday is production planning meeting, 11 o'clock. Supply chain, Friday, 10 o'clock.
If you don't focus on that 52 weeks a year, the progress is very minimal. Extremely disciplined, and that's how we're going to get there. Our model has shown to be extremely supportive in managing these cycles. In the different colors, the bottom of this, you see in dark Besi green, you see the European U.S. headcount. You see in the light green, the Asian headcount, and the gray on top is a flexible layer. You may have noticed that since the mid of last year, we've reduced our headcount by 450 people. No major restructurings, simply stopping certain contracts in the production environment, contracts for a month or three months. Over the quarters, we've reduced that part. Moving more from Europe to Asia in the support side, in the admin side.
Singapore is now a full-fledged limited risk distributor, LRD structure. All the orders come in, all the invoices go out, the money comes in. We've moved that out of Europe to Asia. The whole supply chain sourcing is done in Asia. Back office. Support, as I mentioned already, the sustaining on the installed base is more and more organized out of Singapore. Several key people have moved from Europe with their families to Singapore to further bring our key technology to our customers in Asia. The headcount development is key in that, but not just restructuring, but where do you organize your business and your support, and maintaining your flexibility in managing through the cycles? Next. Development actions, redesign common modules, as I already mentioned, increase standardization, lower unit cost designs and maintenance hours, shorter cycle times. Few words about capital allocation.
Simply to reconfirm how we see that, I already mentioned it at the beginning, all the profit we make belongs to shareholders above a certain baseline, and the baseline is an industry-adopted net cash to revenue ratio. Simply, between 40% and 100% of our profits can be distributed to shareholders, either in form of dividends and also share buyback programs continuously ongoing. Currently, EUR 12 million per quarter, EUR 48 per year. That is to be continued. If you compare that to the peers, slightly more, I would say. I don't know why, but those are simply data. Share repurchase program, I already mentioned the average, the number, just for your models. Then the consolidating in this world. If you look at Besi, we've done certainly our consolidation work since the start, 1st of October 1993. First packaging, plating, then die attach since 2000 until 2009.
If you look at the same time, what have our peers, competitors done? Oh, they've bought much more. Why doesn't Besi buy? Two messages, they've gone more broadly, diversification. We have said all along, because of all the complexity, Ruud's wonderful presentation, our strategy has been to focus on the ever-increasing complexity, which we will continue. You don't solve that by acquisitions. At the same time, and we see that here, that translates into capital allocation over the last six years. The return on average equity. I think investors are always interested in returns. Maybe some are more interested in acquisitions, but this is simply data of the public company information. That's what's happened. We will continue our focus, but at the same time keep a key eye open where we can increase our return on average equity.
That's our key job. For that you need money. Money we have. We raised some more with two convertibles, one end of 2016, one end of 2017. About EUR 290 million net out of those convertibles available for the next still five years. That also gives us ample means to go through any of the cycles. Outlook. When is it, Cor? July 23 or 24, we will tell you the outcome of the second quarter. Guided up slightly with revenue, some seasonal effect maybe. Gross margins flat, a bit disappointing, should be higher, Cor. No, just joking. Gross margins should be similar, which I think is an enormous achievement for the organization. OpEx less. Cor, that could be more, -5%. We should be able to reduce some more costs. That's about it.
With these beautiful bullets, we'd like to end the presentation. Are there any questions? Yes, please.
If market share is-
One second.
If market share is an output of your business strategy, why are you guiding to a 40% objective? What gives you the confidence that you can grow it that much and maintain the margins that you've been talking about?
Well, if we look at the past decade, our overall market share was low-mid 20s. By simply maintaining this very dogmatic focus on die placement, die manipulation, I sometimes call it. You can sort, you can place, you can flip. Very focused on those key customers, pick your winners. Ruud explained in, I think, significant detail, sufficient detail, that that complexity is only increasing. Even more so, it is increasing to a level where our competitors today, the distance is greater than it was 10 years ago. Our cost and supply chain initiatives. The proven track record of our operational model and the operations organization in Asia should give us that opportunity. Of course, we're in a tight, competitive landscape. Don't underestimate the competition in the different areas.
Still, we think it's very reasonable to assume that at least our customers will offer us a chance to perform. Whether we succeed in that is still to be seen. One of the things also mentioned by Ruud quickly, this complexity leads to increased ASPs. Already TCB is practically double a machine price compared to a flip chip machine. The output, however, is significantly lower, so you even need more TCB machines to achieve the same output. One level higher, you are moving to, and we quickly said that, machines which are between EUR 2 million and EUR 3 million a piece. Where you will need, again, more to reach the same volume than compared to previous generations. Those things also add up. That market grows in number of EUR faster, although the volume is growing slower. It's not all equal is my message in comparison. Does that help? Excellent.
Next question.
Thanks. On the EUR 800 million objective or ambition. First, the previous peak, the EUR 600 million. What are the big moving parts there? I can assume that maybe that mobile is a little bit more comparable than it was in the previous peak, then the big growth coming from automotive and computer. How should I see that?
Yeah. Again, take the slides Ruud presented. On all three fronts, whether it's communication, mobile internet devices, or computer space or automotive, they all have significant changes which translate into more complex machines. It's not just the same, let's say, machines, then a next round three, four years later. It's more complex machines which are more expensive and certainly the volumes will grow in the mobile internet device space, computing space, maybe a bit less in automotive, but the content in automotive. That's behind why if we do our homework right, we could achieve a higher next peak.
Then on the gross margin mission. You mentioned also more outsourcing on more higher value parts. How does that compare to your gross margin? You don't have to give away part of your gross margin or just in the mix of the total you can win more on the other end by having the right machines ready, or how should I see that?
First of all, again to explain how is our outsourcing model organized. The unique thing about Besi operating model is that we have for many of our modules, multiple suppliers. We don't have single source, maybe for a few where we haven't found a second one, but for every module developed at the development stage, we organize multiple suppliers, simply to be independent of any of those suppliers. At the same time, to, on an ongoing basis, improve the cost structure. What's the cost structure? Not only the cost, but the delivery time and the number of modules. The scalability, which is equally key. Still, I mentioned quickly, we have 1,000 suppliers plus, and the key is to find in that whole range of suppliers for these 18 different platforms, more cost savings through better selection of those suppliers.
We evaluate all of those suppliers all the time because still compare this to another industry which is far more mature, the automotive industry, our model is very expensive in that sense. There's a lot to be gained. Any gain in that supply chain, combined with the design of more advanced or lower cost of ownership machines improves your gross margin.
Yeah.
If there would not be any improvement feasible, our gross margins would not increase and maybe at best stay because of development success. That's the whole battle.
I think maybe to add a little bit to that for the flavor, Richard mentioned it. We really have every Friday a call with high-level management. Richard, myself, Cor is in there, but also the product group managers, supply chain manager, where we go over specific areas and every week something else. Every week we want to know what can we do more here? What can we do more? That's the kind of discipline. In the beginning, it's very, oh, after a while you see this brings us a lot because every time we analyze, hey, can we get five? Why is this guy more expensive? We do this at a level where a lot of, I will say, other types of managers say, "Ah, I have my people do that. I'm not interested." We are interested also in these small things.
We have on Friday, one element is the supply chain part. We also have our accounts receivable meeting on Friday with Cor. If we have an account receivable CHF 500 that is not paid last two years, we go after it. Yeah. Even at our level, we say, "Hey, why is this not paid?" Because it can be an indication that there is something wrong with the company. It can be an indication we do something wrong in our machine, and we want to know that also. That's a very, yeah, I would say we are very practical on a high level, also busy with the details of running the business.
For instance
Yeah
recently we did all the travel agencies.
Yeah.
It's amazing how much cost we simply save by changing the travel agencies.
Yeah.
We simply want to see that. Why? Why? We change and somebody said, "Oh, but they do less," da. Just to give you sufficient-
Yeah
background about why we think there's still enormous improvements to be achieved.
Yeah.
You have to do something for that. You have to work hard at that. That doesn't simply happen by an objective.
Yeah.
Please, next question. This table first, closer.
Hi there, Wim from ABN. You mentioned on the one hand that you still have about EUR 15 million-EUR 20 million in cost savings that you identified, earmarked, whatever. At the same time, you also mentioned on multiple occasions that the world's becoming more complex, platforms become more complex.
Yeah.
Obviously you're also targeting a significantly higher revenue base. What kind of cost base do you think you need in order to cater to the EUR 800 million ambition?
Well, we mentioned already that the infrastructure is in place, that's no worry. The supply chain is also in place by our whole structure. Simply remember how you guys have all gone to school. My first school was next door here, the Montessori. I was born here across the street. Anyway, four times 170, how much is that? 680, yeah? It's not a big stretch to EUR 800 million with a mix of machines which will have higher ASPs. We don't have to buy market share, which we will never do. In perspective, the only question which is very important, will that be achievable with our product range in this industry? Are we with the winners? Do we have the right development programs? Again, sustainability of existing plus the new technology developments. That's the biggest risk factor.
Are we at the right party or are we in a wrong party? With all the uncertainty you mentioned, that's highly critical. The only way we can manage that, and we haven't said that yet, the R&D programs, every two weeks we review the whole list. Every delta, a change at a customer and his timing, and often that is not just because of end markets, but because of materials, because of other steps in the whole process. That has a significant impact. We have a limited number of people, and you have to constantly review whether we have the people on the right programs. We've added people in the R&D programs simply because we anticipate that there will be more. In some occasions, we've said in the last round we had three major new avenues. Today we have seven new avenues.
Whether they all will be big avenues or very small lanes, you don't know.
Yeah.
That also adds up to potentially a higher revenue level in the next peak. Is that sufficient? Excellent. You had a question? Then we have this table, all three.
Yes. We'll keep it here for the time being. Mike Roeg, Degroof Petercam. I have a question on your R&D, which you mentioned is becoming increasingly complex and expensive.
Yeah.
About half of it is spent on the existing products and half on the new systems, which have all these fancy ASPs. My question is on the existing products, EUR 200,000-EUR 500,000 per unit. They have to go even faster than they already are, even smaller dies, even more complex. Do you see the ASPs there growing faster than, say, in the past 10 years, simply because of that new complexity? That will give you probably also a push from the previous peak of EUR 600 to where you want to be.
That's a good question. The ASP of many of those machines has not grown. We have added certain functionality which customers are willing to pay for. We have significantly reduced the cost by improving the supply chain and our own operations in the number of hours we need to build those machines. The revenue growth has not come from those machines. The revenue growth has always come from next generation.
Basically you've given all the improvements, well, for free to the customer? They get the operational improvements without an extra price to pay.
Look at the margins.
Yeah, that's because you have been very efficient. That's not something they have to do.
No. The margins tell you everything. You would never achieve this growth, this constant increase of margins if it is one or the other. It's a combination. The big jumps, that's what I wanted to say, are in the new systems. In the existing range, they are more gradual.
Okay.
The customer is only willing to pay more if he gets more functionality and lower cost of ownership on existing platforms.
Yeah, there have been improvements. Suppose that you have 20% higher output with an improved version of-
Yeah
your existing machine.
Yeah.
The client still pays the same ASP.
No. He pays. You explain, well, cost of ownership is how they evaluate us against the competition. You have to demonstrate clearly on existing platforms, even more than on new things. Is it nice PowerPoint presentations or are these real demonstrated data? That's how you sell those machines. Whether the price goes up by 20 or by 15 or by 30, we sell it based on cost of ownership.
Yeah.
The cost of ownership improvements translate in ever growing gross margin, even in down cycles.
Yeah.
What we haven't said, if a customer doesn't want that, if there's a short business horizon, maybe a year or a customer program, we are very happy that he buys a lower cost machine. We're not interested to maintain just that market share because. Very important question to understand about our strategy.
Yeah.
Many may do it differently. They may say, "Well, we can get much more out of this market if we have a lower margin.
Yeah, I think it's also very simple. If a machine is, say, 20% better and you basically increase the price also with 20%, then the customer will say, "What do I get for this?" It's not a big reason then to go. You have to find this mix, because we have been improving the machines all the time. Yes, we have been improving step by step also pricing you. One of the elements you see is many of our machines have different names. Simply if you go to an Asian buyer and you say, "Oh, I have an EVO 2200 and it's now 10% better and I have to charge you 10% more or 8%" or whatever. They go, "It's the same machine. Why should I pay more?" You have to make a new version number of it.
You have to do a lot of things to keep that. There's a lot of pressure on that. We have been able to do that very well. Otherwise, we would not be able to keep these margins at that level. We've always had this performance, speed, accuracy. Speed, accuracy, always one step so, it has been working very well for us.
Yeah.
A quick one.
No.
Well-
A quick follow-up. Yeah. Basically, the idea I wanted to share is.
Yeah
if you boost it by 20%, you share it with the client, so 10% each.
Yeah.
Both companies win.
No, we don't share.
Yeah. You share the operational benefit. That is part of the way to get from EUR 600-EUR 800, simply because your ASPs are probably going to be going up if you spend all that money on your R&D.
Yeah, my message is that growth is less coming from the existing product ranges. It is more coming from the.
Yeah
New product ranges.
Yeah. A big dream.
gross margin before revenue growth. Okay. Next question.
Yes.
Mark.
How difficult is it to prepare?
Oh, I think you were online. You were waiting very long time there.
No problem.
Oh, okay.
How difficult is it to prepare the organization for, on the one hand, the cyclical slowdown risk, as you mentioned, and on the other hand, the secular growth opportunity in AI and IoT and big data and gaming, crypto mining, autonomous vehicles, and probably I forget a few. Especially if they happen simultaneously at the same time. How difficult is it then to prepare on the one hand, for the upside, whereas at the same time you also want to, well, keep a lid on the cost for maybe the cyclical downside that you encounter as well?
It's an excellent question. We said at some point the past days, we can write a book about that. The answer is, if you manage the business on a weekly basis and you see negative or positive changes, you immediately adjust your organization. I mentioned quickly, we have 450 people lesser. Those times when we waited for that we were really in the downturn, and you had a massive restructuring, everything stops. It's all communication as well. If you take people along that on the one hand, we have less machines, less business, on the other hand, we have increased development requirements. I can only say the more we do that in line with what is happening, the better the organization understands. Also, we have split the organization. R&D is here in Holland, packaging, plating, in Austria and Switzerland, Diatech, and we have operations in Asia.
We have the support in Asia, operations in Malaysia, China, support center in Singapore. Many would say, "Well, all in one factory is much lower cost." Also this focus has brought us a lot, and at the same time has brought us that you can manage much more easily the cycles both up and down. In the past, when it was in one factory, any decline would have a very negative impact on the total. In an upturn, you would pull all the people out of R&D to support the operations. Today, that's not the case. Anyway, it's very fundamental. Our customers do the same. If you look, for instance, at Intel or Infineon or STMicroelectronics, everyone has learned how you manage the cyclical business is by splitting those parts of the organization, which might seem very ineffective to other businesses. Next.
At the end of Q1, you talked about 1,695 employees. You said that you are meeting a lot weekly with management. Can I have an idea how big the management is? If you talk about how many people?
Easy. Nine people.
Sounds like.
Our management team is only nine people, and then four.
Salespeople?
Are close to sales. We have sales in Asia. We have sales in the U.S. Asia in total, sales and service together, is EUR 190.
100?
90. In Europe and U.S., it's about EUR 60
62. EUR 60-something.
62?
62.
In R&D, of what amount of people do we have to think? Is that gradually going up?
Yeah, it's gradually going up. We just added another 27 in this year so far. Well, if you look at die attach packaging and plating-
About four
Singapore. First of all, My last count was we were at 380 people in R&D.
one other small question. When I met you a few times, last year or the year before, you were talking about new applications with growth potential in solar business.
Yeah.
you tend to talk less about that. Can you give some flavor on it?
Maybe Ruud, you can give.
No, we have been a big proponent for solar because I've been quite involved in that. We simply see that the solar market is more difficult. It goes really up and down. As soon as it becomes successful, then often the business models are corrected by the government. Even in Holland this happened, and suddenly things went down. it's been more difficult, but we slowly but certainly increase our momentum in there. We've now quite a bunch of systems in the meantime there. We're working on two big new project. The whole area is our technology is very much geared to the high endVery efficient solar cells. Back contact cells, for example. That area, that part is growing. China has reduced subsidies so that thing slowed down a little bit.
I would say it's been more difficult than we initially anticipated, but we see a steady growth in that. These are also very big lines, so they are 80 to 100 meters long and have quite a nice price tag. It takes a long time to get into the qualification round because in a solar cell, if you buy a solar cell, you want that this thing works for 20, 25 years on your roof. Otherwise, the financial model doesn't fit. Getting into the qualifications takes a long time. It took longer than we thought. It's even worse than, I think, in automotive. It's already very tough to get in. In the solar, it's even more. We have now quite a bunch of lines now.
I think here we show a little bit of the new things we are working on that like what we show for automotive or this wettable flank, for example, in plating or the other part, this whole flexible thing is very interesting. Also high-end machines. Those are some of the new things, but the solar is steadily growing. We'll have every year now two, three lines going in.
Hilko or Nigel? It's your turn.
I have a couple, but I'll limit myself. First, on the next generation flip chip machine.
Yeah.
Maybe it's a follow-up on new product qualifications, et cetera. First evaluations are positive. How does that typically work? How long does that take until customers are-
Typically 3 to 6 months. You have to put the first one in, then they test it, they try it. Then they say, "Oh, it's not working, and this is wrong." Then you go 2 or 3 times there, you fix this, you fix that, and then suddenly it is there where it should be. Then, Richard has shown our competitors. Typically, competitor has a machine also there, or both competitors are there. They measure the performance of these machines against each other. Then what we typically win out at the end is not only speed and accuracy, but at the end of the day, also reliability of the machines. Because our operators, the customer's operators, they like these machines because they have a tendency to run always in spec.
Yeah, that is a thing where we think that we keep our R&D efforts in Europe because especially the Swiss and the Austrians also, of course, the Dutch also, they have this built-in quality feel and a design. That we definitely want to keep. All in all, yeah, 3 to 6 months, sometimes longer. Does that answer it to some extent?
Ruud, you can answer that.
Yeah.
There are some specific differences in design of the three parties. Ours is the most accurate in placement integrity. The other two are, and that's already demonstrated, less in that sense. All three are new standards. All have increased the output by multiple bond heads. As I mentioned already, it's not just evaluation. We also received orders, and also the data. It looks very positive.
This accuracy issue, sometimes if you think about it, you have to put this die always there. The better you do that, if we have 3 micron, another has 4 micron, and say the spec is 5 micron, then with the 3 micron, you're mostly more in spec than if it's a little bit broader. That's also a reason why operators like these machines, because they are continuously putting it on the right spot, so to speak.
Maybe a different type of question. You've shown a couple of slides from ASML. I think every earnings call from AMAT, et cetera, people are talking about advanced packaging. KLA acquired Orbotech.
Yeah.
Just generally, how has that changed maybe your conversations with your customers? Are there new peers now, new competitors? Just in general, how that sort of translates.
Now you have to look at it, in my view, they are completely different worlds. Nothing has changed in the past 20 years since the debate, especially here in Holland. This world is nicely segmented. You have the lead frame assembly world. You have the substrate world. You have the flip chip world. You have a little TCB effort. You have a small growing wafer level, rerouting different processes, different companies. That wafer level is closer to the front-end world. Yes, front-end companies are looking at that. Customers are, of course, saying to front-end companies, "Why don't you do this?" They say, "Well, this is a Mickey Mouse market compared to my EUR 10 billion market." They won't say it in that way, but that's basically what it means. There's a gap between the worlds.
It'll be interesting to see what will happen in the next three to five years. Will that really become part of the front-end world? Or will it be a world in itself? Why is that not clear? Because the accuracies all the time, mentioned also by Ruud. It's a complete different challenge to accomplish wafer level rerouting interconnects than to develop a 95-layer whatever chip with seven nanometers. That's a whole different world and different people. To have that in the same organization is probably hard to imagine at this time.
I think if you look at the last five to eight years, the level of people we talk with now is really increasing. It's really going up.
In the past, you would talk to the old guards, now it needs to be cheap and fast, that's it. Now you start look at where the investments are happening. TSMC is investing heavily in back-end. Samsung is heavily in back-end because they see it as more and more an essential product feature for them to do these 3D structures and so on. You talk at a much higher level. That has changed, in my opinion. We're at a much higher level with these people now.
Yeah. That's certainly true.
Yeah.
That's why you also see these slides and much more. Simply read what they explain to the world. Packaging interconnect has become a bottleneck and will become-
Yeah
an even greater bottleneck. The key question is, how is it going to be organized? Will we, as a simple little Besi company, fit into that world? Or will the customers dictate that'll be an Applied or a Lam world? That's very important. In the meantime, the only thing we can do is simply continue with our wonderful development in those areas. Hilko, you are the next.
Hilko Wiersma at Savant. We have two questions. What can you say about the opportunity of the MicroLED screens for the coming future? Will that be the next game changer for Besi? The second question is about the Huawei ban at this moment. What can you say about the impact so far? We saw two days ago Skyworks Solutions. What can you say at this moment?
Two questions. One on the MicroLEDs. We see a clear interest in that in our big panel machine. This very high accuracy is very attractive for that spot. Whether MicroLEDs will finally be big time. We think one of the most logical ones is starting with watches, for example. The big advantage from the MicroLEDs is that you have a much brighter picture in daylight. You have a better angle. You can look at a watch. Now sometimes you only need to look like this, and that you don't have these things. We see big progress in there. It is a complicated process because I've run big LCD machines. I know one pixel missing, and you see this. Your eye is very sensitive to that. There's a lot around it to make it work.
We see a good chance in that. I cannot say more than this. I can only say we are deeply involved in it. The other question on Huawei is it's a mixed blessing, I think at this moment. What we see is, yes, issues, but we should also move to some of our Chinese customers say, depending on this whole situation, we need to build up more locally now. We're actually starting to see, hey, we get more requests. We also see some movement to Southeast Asia where people say, "Oh, let's move it there." It's difficult. It has pluses and minuses in it. We are not, as Richard said, not deeply in the Huawei chain. We were more in the other chain. We are progressing in that area also.
At this moment, it's maybe not too bad that we were not too deeply in it. Long term, we still want to get also more in the Android phones.
Hello. Anthony Ong, NIBC. Two questions also. Maybe to start off with the mobile phone market. It's getting clear now probably what the new features will be for the next generation phones. Can you maybe elaborate a little bit what that means for you at this moment? Maybe especially with a focus on 3D sensing and touch.
Okay. In the 3D sensing world, that part, there are further improvements in it, and we are working actively on that. Also on the cameras, there are new lens types, for example, being worked on. This whole alignment of the lenses we're working on. You see these features, but I think it will be more and more difficult because the camera is already very good. The incremental improvements are slightly smaller. It will go a little bit slower. We see in the 5G, this whole antenna structure, how you have to do that is getting complicated. We see more fan-out coming in the phones, which is good for us also. There's a lot of things going on in the phone, but I think the next upcycle will probably be 2021 on the phone side.
The 3D sensing and the face recognition? I think now more Android.
We're getting more requests for that because we are on the non-Android side very well in there. People say, "Okay, do this also here." We have to do this very carefully.
I got triggered a little bit by your remark on the epoxy die bonder, where you said that you were working on AI and self-learning.
That's in the machines itself.
Yeah.
We have a few projects now in that direction. You need much more sensors in the machine itself.
Yeah.
You need more camera to look at things, and then machine learns itself, or something go wrong, then I adjust. Normally an operator will see, "That goes wrong. I have to adjust something." These things we are putting in there. A current die bonder already has close to 3 million lines of software in the machine. Half our R&D people are software people by now. This becomes bigger and bigger. These are the things we're doing. These requests come from everywhere, also from China. They come from, of course, clearly from Japan, from Asia, because it becomes more and more difficult to find operators also for these machines. Another element that is, I didn't mention that too much, but you start to see more automation in the machines now. They're fully automatically loaded.
Actually, they don't want operators on the machines anymore.
Is it something that's only for the epoxy tool, or is that also something that can be used for the other-
That's in general. We see this on all our machines, that level of automation. Also this intelligence in the machines is really increasing at this moment. Yeah.
Well, it's good to add still something to that. Forever customers would dream of having lights out factories. Fully automated lines and programmable from a distance. It's fair to say that we have come a long way in that. We can program, for instance, die bonder department 40 machines from one station. As we had mentioned, more and more of these machines become self-diagnostic and self-adjusting. That is based on a philosophy of islands of automation operated, let's say Bluetooth. The other concept, Industry 4.0 or whatever, having all systems linked in an industry where supply and demand is volatile, plus there are a lot of changes to the products. Any program which has started to that concept has not reached the goals which were set at the beginning.
This is an ongoing battle, which to our view, with the ever-expanding products scopes, with the ever-increasing variety of semiconductors, which require all kinds of different settings in the assembly steps, that more the island concept prevails than the direct line integration prevails.
All right. Also the place of software within the supply chain. I guess we talked a lot about hardware side in the supply chain. Do you also have a wide variety in software vendors in there or?
Too much.
Too much.
Still, if you look at all these machines, also there, of course over time, there's a lot of development on feature. As Ruud said, there are so many lines in all these machines, and everyone says, "If I use this, I can combine that and" There's a diversification, and you should expect over time. You have to be very careful, because development of all these software programs takes years.
Also-
We've done some efforts in the past which have also been very disappointing. It's a similar challenge.
Exactly. Thank you.
Yeah.
Just one add-on question. Growing via timely acquisitions.
Yes.
Every time I meet you say me that the market is too expensive.
Yeah. Still the case.
Is that still the case?
Still the case. If you look at the past. First of all, I've said that, and I will say that again. An acquisition only makes sense if you can 100% integrate that product with its entire organization into your existing organization. Otherwise, the words are nice, the PowerPoints are beautiful, but you will not have the benefit of integrated technology, manufacturing in one facility, developing common modules, common platforms. You'd asked about sales for a service, et cetera. All that has to be possible. That's number one. Number two is it's basically number one, but anyway, in my explanation now it's number two, but it's number one by far, is you only do that when a customer-
Asks you or more or less forces you to do that. If you look at our history, why have we done all this? Because the Intels and STs, Infineon, and all the communication people, they have, and it started with Motorola, and then Orbotech, that they drove our acquisition strategy. It's not us.
No.
We only did those steps when we clearly saw what that benefit would be to us. Is it only Intel, or can we use that for all the winners?
Yeah. I think in this nice slide, with say, the acquisition history in the industry over the last years, we've looked, of course, also at all of these ones.
We are looking today every single day.
Yeah.
We found
It's very interesting today, people are all re-evaluating their strategy.
Yeah.
Simply imagine all boards sitting and stock prices have gone down by 50%, and what next?
Yeah.
Those discussions are always happening. Some say arrogantly, "Oh, no, the next time will be double." Others will say, "Well, let's look at it. What are the pluses and the minuses?" There are bigger pictures. The wafer level pictures. All those discussions happen in downturn. In upturns, they happen because people want to cash out, they want to sell. Banks are happily, all your were still part of banks, they're happy to do those deals. 2% of a total, ta-da. Bingo. No.
Yeah.
That's in a nutshell how the world operates. You constantly are on the lookout, how can you improve? We mentioned operating leverage, huh? It also has to add to your leverage. To buy companies on the side which are separate, which you also have to manage. Managing a company may seem not so difficult, but it's extremely difficult. That's the answer why it doesn't happen.
Yeah.
There hasn't been a customer who says, "You have to help us, and we will help you by offering you whatever if you solve that for us." That's how it works.
Yeah. A lot of these cases that you have seen, we hold also to the light. Then, discussed this also with some of our key customers, said, "Does this make sense? Yes or no?" Then, yeah, if it didn't make enough sense, then we stayed away from it. You also, if you have to manage an integration, it's a lot of work for the management. You have to separate from the other tasks. We are a very lean company by now. I hope you understood that we have a very small effective management team, you have to be very careful what you can do with it.
Yeah. Trio.
Thank you. Yes. I thought maybe you could comment about your spares and services business, You've got it here as 15% of sales. I think last year it was less. Obviously, your revenue base has gone down, That business is growing. I assume it's higher margin than your 55%-60% target. In this EUR 800 million mix, can you keep that 15? Can that 15 go higher? What's the thinking there?
Certainly, if you benchmark, we benchmark on everything, one of the things is, of course, customer support. There again, you have to look carefully in who can you make happy doing that? For a long time, there was this policy, we were always late, so to help to save our reputation, we would help with spare parts and service, mostly on a free of charge basis. That was historically the industry. That has changed over time. We've developed that also more in, that varies per customer, how much do they want? We share openly. There are two types of parts. There are vendor parts. Those are parts which are from third parties, which through the internet today, anyone can find them. Then you have custom parts, which we manufacture.
On the vendor parts, those customers who really want us to support that, it's fine, the margins on that are low. On the custom parts, the margins are, of course, high. It's a mix. Then again, through the basic concept, it depends on whether customers want more support or less. We have some customers who ask us to support every quarter. That differs because, or varies because of their end demand, simply to keep their lines running at maximum output. They're paying for service people support. So there's a mix. The 15% is just a number, again. If you benchmark that to others in the industry, in the front end, we know it's more in the 20% range. With the complexity of the machines increasing, you can expect that that would need more support. Over time, it's likely to grow higher.
Also there's a very interesting case. About 10 years ago, we decided to separate that. It's become an independent business, run independently by our own people. It's amazing to see how that has improved the support to those customers who are using that over time. You can easily imagine in the past, any part which was needed for production had a priority over parts needed for spares. At the same time, I think we were the first company to offer a website access to our inventory. Our customers, having certain, of course, entry protection for us, that has increased our support dramatically. In the past, an operator who would need a part had to go to the purchasing department, they had to find whatever, it took forever to get these parts.
Now they have direct link to our stock, everything like that has a value. The summary is simple. In this world, complexity determines the support customers need. With the increased complexity, they will need more support. Any further questions? Next. First row.
Just a quick question on the share buyback program, because it has been almost EUR 100 million the last three or four years. You continue. Do you cancel those shares or do you keep them in treasury?
We keep them in treasury.
Why is that? Why you prefer that versus canceling them?
We simply keep them. In the past, we've used them in terms of acquisition, as a part of acquisition consideration. We also hold them for the convertibles in case we need that. There are several reasons to keep those shares. Next question in the back. Yes, please.
Martin from Gaarde. When looking at your profitability and comparing it to ASMPT and Kulicke & Soffa and your whole product portfolio, do you agree that you act much more at the high end of the market, much more technology driven than your peers, and should therefore be somewhat less cyclical compared to the others?
No. The cyclicality is across the board in equipment, because every equipment is a capital goods. You have less volatility in materials and in support. Over time, we show that in one of the slides, there's no place to hide. It varies. Sometimes in a down cycle, you may have less or more impact. Again, it may sound strange to you, but that is the least of our worries. The key is, where does this product fit into the customer's roadmap? What is the value of that to that roadmap? Whether they buy 10 or one or two, over time, of course, it should be more because we are mainstream focused, not on specialties, also very dogmatic, mainstream focused. There's no direct link between whether you are on the high end, that you have less impact than on the low end.
Fine. When looking at the growth rate of the last 12 months you recorded compared to peers, you have done much worse than them. What's the explanation then for that?
Well, you said it partly already. It was because of the 2017 enormous growth in certain applications where we have grown double, nearly 50% year-over-year. That did not continue in 2018. If you look at that separately, yes, in 2018, we grew less, but our margins improved, our cash generation improved. That's the only thing which matters.
Yeah.
Excellent. No further questions? One last one. Michael, first row, second row, sir.
I had a couple left.
Okay.
Maybe we were talking before the presentation started about the infrastructure rollout of 5G that seems to be happening right now. Maybe a couple of years, but 4G didn't seem to be that much of a infrastructure investment. On the smartphone itself, it was huge, obviously. How does that look in the next two to three years? Do you see maybe smartphones coming towards 2021, but before that there's already some demand?
Yeah, it's a chicken or the egg problem. If you don't have the infrastructure, you buy a 5G phone, but you can't do anything with it. Why buy a 5G phone? The infrastructure buildup is more costly than if you talk to the Nokia, the Ericssons, and so on. It was more costly than they originally anticipated, simply because you need a lot of new things there. That's now happening. I personally expect that in 2021, the next cycle for the phones will come. The other part is we were very well into 2017 to 2018 in some of this new 3D face recognition, password. That's done a big step forward, and then for a while it's flat. I expect also that in the timeframe 2021, we'll get the next thing in the mobile phones again.
We cannot say much about it at this moment. Yeah.
Maybe more specific, is there more investment needed for, or especially more exposed maybe to 5G infrastructure ramp than before? Is that still?
Yeah. I think we're a little bit more dependent on that now, because that has to come before these next steps will happen on our side also. Yeah, the telephone companies are the Verizon, or you name them, they all are struggling a little bit with getting this infrastructure.
Yep
In place. It's happening. If you look at world scale, it's happening everywhere at this moment. Korea is there. Japan is moving on. There's a lot of movement on going. I think the moment is ticking up now.
Big cities.
Yeah.
As Ruud said, and on roadshows, you hear that from investors, it's a major investment of the telephone companies.
Yeah
To set up these infrastructures. Who's going to pay for that? It will not be a straight line.
That's clear.
Yeah.
Maybe last question on one of your strategic initiatives is to roll out Korean smartphones. What about Korean logic and Korean memory?
Korea is, of course, a small country with only very few customers.
Yep.
Very big ones. We are continuously working with them in the memory space to get in there. We have in both, I would say both of the big ones, we are getting in step by step. We've delivered to the biggest one quite some equipment now, fully automated also. That's a positive development, especially in our 8800 Series tools. That's going well. We have some other areas in, for example, there's a company working on panels that has very interest in what we're doing. There's Korea's, a lot is happening, but very often it's important to be in Korea because other ones look at what is Korea doing now. In the past, that would be sometimes Japan, but now often people look also to Korea. I'm very regularly there. Yeah.
Thanks. Thank you all for taking the time, if digesting the material, there are more questions, don't hesitate to contact us. Thank you.
Thanks.