Good morning, good afternoon, ladies and gentlemen, and welcome to Besi's Analyst Session. Joining us today are Mr. Richard Blickman, CEO, Mr. Ruurd Boomsma, CTO, and Ms. Hetwig van Kerkhof, SVP Finance, whom will provide a review of Besi's business, development, and capital allocation strategy, as well as a discussion of trends in the semiconductor assembly equipment market. At this time, all participants are in listen-only mode. Later, we will conduct a question and answer session, and instructions will follow at that time. As a reminder, ladies and gentlemen, this conference is being recorded and cannot be reproduced in whole or in part without written permission from the company. The audio webcast will remain available at Besi's website. I would now like to turn the call over to Mr. Richard Blickman. Go ahead, please, sir.
Hi. Good morning, good afternoon, ladies, gentlemen, and welcome to our Annual Analyst Presentation. The presentation you can find on the Besi website. Also available in this webcast. The agenda for today on slide three. I will start with a brief strategic overview. Ruurd will take over and focus on what's happening in our end markets and specifically, what are the drivers, followed by product development strategy. I will take over again and discuss with you strategic initiatives, progress, capital allocation strategy, and then we end the session with some outlook comments and a summary. Before we start, this is an analyst presentation which is focusing on our technology, our market position on the company, and all numbers we present are numbers related to the latest press release of April 30, our Q1, and our guidance for Q2.
We will not provide any additional information about the development of the market in relation to our numbers. We will present the Q2 numbers end of July, and you're most welcome to focus on those numbers at that point in time. Let's start with slide five. The H1 2020, business prospects have surprised to the upside. If we simply look end of April, the guidance we provided for the second quarter, and we look at what happened in the market in May, public information tells you that the semiconductor industry so far has fared pretty well. The targets for 2021, 2023, a rollout of 5G and also new features in mobile and cloud infrastructure. We refreshed our strategy in Q1 2020 and also continued the exercise in implementing certain impacts of COVID-19 to our strategy, and the initiatives are clearly defined and well on the way.
There's definitely engagement with the leading mobile and logic players for the next-generation systems. We have attractive capital allocation to enhance shareholder value. Simply, the dividend over 2019 results we have paid out after the annual meeting, April 30, early May. As we guided for the second quarter, this segment in the world is still faring pretty well. There are favorable long-term drivers in place to generate strong advanced packaging growth, where we will focus on in particular in this analyst update. If we go to the next slide six, just as a reminder, the cyclical behavior of this industry and what you can see very nicely is periods of about four-year average. Every average of those four years, Besi has improved its performance, its market position, its margins.
If we simply take the H1 2020 midpoint of our guidance, that would imply roughly a 12.8% increase versus the H1 of 2019. Basically, a cycle return to growth, albeit due to Corona, somewhat less certain. If we go to the next slide, also clearly visible is improved performance cycle over cycle. In 2015, our net was about 16.6% over revenue of EUR 349. And last year, 2019, if that would be the trough in retrospect over revenue, which is pretty similar, our operating income was 25.8%, which is an increase of 58.7%. The next slide, eight, shows you in some more detail the quarterly developments, which is quite interesting here to see, that this recent downturn has lasted from Q3 2018 until Q4, Q1 2020.
As we simply follow the midpoint of the guidance, we saw that with the order intake in Q4 and also in Q1, we see a gradual return to a growth cycle. If we go to the next slide nine, and we analyze the gross margin development of Besi and the operating margin development compared to our closest competitors, ASMPT and K&S, you see a very nice development over the years, Q1 2016, 2017, 2018, 2019, and 2020, and the developments of both gross margin and operating compared to our peers. Next slide, 10. If we look at the spending forecast, 2020 versus 2019, the latest numbers show a similar, roughly flat versus 2019 spending.
A comment here is clearly that in the current uncertain world, it's difficult to really understand how the impact of COVID-19 will be on our industry in the next 6 to 12 months looking forward. In the comments here, we've seen a significant upwards revision in Q2 after the Q1 earnings of the sector. VLSI Research took down its forecast for 2020 significantly directly after COVID hit, have revised it upwards in the second quarter recently. There's a continued strength in logic cloud application, probably also because of the different way of working in the world, offset by a weakness in memory discrete analog, and in particular in automotive applications. If we go to the next slide, we see that also graphically. We see a beautiful development since 2013, all cycles and also some key points in those cycles.
We see the bottom, the last bottom, directly after everyone understood that COVID is for real, in February 2019. We've seen, first in 2020, sorry, then it improved. It showed significant improvement after Q1 numbers and in early Q2. The outlook has improved, but the question is, of course, is this holding up in the current environment? Careful we should always remain. If we go to the next slide 12, we simply see here the assembly market was set to rebound in 2020 despite challenges, which simply from general information in the market, seems to be the status as it is today. It's very interesting to simply follow the uncertainty in the sector, in the world, and so far for the semiconductor industry, that has worked out positive.
Next slide 13 also tells you the current environment, assembly markets turned upwards in Q4, post 35% down in 2019. Recovery interrupted simply by COVID-19, as mentioned earlier. The recent upturn is fueled by renewed mobile investment for the next cycle, increased Chinese investments since second half of last year, and that continued, and continued logic cloud expansion. If we look forward, repetitive to say, a volatile situation, H2 is difficult to predict. VLSI Research expects a significant rebound in 2021 till 2023. The assembly market fundamentals, however, are very strong. 5G could be a game changer. Chinese want to be as much as possible independent from outside semiconductor devices. Mobile makers pack more features, functionality into the handsets. Further advances in the cloud, artificial intelligence, and logic applications. A brief update on COVID-19. If you look at Besi, it's remarkable. Our business is fully operational.
We had in February some slowdown in China after a sort of lockdown in different areas in China, that quickly reopened early March. In the meantime, Malaysia took over, difficulties started in Malaysia, where China took over from Malaysia. All in all, we have been, I would say, very successful and lucky that we had no operational issues so far. European facilities are open. R&D has continued. There's a lot working from home, also alternative schedules, people coming back to the offices. That's all basically working well. Our production cycle is functioning simply because we also have dual sourcing for many of our modules in place. Immediately when COVID became serious, we expanded our supply base simply by minimizing the risk of not being able to supply systems to the market.
We have stacked up a little bit in critical components, simply to be also prepared, if a second round hits, that we can continue to deliver. Simply, if we look how we have operated so far, we can say we are pretty healthy. We had one false case on 1,600 employees, which is amazing, and so simply, Besi is holding up very well in this COVID-19 environment. That brings me to the next slide, the impact of U.S.-China trade disputes, slide 15. It can be said that this conflict is reshaping the global semiconductor supply chains. Since it is highly capital-intensive, it will take some time to see the real impact of how the infrastructure will reshape. We have seen already since the past two years, some further capacities being built outside of China, and on the other hand, China becoming less dependent upon imports.
Besi is well-prepared that if we are confronted with simply less U.S. parts in our machines, all of our systems are being tested for that so that we are able to continue supply even when the maximum content of 25% is lowered to 10%. Our exposure to China is very well manageable. In 2019, it was about 31.5% of our total revenue, which consists mainly of foreign IDMs with operations or through subcontractors in China, and we have a limited exposure to high-end Chinese smartphone manufacturers. If we look at the next slide 16, and some market trends. The latest numbers from Viavi have been released recently, and the basic message is that although the market dropped by 32.5% in 2019, our shares have remained relatively flat in that environment. That's not unusual.
In the past, many cycles, in downturns, we have sometimes lost slightly market shares, but in upturns, we have gained significantly more share than the overall market. If we go to the next slide 17, financial performance improved in the H1 of 2020, if we simply follow the guidance. What we mentioned earlier, in the midpoint of the guidance, revenue should then increase by about 13% versus the first half of 2019, simply driven by renewed mobile investments and continued demand for logic and cloud applications. We see a slight improvement on projected gross margins year-over-year. Operating margins are expected to reach 25% level. OpEx under control, 31.5% of revenue in H1 2020 versus 33% in the comparable H1 of 2019.
Our next slide, our business model objective remains in a next upcycle to reach levels beyond the past peak, where we reached close to EUR 600 million revenue. With the big drivers in place, we should be prepared for higher revenues than what we achieved in the last up cycle. Of course, with similar gross margins, a higher market share, as mentioned earlier, as we demonstrated in previous up cycles, and that then leads to net margins which are comparable to what we achieved in previous peak cycles. Headcount split, 80% Asia, 20% Europe. Currently, it's about 71% Asia and 29% Europe. If we go to the next slide. I will hand over now to Ruurd, who will take you in a wonderful journey through what is happening in our end markets, and in particular, what are the drivers for the next cycle where our growth will come from. Ruurd?
Yeah. Thank you, Richard. It is a special way of doing this presentation, in a virtual way, and I hope you can all follow the screen very well. I'll start with a little bit, slide number 20. Of course, advanced packaging is clearly critical to the next generation of applications. We see the mobile revolution going on, the digital society coming up, and that means, in principle, higher accuracies, greater miniaturization, complexity increasing, density increasing, of course, performance, and also lower power consumptions. Advanced packaging has always been our key element. If you go to the next slide, even Gordon Moore, famous for his law on the advanced side of the front end, where he predicted every two years a doubling, he actually already mentioned in his original publication in April 1965, that back end could be actually also very important.
He literally wrote, "It may prove to be more economical to build large systems out of smaller functions which are separately packaged and interconnected." 1965. Very good that we sometimes remember this. On the right, you see some of the ideas that are playing today, where we split up monolithic chips in chiplets and try to make a complete library of functions that can be adapted. Next slide. We also see that the key players, you see some of them on the sheet here, recognize the importance of advanced packaging in designing their next generation devices. On the left, you see the company in Taiwan, bringing up a complete new class of devices called SoIC, system on integrated chip. They have their InFO, which you see on the bottom side, their CoWoS. Intel, very active, bringing up new classes of packaging.
Foveros, you may have heard this term, which is a vertical stacking. EMIB, which is a bridge connection, and combinations of those. Samsung active. I will not go over all of those, you see also the Microns, the Hynixes, all the ones are busy bringing advanced packaging to a next level. Next slide. What we also see, this was my conference on 3D integration in Dresden in end of January this year. This was the last physical conference we had. After that, COVID struck, it's all now virtual. Even the ECTC, an important one, is spread over three weeks of presentations now. ASE presented something nice. They said, "The value of semiconductor packaging is increasing actually faster than the value of semiconductor revenue." That's a strong statement that packaging, especially with the smaller dimensions, is becoming a more important part in our semiconductor world.
If I go to the next slide, then. We would like to stress, Besi has always focused consistently on the advanced packaging world. We have never digressed to other areas, to SMT type of things or whatever. We stayed very well-focused on advanced packaging. Basically, key is, of course, process knowledge, of course, in die attach, but also in molding and plating. Extreme precision at very high speeds, that is then typically for die attach. Mechatronics is a key element. An item or an issue that becomes more important is optical system development, metrology and vision. We just developed a very nice 2D, 3D vision system where you can see very precisely how a surface is of a die that you are working on. Sometimes it's good to remember that our company in Austria started as a vision company, recognizing number plates for the first time.
Another part that is important and sometimes more important than in front end is, all these processes run very fast, so you need also very fast real-time software, especially in the pattern recognition. All these things are, we are day in, day out working and bringing this to a higher level. On the top, you see some examples, an RF circuit, some other circuits. You see hybrid bonding on the right also. This is our world. On the next slide, I add one more thing to it. We have also focused always on operational and financial excellence. It is nice to do all the technology, but at the end of the day, we also have to show performance because that's what we do for our shareholders. This is based on best-in-class technology, being close to our customers. It's very funny, we do a lot of contacts now over WebEx.
We just had this morning a very big one. We do this very regularly now, but I hope we can start traveling back and see people physically also. Operational excellence is key. We're always busy with cost reduction. An element that we look strictly in is the active cycle management. Our business, you have seen that in the presentation part from Richard, goes up and down very fast, and we are always looking at can we adapt fast enough and not have too much ballast in the company. Another element is the other way around. We have to bring our products fast to the market, and we also have to ramp up the production very fast. We have shown that we can do this very well.
On the right, you see a graph of how often new phones are introduced. That means in the past you may have four or five years. Now, every year, we have to bring new additions. We go to the next sheet, that is slide 26, you see where our revenue is generated. The largest part in the mobile world with the mobile internet devices. Second is the whole computing section. Actually spares and services has becoming a substantial part, very regular and solid income for us. Automotive, 17%. We have industrial and some high-end LED on 9%. I would like to go a little bit over the development in the three main areas, mobile, compute, and automotive. Dive a little bit deeper there. We go to slide 27, you see our key end user markets. These are basically our growth drivers.
On the left, mobile internet. The 5G ramp is really on its way now. Of course, this means more mobile content, more features, but it also opens up a whole spectrum of new applications. On the left, you see things like the massive machine type communications. You can think about smart city, smart homes, gigabytes in seconds that will move. On the right side, you see 3D, ultra-high-density television, and you see 8K coming up. It may go further. Cloud services, augmented reality, industrial automation, mission-critical applications. The idea is that 5G will be also very fast. Of course, in the bottom, we still have the self-driving car. Some other areas, maybe Micro LED display. I will come up later. The second group is computing, and the digital society is clearly accelerating. Cloud infrastructure is growing 12%-13% per year.
Areas like artificial intelligence are everywhere now in our application field. Data mining. Medical is a large area, On the bottom, I put a work-at-home economy because I think a lot of our digital transformation has been going faster than we expected, and many of you will actually probably like that you can work from home and that life is actually continuing this way. Third element, automotive. Though the automotive business at this moment is somewhat down because the number of cars sold is reduced strongly, we still believe this will continue. Whether autonomous driving will be there tomorrow, it probably will take some time, but it's certainly continuing. In the car, you will see more powerful compute systems. If you want to drive autonomous, you have to have strong vision systems around you that have to be very fast.
You see new sensors coming in like LiDAR, also a big part is electrification and also building up the whole infrastructure to do all the charging. All of these, I would say, developments are positive for our business. If I go now to the next slide, a little bit about mobile. We are getting into a phase II. The top sheet was from the latest Samsung investor presentation. We get now mass market 5G devices coming in with several 5G chipsets, standalone network deployment, the adoption of millimeter waves, and of course, industrial IoT. On the bottom, you see this is from the Ericsson Mobility Report. That's always an interesting guideline to see how are things developing subscription-wise. The estimate between now and 2025, there will be 2.5 billion additional 5G subscriptions.
Of course, on the right side, you see another element that is the number of IoT connected objects that will be connected to the net is also growing very strongly. The numbers are fluctuating. People talk about 30 billion, 50 billion. A large number of devices connected to the net. If then I go to page 29, this explains a little bit 5G. One of the new things in 5G is that we will be using higher frequencies. On the top, you see frequency 6 gigahertz already, 28 up to 39 gigahertz. In that range, you also talk about millimeter waves. To do this, you have to build up a complete new network infrastructure. New in this is, besides the massive MIMO stations, you also get small cells and even indoor radios, as they are called. In the phones, to accept 5G, you need many more components.
On the right is a nice plot where you have a 4G phone that had about 40 filters. A 5G phone needs about 70 filters. The same with bands, 15-30. The filters go from 30-75. The switches inside the phone triple, and also, the MIMO devices quadruple. That means all this has to be all mounted in a phone, and hopefully with our equipment. On the next slide, I go a little bit deeper. On the left side, you see the 5G network expansion. This is a slide from Infineon. That's why you see AC/DC components for the power supplies in here. If you see to the left, you see something new that are the 5G millimeter wave antennas, and that's a complete new class of antennas. These millimeter waves have a penetration radius of maybe a couple of hundred meters.
That means in a square kilometer, you need several of these cells to cover. That means an enormous growth rate in that application. Inside the building, you have the same issue also there. You have to put a new type of transmitter working with millimeter waves, and also there is a big growth area. Another new element you see here is called fog computing. The network with all the devices creates so much data that you cannot simply dump this all in the cloud to a big data center. You have to do more local computing, and that means more powerful computing close to where it's being generated. A lot of our customers in the infrastructure, you see to the right, these are all famous names. To the left, you see also something new.
We have all heard of graphic processors, of CPUs, but there's also new infrastructure processors are being developed that are actually also very powerful. On the phone side, to the right, you see, of course, next generation 5G phones. Obviously, they have to have a 5G reception and transmission part. Companies like Qualcomm are supplying this. There's much more handling of power inside the phone because there's more applications running, so that there are new devices being developed there. You get new Wi-Fi. The whole GMS, GNSS networks have improvements there. Even the Bluetooth is coming further. We also see more complicated sensors in the phone. Very many sensors get fused, as they say. That means in one sensor package, you have more sensors built in, which makes actually the handling of the packaging and mounting these devices more complex.
There on the bottom, you see a number of customers we are working with. If I continue now to the next sheet, this is sheet 31. Mobile also creates many new opportunities in what we call the Internet of Things world. To handle all these devices, you see on the bottom left, you see something that is called the edge layer. That is the first layer of calculation levels. This is typically located in buildings themselves. These are quite powerful computing stations. Companies like NVIDIA, Google, many others are developing special boards with high-power computing systems to do the calculation close by. If you move to the right, you see a fully automated factory. That's the dream, of course.
You see a picture where people are being observed, and you see a little picture there with a red line where somebody goes outside an area who should not be there. This type of location calculations needs to be done very fast. You want to do that close by the position. This is typically an edge application. This edge area is growing with about 40% per year, and the expectation in 2030 is that edge computing will be actually bigger than cloud computing. All these devices need quite complicated package processes. I move a little bit to the general data world. We all know that data generating is still exploding. Expectations around 2025 is under 75 zettabytes being available. If you think about it, 10% is original data, the rest is being transferred up and down, and about 30% will be even real-time data.
This whole drive to more data is continuing. If we don't want to drown in energy usage for this, of course, we have to keep developing more energy-efficient packages and more powerful packages. This computing world, the advanced computing, the HPC is another area of real growth besides the standard Internet of Things. Cloud computing, of course, already a EUR 230 billion business. Automotive and manufacturing and healthcare are also adopting high-power computing at a very fast rate. These are actually massive markets that are coming up. Gaming is, of course, an interesting market with under EUR 60 billion. All these markets need computing power, faster data transfer, artificial intelligence, and high-level vision systems, of course, graphics. All based on ever more powerful processors and also clearly on memory needed for this.
On the bottom, you see some cases, of course, the Intels, the NVIDIAs, the AMDs, the NXPs, the Googles are all working on this. In all these projects, we are involved. Going to the next slide. We're now at slide 34. You see on the computing, the data growth requiring more powerful and efficient devices. On the left, you see the latest NVIDIA A100 processor that's built by TSMC on seven nanometer technology. It's CoWoS-based. That means there's an interposer below this. Has 54 billion transistors and works with HBM2 memory and can do 1.6 terabytes per second data transfer. In the bottom, you see the effect of that. A classical data center station would be about EUR 11 million, 630 kilowatts. You see about eight of these big stacks full of servers.
With this new processor, you can bring this down to one stack at EUR 1 million and only at 28 kilowatts energy usage compared to 630 kilowatts. Initially, you could say, "No, that's bad because we'll sell less." We have always seen that these improvements actually generate bigger markets. The model of such an advanced package you see in the middle. This is a substrate. On top of that is an interposer that is typically connected with classical C4 bumps. On top, you see microbumps where you have different chiplets, and you may have even a bridge connecting that. In the below side, you see an application like AMD is doing, where they combine 40 nanometer input/output area with seven nanometer cores. This drive to going to more chiplet-based is only accelerating. If we go a little bit to why this chiplet area.
On the left, you see the very well-known yield versus die area curve. If you want more powerful computing, the most logical thing is to make the die area bigger, because that's the fastest way to get more computing power in the die. Bigger dies have lower yields. That is very simple because of particles and defects. If you make the die bigger, there's a bigger chance you have a defect in that die, and the yield goes down. If you break up this chip in smaller parts, you can actually grow in yield. The improvement in yield, of course, is eaten away partly by the extra cost for now having communication with each of the chiplets. We start to see that it is cost-effective now to go to this chiplet structure.
You see AMD is doing this with an horizontal infrastructure lately, and many companies doing this now with interposers, even active interposers, and Intel is going even in a three-dimensional direction. The advanced computing part is developing very well for us. There's a lot of handling in there for our machines. We have high expectations of this area. If we now go to the next part, which is a little bit on automotive. Despite that automotive is a little down, on the left, you see a curve saying electrical car sales growing strongly. Expectation is that by 2040, the number of electrical cars sold per year will be higher than the number of cars sold with the conventional engines. At the same time, you also have to build a substantial infrastructure.
On the bottom, you see a curve where you say that by 2040, there are more than 300 million charging stations should be available. Infineon always publish very nicely in their presentations, the power content in an electric car. If you then multiply this simply by the number of cars expected, you see an enormous growth in power devices. That's what we are also seeing. We see much more demand now for that, and that's, for us, a positive thing because we have a good position there. We also have the, of course, focus on efficiency and power conversion. You see more silicon carbide, for example, coming up, which has an higher efficiency than classical silicon. We also see that Tesla is, of course, in everybody's wording. Everybody follows this. We also expect that conventional cars will strongly come back probably in 2025.
Maybe we sit together then. It is expected that Volkswagen is actually the leader. Of course, significant infrastructure investments will be required. Going to the next slide.
Slide 37.
We are now at slide 37. On one part, electrification is a strong grower. On the other side, everybody keeps talking on autonomous driving. I drive a lot, so I like this idea. I also think it will still take a long time because before you have everything legally arranged, before the whole infrastructure is really working, probably another 10, 15 years. There will be a slow increase from what we call level two to level three, four, and five. To drive autonomous, you need many more sensors on the outside of the car. You talk about long-range radar, you talk about LiDAR, you talk about camera systems, you talk about short, medium-range radars, even for the parking, the ultrasound, which you know very well, the beeping that you get when you drive somewhere.
If you see what the level of sensors in a level two car, it's about EUR 160. If you go to level four or five, which is more or less autonomous driving, you talk about EUR 1,000 of sensors in there. Second part that comes into an autonomous car is, of course, a very strong supercomputer. You see two examples there for a level three and a level four or five with many more cores to keep this going. Further, you need to further develop the internet to have the connections between the cars and to get the traffic things. We also see there is more collaboration needed between the large players, because also the software effort to get all of this done is substantial. We see this long term also as a good growth opportunity. Let me now turn a little bit to the product development strategy.
If we go to page 39. In short, our main activities in 2019, we're developing a nano-accurate hybrid die bonding system, designing our next generation TCB systems, and also work on the next generation high-speed and high-accuracy fan-out and flip-chip systems. In 2020, we continue on our nano-accurate hybrid system. We have excellent initial results. We also started some time ago with working on Micro LED. I have a few slides on that also. Our next generation TCB system nears completion. Of course, we're also continuing developing our next generation soft solder and plating systems for the automotive and the power applications. Slide 40. There's an overview of the next generation bonding technologies. I mentioned here hybrid bonding. If we see historically, wire bonding is a long time around, 1975, more or less, started. We got into the flip chip world that started 1995.
Pitches of products between 150 going down to about 80 micrometers. The big difference between wire bonding, if you see the left picture down, you see the little wires coming from the chip, and you can only connect wires on the outside of the chip. If you go to flip chip, you can use the whole area of the chip to make connections. That's all the little balls you see there. The next step was fan-out, where you actually expand the area of the chip by molding around it, so you have, again, more surface area to make connections. The next step we saw was the TCB bonding, where you go to a much smaller pitch, up to 80, 50, 30 micrometers. You start at stacking dies. These are some examples, in this case, from Micron, published. Some stacking goes now to seven or eight.
12 is coming up. This technology will certainly stay around a while. The next thing we start to see is what is called hybrid bonding, and especially die-to-wafer hybrid bonding, which is used for pitches below 20 micrometers. I will go a little bit deeper in what is this hybrid bonding. If we turn to page 41. This is a little explanation. On the left, you see two die areas, the top die and the bottom area, and you see a dielectric material and the metal contact. If you polish these surfaces very flat, and I talk about five nanometer flatness, almost atomic flatness, then if you imagine that you put two very flat surface on top of each other, they more or less stick. In the original physics, this was van der Waals force. They call it also dangling bonds bonding.
Basically, you stick two very flat surfaces together, and you cannot take them apart anymore. That's what we are doing here. We create a very flat surface. We activate the surface with a plasma step. We bring these surfaces together. The dielectric in the middle bonds immediately, and this dielectric is already a very strong bond. This whole bond is done at room temperature. You don't have to heat up anything. You just connect it. As a last step, once the dies are bonded with the dielectric, then you bring them to a temperature step, and then the metal that is slightly recessed will expand and will make the final contact. This is a very elegant way. The two new things that come in here is, of course, the surface activation, the very high precision, and with the high precision, also cleanliness.
In the middle part, you see the difference. You see on the bottom picture, you see a classical solder bonded die. You see the copper pillar and the solder cap. You can imagine that if you bring the distance between these contacts smaller, say you bring it to 20 or 10 micrometers, the solder will actually contact the next contact, and you have a short. This limits simply how far you can scale this down. With a direct copper-to-copper bonding, you are not having any solder anymore. You have nothing that squeezes out there. It's a direct, very clean bond, and it's also, you are not needed anymore underfill. The distance is very small now, and the heat transfer also much better. Of course, we see this especially for very dense pitches, say 10 micrometers pitches or less.
That means a lot of contact, means also you have to very accurately bond this. If I go now to the next sheet. This hybrid die bonding compared to flip chip and TCB has substantially better properties. You see on the left table, you see classical 2.5D or 3D or hybrid bonding. You see the interconnect type, it's a micro bump or it's a hybrid bond. The chip distance, 100 micrometer, 30 micrometer, basically zero in the hybrid bond case, and also the distance between the bond pad pitches goes down substantially. The net effect of that, you see the speed compared to a classical 3D IC goes up a factor of 12. The bandwidth density goes up almost a factor of 200, the power efficiency goes up a factor of 20.
To make it more visible on the bottom, you see a classical bump for flip chip. You can do about 120 of such bonds in a square millimeter. TCB, about 800 of these bumps. The bumps are already much smaller. In the hybrid, you can see you can do 12,000, and there is already a publication coming out that it will go to 1.2 million contacts per square millimeter. This also allows, and that is what you see on the right side, for a complete new class of devices where you have chip 1, chip 2, and chip 3 connected almost without any distances and giving very fast data transfers. System-on-integrated chip, this is called, some also call this hybrid bonding, and it opens up a new class of devices. How do we do this very accurate placement? On page 43, you see the base principle of this.
We have developed an optical system. We call it Van Gogh substrate camera system, where we see the green part on the bottom is the substrate where we have to put a die. The blue part is the die you see on here, and we are able, with our camera system, to see both the substrate fiducials and the tool die fiducials at the same time. With the bond head, we can go down to the bonding position in a controlled way. With piezoelectric motors, we then follow the position very precisely, and we've been able to do this at about 150 nanometer now accuracy. You have to imagine, 20 years ago, 30 years ago, this was front-end maximum. We do this now in back-end. What is more, and this is important, we can do this at high speed.
There were systems on the market that do this at 100 dies per hour. We are currently at about 2,000 dies per hour, and we think we can go further than this. On the right, you see a few pictures how this looks then in the real world. Going to page 43. There's, of course, an estimate of how big will this market become. This depends, of course, on the adoption rate. We think that the number of big ones will start adopting this now. Bigger IDMs and OSATs will probably follow, and we'll also get at a certain moment even replacements. Cumulative, we see this as a market in 2030 that the total installed of between EUR 1 billion and EUR 2 billion of tools is possible. We talk about 700- 1,400 systems. Let me turn to another subject. A little bit about Micro LED.
Micro LED, it's a little buzzword at this moment. You think about watches and smartphones. In the middle, you see applications like virtual reality and augmented reality. Here you might have a Micro LED device in glass, and you will be guided. If you're working in the stockroom, you might be guided to where the product is lying. This could be one application. Of course, to the right are probably later on larger sizes TVs and TVs with very high density. Micro LED, I have a little slide on this, slide 46, trying to explain a little bit the basics, what are we talking about here. On the left, a classical LCD display. This display, you have to imagine in the middle, where the yellow arrows are, this is a liquid crystal.
If you put voltage on this crystal, it will turn a little bit, and then it will allow light to pass through, or if you put more voltage, it will stop the light. On the back, you have a backlight, that's the light source. The filter is nothing. It's a big filter that you can steer, and then you get the colors. This is well-established technology, low cost in the meantime. The brightness is somewhat limited. It needs quite some energy. At this moment, a lot of people are working on better backlight sources, Mini LED, not Micro LED, but Mini LED, for example, is used as a backlight for improving LCDs. Yeah, I think developments in LCDs will continue. A second technology that came up over the last years, OLED display. Here, you don't need a backlight anymore. The light is generated in the device itself.
It's basically a light-emitting diode. To get the colors, you have to add an organic layer. This organic layer is also the problem. This one still has lifetime problems. That's why you don't see these applications in, for example, your PC, because if you have a taskbar all the time, over time, this taskbar will not disappear anymore. It's somewhat higher brightness and also somewhat lower energy than a classical LCD, but still has limitations. The next step is Micro LED, that you see to the right. Here, the light is generated directly in the device, but here you use really III-V semiconductors. You have no organic material anymore. Sizes of these LEDs, that's why they're called Micro LEDs, are 10- 50 micrometer. It's clearly an emerging technology. It has at this moment still a high cost, but it also offers the highest brightness, lowest energy.
At this moment, people also expect that lifetime issues will not be there. To build these Micro LEDs, there are two basic production methods. You see this on the right. One is called mass transfer. Here, you pick up a whole bunch of these little Mini LEDs and put them on a back plane. The second way is what is called monolithic, where you create more or less the Micro LED on a wafer, and then take it out from the wafer and mount a whole device on a back plane. This is much more like a classical TCB flip chip type 2 process. Today, there are more than 20 processes under development. There is not yet any standardization. All the big companies work on their specific technology. For us, it is difficult to see which one of these will be the winner.
If we go to the next page 47. On the Micro LED front, we are working on two concepts. One is for the mass transfer. This is a system based on what we call our jumbo concept. This is a large area, die attach machine with very precise placement over a wide area. First system is delivered and accepted. There's interest from additional customers, and this is probably a potential application for smartphones and wearables. On the other side, we're also working on monolithic Micro LEDs utilizing our next generation TCB system that fits very well for that. Probably there, the applications are more like the augmented reality and the virtual reality glasses. We're working on both sides. Success in the whole Micro LED environment depends significantly on improving the productivity. Market potential is substantial, but there's all still a lot of development needed for this.
I now turn, page 48. Two elements we discussed, the hybrids and the Micro LED. We now go to our more classical product portfolio review. You know we have the die attach equipment, die bonding, multi-module, flip chip, die sorting, die attach. We have the molding equipment, the trim and form, and we have also the singulation. Then, of course, our plating equipment. We'll go over these one by one. First of all, in the die attach world, our EVO system. This is by far the most successful multi-chip module system. We have now more than 6,400 installed. You can check the last one from last year. We've progressed very well with this. The key competitive advantages of these systems are the combination, of course, of accuracy and high speed. We now have a tool with three micrometer accuracy available.
It's multi-die, multi-wafer. It has an extreme broad application field. It's very quickly to develop a new application and bring it from R&D to production. What is also nice with these machines, you can actually put them side by side. We have done up to four side by side, that you can do multiple steps to manufacture a quite complex device. Some examples you see here on the bottom, which are some optical devices that we have done with these tools, which have multiple steps in there. New developments in 2020 for this tool is further optical inspection, especially 3D inspection. Next steps in factory automation, Industry 4.0. We get more and more requests for power applications with higher bond forces. This has to do with the upcoming silicon carbide technology and power applications.
Target markets 2022, camera modules, 3D structured lights, infrared, fingerprint sensors, motion sensors in the automotive, LiDAR, VCSEL placements, MEMS mirrors, radar, IGBTs, sintering. Industrial, of course, power IGBTs. In the high power computing, 2.5D interposers. We place interposers with this tool larger than 70 millimeters, which is the only one in the world that can do this at this moment. Of course, in the 5G world for the cloud storage, the photonics, the placement of the photodiodes and the VCSELs. If I go to the next system, we are at the epoxy systems. This is our second high runner. Over 3,300 installed now. The most successful epoxy system in the industry. High accuracy and speed. This system has running capabilities of up to 18,000. What we see more and more is that the very high control for the bond line thickness.
When you put a die in the epoxy, you want to be sure that it is very parallel to the substrate and that also the bonding height is very constant, and we excel in this part. We added in the last year a high level of self-diagnostics and machine learning. We see that the dies get thinner and thinner. We developed a very nice new ultra-thin die ejector, and we now also add confocal inspection on these tools. A new development in 2020 will be six-sided inspection, where with a new optical device, including a liquid lens, we can inspect the die from every side and give a signal everything went well, or you have to divert this one. Also there we see higher bond forces. NAND was a classical market. MEMS and sensors, flip chip on lead frame packages, high-end ICs.
A nice one is also we do some special cooling elements. These are very small Peltier elements. With a very vertically directed silicon part, and this is used to cool lasers used for photonics transmission. Our poka-yoke, our intelligence of the machine, has brought the setup time back from 45 minutes to less than 10 minutes, about 75% reduction in setup time. In the high-speed world for flip chip, our 8800 FCQ is in the market now. From the original machine, the single-head flip chip machine, we have delivered more than 1,000 machines. Currently, we think it's the highest speed on the market. We have a novel quattro principle in die handover, and that minimizes the times you have to touch the die. We also developed an ejection mechanism for ejecting four dies at the same time.
In 2020, we will expand this application, add machine learning to it, and also new inspection systems. Markets, especially flip chip DRAM, there's a large shift in bringing DRAM from wire bond to flip chip, and of course, the general flip chip and fan-out market. Systems are very well qualified now with customers. To the next system we go is the soft solder. There, a big step we are making is we're bringing our classical 2009 system on the level to the 2100 system. It will be a next generation soft solder system. The 2009 was already the standard in the industry. Almost all automotive power devices are made on that tool. We have continuously upgraded this, but we now see a need for a next generation tool, and we're developing this.
There will be also plasma treatment included in this machine and, of course, the fully automated qualified, including traceability. The 2100 soft solder system will have a new tunnel design in that. It's in the left middle picture. We have shown that this tunnel has a much higher cleanliness level than our former tunnel. This has been highly tested now with a number of customers and well appreciated. Target market, of course, automotive, high reliability power packages, silicon carbide, and in industrial, all power devices and the IGBTs. We turn to the next round of systems. We go to the TCB, our next generation tool. We expect market introduction end 2020, especially developed for logic applications, but we're now moving also to memory applications. The competitive advantage are it's both a chip to substrate and a chip to wafer system.
Precision level will be in the 1 micron range, high productivity. The system on the left top you see would be a 5,000 per hour system. We've also developed the capability to use very large dies, up to 70 by 70 millimeter die size. Here we also will implement a tunnel concept to create protective atmosphere with an ultra low gas usage. There are tunnels currently available, but they use much too much gas. We have developed a very nice concept to reduce this. Target markets, of course, HPC, memory, and in some cases, also monolithic Micro LEDs. We have been working on die to wafer with thermal compression bonding already since 2016. On the bottom right pic you see die to wafer with 20 micrometer pitch already done in 2016. Last part here is the die attach large panel. I'm now on page 54.
We have here the largest area in the industry. Very good accuracy, 1.5 micrometers and 1 micrometer. Use a unique measurement frame. We're capable of using all our basic bond heads and high speed with special cooled linear motors. We'll further expand this system. On the bottom, you see a special bond head we designed for this. You also see the enormous good results we get in accuracy in 0.3- 0.5 we have seen already. Auto build, auto tilt is a new thing we're going to introduce in this machine. We talk about HPC, Micro LED, and general ICs. Let me now turn a little bit to the packaging world, to the molding world. Our LM system with over 325 systems, we see a lot of activity at this moment. It's the most cost-effective system in the industry.
It's a standard for thin die molding, also exposed die molding, where we think we have about 90% market share. The strong element of this tool is the dynamic clamp force control and the possibility to compensate for board thickness variation. Double-sided molding and foil-assisted molding are in high demand at this moment. Also, this tool has a very low power consumption and low weight. We'll introduce 180 ton molding system for larger area this year, which has four individual clamp forces, so we can, on this large area, balance the clamp force and get a very even molding performance. Vision system for zero defects and also ESD capability has improved. Applications, 5G, wearables, automotive, memory. On the next page 56, you see some applications. Antenna packages, wearables, all of these done almost exclusively on our tools. On the wafer molding.
We're further experimenting with wafer molding. On the bottom left, you see a case where we do multiple different dies in an exposed version. We continue with that. We introduce exposed die on glass and metal carrier this year. Also here we built in a new particle detection system because for these very advanced devices, you want to be sure there are no particles. 5G and wearables are the main application areas. Next, we come to our high precision trim and form tools. We have now about 1,050 of these tools installed. Handles the most complex high density lead frames. Full product traceability has been introduced. Many offload configurations. Laser marking has been added. Selective punching, if there's one defect, you can push this out in a separate way.
Single product handling will be added because we start to see very big products that come in this machine, also laser deflash. Mainly, it's all lead frame-based power IGBTs and general lead frame-based packages. Basically, automotive, industrial applications. As last one, plating systems. This tool, the top line you see is our standard plating system, over 820 installed by now. It's an industry standard for lead frame plating. If you follow VLSI market share numbers, you see extreme high numbers. Sometimes we're even a little bit puzzled ourselves by this. A key competitive advantage of this tool is the very efficient chemical usage and very little spillover, high precision plating. We also have a solar version for high efficiency cells, and we keep upgrading this tool. We just added the new selective veto system to it.
New efficient power supplies, flexible connector applications, one very interesting for inside the phones. Target markets, industrial, automotive, and also solar. That brings me to the end of the review of, say, what are the markets doing, what key technologies are being developed, where are we with our systems. In a short summary on page 60, we could say advanced packaging is the clear solution for smaller, thinner, more complex, and more functional devices. Assembly is a very important gating step by now. Specifications start to go front-end light. The value of packaging increasing. Decreased node sizes make the necessity for packaging very clear. 5G happening, data generation creating the need for more complex devices. Hybrid bonding micro LED upcoming. We are at the forefront of these developments. I think it's more than enough. I hope you could follow this part.
I think we'll return now to Richard, and questions probably at the end of this talk.
Thanks, Ruurd. Let me switch on slide 62 to take you through our strategic objectives. First of all, in the top you see maintain best-in-class technology leadership. Number two is increase market presence in addressable markets. I think Ruurd gave you many directions which should accomplish that. Enhance scalability, reduce structural costs. In a few slides, some more about that. Balance our business objectives with social, ecological responsibilities. Also look beyond our current product range to be better positioned in the strong, growing wafer-level assembly market, maybe through acquisitions. Also continue to reward shareholders via capital allocation policy above the average of the sector. On the next slide 63, revenue initiatives, simply confirming what Ruurd said again, through partnerships with next-generation leaders in our industry with ever more promising developments, especially in the hybrid arena, but also through ever-growing end markets. Exploit those opportunities.
Increase mainstream penetration with high quality mid-range systems, and expand the presence and share of wallet in China. Our service, combined with operational excellence, we can see over the years a significant growth in the percentage of spares and service as a percentage of revenue. Ten years ago, that was below 5%. Today, it's closer to 20%. On R&D, we have started to invest more, simply anticipating on the major expansion of semiconductor advanced packaging requirements, which Ruurd showed you many of the different activities going on. We added last year about 40 people. Far this year, another nine people. Simply all focused on targeted customer developments with high growth potential, and that should bring us into a next level of revenue in the years to come.
On the cost reduction, in a similar way, every year we define key targets to reduce our cost, and where are those costs in the three areas: supply chain, product design, and overhead. Systems manufacturing rule of thumb, one-third of the cost is materials. The supply chain is extremely critical to continuously focus on cost reduction in all the various components, modules in today about 18 different platforms in the Besi portfolio. Product design, all cost starts where the ink dries up, and that's in R&D. Constantly, our engineers are focused on developing new generation products with ever lower cost, and we can do that through a reduction of platform standardization, reduce cycle times using better subcontractors. We have seen over time an enormous improvement, but we still are certain that we're only at the beginning.
Cost down engineering, the focus not only on technology leadership, but also on cost leadership, is a major initiative for the coming years. The third one is the overhead. We've come a long way in moving from west to east, where our end customers have their operations, and the target is to have in the future, roughly 80% of our headcount in Asia. That's partly due to growth, which we anticipate will come in the next cycle, but also still to further reduce European facilities overhead in the Netherlands, but also in Switzerland. Reduce also our overhead by the new way of working and gain efficiencies, further efficiencies from our spare parts consolidation in Asia, and also more web-based service of spares to our customers. A few words about CSR objectives.
You see that on the next slide 65, where around the circle of CSR, we see in the top, safeguard safe and healthy working conditions, maintain best practices in environmental and ethical behavior, reduce environmental impact of products and operations, promote employee talent training and diversity, conserve natural resources, develop sustainable supply chain, minimize impact of conflict minerals, reduce packaging waste, transportation, and energy, and last but not least, responsible tax practices in all jurisdictions. You've seen these goals each year. The change which will happen this year is that we will add more KPIs to those goals simply to demonstrate our serious approach to our CSR responsibilities. You see that in the next slide 66, where we have the highlights listed 2019, and also what we have achieved and the objectives for 2020.
Review sustainability strategy overall, improve the reporting and establish KPIs and a monitoring, and upskill sustainability initiatives. Next slide. That brings me to the capital allocation strategy. Most of you know that we have a very simple strategy. We distribute between 40% and 100% of our net profit is available to distribute in dividend. Underlying that policy is that we maintain roughly 20% of our revenue in net cash, simply to be prepared in the unpredictable cycles in this industry, the ever-increasing capital required for development programs. Ruurd explained all of those beautiful developments. I added to that how many people we have increased our staff in R&D and simply to avoid any dilution risk for shareholders, we maintain a net cash position of around 20% of revenue. Everything above that, we distribute to shareholders either in form of dividend and share buybacks.
You can see in the next slide as well, slide 69, that over the years we have not only defined that strategy, but we also applied that. We see in very strong years an enormous return to shareholders, but also in down years with maintaining our gross margins in the mid-50s, net margins around 20%, allowing significant returns to shareholders. Since 2011, we have distributed about EUR 732 million to shareholders. We see in the next slide 70 how much that is roughly in dividend trends. You see per year the dividend per share, which we have distributed. It's amazing, in five years, above 90% of our net income has beenReturn to shareholders, of which two years are even 100% over our net income. Anyway, next slide. Share repurchase program, also simply continuing the program year after year.
Far this year, EUR 5.6 million with an average share price of about EUR 33. What we can see here in total since 2015, EUR 72.1 million purchased under the current EUR 75 million program, 3.2 million shares, average price over that period, EUR 22.51. Next slide 72 shows you relatively our return on equity versus ASM Pacific and K&S. It's fair to say that since 2013, we have gradually increased our returns vis-a-vis those two competitors, and that has led to a significant return on equity over time. That brings me to the last slide to finalize with the outlook for the second quarter, slide 74. We guided April 30 that our revenue could be up by 5%-25% versus Q1. Our gross margin in the range between 56% and 58%, and then our operating expenses down by 10%-15% compared to the first quarter.
These are our prepared slides, and if you have any questions, we're happy to answer them. Operator?
Ladies and gentlemen, we will start the question and answer session now. To be registered for the question and answer queue, please press star one. Star one for your questions. Go ahead, please. The first question is from Mr. Marc Hesselink, ING. Go ahead please, sir.
Thank you. Maybe on the EUR 800 million ambition and what we're now seeing with COVID-19. As far as you can see right now, does it have a positive or a negative effect on the future opportunity that you have? Maybe also taking into account the competitive position with, for you, the flexibility to have production both in China and in Malaysia.
Well, in the current time, it won't surprise anyone that there are different views varying from a significant acceleration in demand of semiconductors, especially the acceleration of the rollout of 5G, also data centers, simply new business models, so expected growth. On the other hand, there are also views that maybe the impact of COVID-19 on global GDP could be significantly negative, and that would put some, let's say, delay on the next cycle. I hope that in the slides we presented, Besi's position in those growth areas has improved over the years and is currently very strong, and we are also benefiting from them. We saw that in the turn of tide in Q4, Q1, continued order increase. End of April, we're all very careful.
What I mentioned, CLSA, in the meantime, has changed its view from a negative year 2020 to an expected growth in 2020 for assembly equipment, even double digits. I continue to say it's uncertain times, and who are we to be able to tell anyone which way it will go? Besi is very well prepared for that. We can accelerate very rapidly. We've also demonstrated with downturns that we maintain our margins. We are pretty comfortable in what the future will bring us.
Okay, thanks. My second question is. Thanks for the extensive update on the new products and the even more advanced techniques. Could you maybe indicate a bit the relative size of those markets and how quickly it's moving from your stronghold in flip chip today towards the new techniques?
Yeah. Thank you for this question. We tried to explain that these new technologies are upcoming, and I showed one graph where we estimated how quickly will this go?
I think it will still take a number of years before this becomes a real volume, because it is always with new technology. I think in the coming period, 2023, 2025, we should see this in a much higher volume entry. It does not say that all these other technologies will disappear. What you always see is that, we have seen this even with our soldering systems, they have a very long lifetime because a lot of products keep coming and also have, as Richard mentioned this already, as mainstream, maybe a little bit more middle class. These are large volumes that also keep generating a very good business for us.
Final question. You mentioned first still the opportunity to move more into the mid-market. Is that related to what you earlier talked about, getting a stronger presence in the Android supply chain? If so, can you give an update on where you stand and if you think there is indeed a quick win in market share possible?
Well, quick wins is not, let's say, to be expected. Big wins are always in new technology applications or in new features, where you grow from scratch to a significant mainstream application like 5G antennas, also the AirPods. There are many more devices, and what Ruurd explained is clearly on every product, there's a certain specific application, always on the high end, always on offering customers accuracy and speed as a combination where they simply need less VSM machines than that of our competitors. That has proven over time, if you go back to our second slide, the development since 2006. Cycles whereby we double outpace the growth in an up cycle, and we are more or less stable in a down cycle. That's simply focus on the growth elements, products in every next cycle.
Okay. Thank you.
The next question is from Mr. Peter Olofsen, Kepler Cheuvreux. Go ahead, please sir.
Yes, thanks for letting me ask a couple of questions. The first is on the gross margin. If we look historically, what we have clearly seen is that you have been able to grow your revenues from cycle to cycle, and with a growth in revenues, also your gross margin expanded. If we then look at your ambitions, the EUR 800 million basically implies a doubling of revenues. If I then look at your gross margin target to 55%-60%, you are already in that range. Well, I would say usually if you have a doubling of revenues, you would usually assume also some leverage and in cross-margin expansion. I know you already have the highest gross margin in the industry, but why wouldn't we see any leverage if you would double your revenues? That's my first question.
My second question is on plating, where we have been talking about the solar opportunity, for a couple of years, which I think is potentially pretty sizable opportunity. It seems that we haven't seen it really take off yet. Correct me if I'm wrong, what's keeping that particular market from really taking off? My final question is on R&D. You mentioned that you're adding people in R&D. I recall from a conference call last year, that you were expecting R&D expenses in the P&L, to increase towards a figure in the mid EUR 40 million in a number of years from now. Is that still what you are expecting?
Yes. First of all, gross margin. That's a good point. Clearly when revenues increase and you have better utilization rates in your operations, also the product mix is important. It's very fair to assume that from the current levels, there is certainly an upside to the gross margins. We've also mentioned that before.
You're right that we do not guide for that. It would be a possibility, certainly, if revenues start to increase again. We'll see. On the second question, solar, Ruurd will respond, but I can also respond to your R&D. Last year, we were just below EUR 40 million, about EUR 38 million on an annual basis. We added roughly 40 people. That's EUR 1 million per quarter cost on average. Gradually, we grow to above EUR 40 million, and if all goes well, that also adds to our revenue in the years to come. Otherwise, we've done something wrong. Your math is very accurate. By the end of this year, with adding, as we mentioned, about nine, 10 people more this year, we should gradually grow to mid EUR 40 million on an annual basis. That's the R&D cost figure. Ruurd, on solar.
Yeah. Maybe one remark about this gross margin and the volume. Of course, it's logical, more volume should also finally bring a better gross margin. Over the last couple of months, we have also seen that it's, for us, very wise to have somewhat broader supply chain base to secure everything, and that may have a little bit of an impact on the other way. I think it's possible because I'm heavily also involved in the supply chain, and it certainly is the goal to bring it up. On the solar thing, on the solar systems, I've been always pushing very hard for that because I still firmly believe in this market, but it has been indeed slower than we anticipated.
One of the issues is that for solar cells, you have to give a guarantee of 20+ years that they keep functioning, and people have been very careful doing step by step going in, say, next technology. It took longer for us to, I would say, establish. We've also seen that the classical cells have kept moving up, but now we see clearly that the high-efficiency cells become the next area of interest, and we're working both on the back junction cells, so to speak, and what is called heterojunction cells. Yeah, I see in the coming time, it should bring more than it did in the past.
Okay. Thank you.
Thank you.
The next question is from Mr. Wim Gille, ABN AMRO. Go ahead, please.
Yes, good afternoon. Can you hear me?
Yes.
Very good. Thanks for the extensive presentation. I would like to start off on slide five, which is the well-known chart whereby you basically see that there's a cycle of about four years, where you keep improving your margins in the upcycle. If we would extrapolate this chart, we would actually enter a significant period of growth again as of, well, this year, next year. Obviously, there's two moving parts here, but, I understand that there is a lot of short-term uncertainty in relation to COVID-19 and a possible recession, et cetera.
Given the big trends that you've referred to in your presentation, including 5G, your increased share of wallet in mobile applications, the strong drive for compute, and possibly a return of the automotive market, is it conceivable that we would be kind of replicating this chart in the coming years as well in terms of growth, for the coming four years? Should we take it a bit more cautious and say that the current slow markets will last for a few more quarters to come? Second question would be on M&A. I noticed that you put it in your capital allocation policy. Obviously, it always takes two to tango for M&A, and in the semiconductor industry, it even takes three to tango because you need to have a buyer and a seller, but also your clients need to approve of it.
Would you say that COVID-19 is increasing the optionalities or increasing the chances of M&A in your specific end market? Which kind of M&A should we be looking for in terms of applications that you are targeting? The last one is, on the cost reductions that you mentioned, EUR 10 million, would that be kind of a net number, or is that a gross number that will be eaten away again by natural inflation that you see in any cost base? Thanks.
Excellent. Well, the first question, every company in this industry predicts that the period we are entering right now is a significant higher growth environment than the previous cycle, whether we call artificial intelligence, et cetera. Combined with the package development, the advanced packaging development, what Ruud explained in some detail, the growth drivers for the years to come, combined with the focus on advanced packaging, should offer us a similar growth model and maybe even higher than what we have had in the past. The big picture, longer term, looks, you could say very bright. The current situation due to COVID-19, whether the world will be confronted with negative GDP or whatever disruptions in the global world may occur, historically, that always has a negative impact on the semiconductor industry.
It usually means delays of a year or two years, but the basic underlying growth of technology continues. BESI is ever more positioned in those growth drivers over the past, you could say, 25 years. On your other question on M&A, of course, this industry, like many other industries, is a consolidating environment. We see that with the end customers. We see that also on the supplier front. The difference with other industries is that M&A is driven by technology. Technology in integrating processes in new advanced applications, those are the drivers for M&A. There may be M&As on the lower technology, or you could say, sunset world of this industry, but that's not a world where we are looking at. Our future lies in evermore expanding our offering of products in the applications in advanced packaging.
We are well-prepared if opportunities do occur, that we can act independently, but also offer any potential combination, the synergies from the start. Very focused considerations in every downturn. Many companies are revisiting their long-term strategy, especially the ones who are not so well financially positioned. It's an interesting period where many opportunities are visited or revisited, and time will tell.
On the cost reduction?
On the cost reduction, EUR 10 million. We always formulate every year in our strategic paragraph, on the one hand, you have revenue growth, on the other hand, you have cost savings. The target is clearly an addition of all the programs we are working on to reduce costs. In the three major areas which we have listed, there are continuous programs going on to reduce cost over time. Of course, some of that reduction is absorbed by whatever pressures in product cycles. You can fairly conclude over the past years, and again, back to slide five. Or what is it? Slide six, where you see increased margins over time. Always part is improved product position in the market. The other part is the benefit of ongoing cost savings.
This EUR 10 million is an identified number based on today's situation in all those three areas, and some will be accomplished this year, some next year. It's always over a period of three years.
Very clear. Thank you very much.
The next question is from Mr. Nigel van Putten, Kempen & Co. Go ahead, please.
Hi. Good afternoon, gentlemen. Question on the packaging side. It seems from the slides that the molding business is seeing quite some impetus from 5G millimeter wave, but also DRAM and also power products, if I'm correct, which are probably three of the four strongest driver of the business in the coming years. Do you expect that business to grow more significantly than the attach business? Also in terms of market share, do you expect that to rise to levels similar as the attach business?
That is a complicated question. First of all, in the molding area, we've seen a real strong position now in devices you mentioned, but also in all kinds of advanced SiP, system and package devices, and we've built up a very good position now in there. Of course, we're trying to expand that further. Our market share in the packaging and the molding area has been smaller than we have in the die attach area, and of course, we're trying to get that share up by new applications and by also introducing a number of new features. The die attach area itself, we are really expanding. The thing that Richard didn't mention, but with these new applications, in the die attach, we also see that the equipment price is in a different class now.
A tool for, say, the 200 nanometer world is definitely a different price class than a tool that's in the three micrometer area. That helps also in bringing the financial volume up there.
Well, what you haven't said, Ruurd, in the applications which are shown on those slides. In the ultra-thin molding, our estimated market share is around 90%. That tells you that the competition is currently not able, and that's because of adaptive clamping. That is because of also keeping one side cleaner. Those capabilities offer significant market growth in the years to come, because everything is becoming thinner, smaller. Yeah, we have demonstrated also with margin. Simply again, remember, we haven't said that. It's margin before market share. If you focus on the high-end margin, which is always at the forefront of packaging development, and you maintain that position, market share follows. We have made significant progress in the molding application simply because of ever smaller, thinner applications.
A number of these applications that we show, we are basically 100%, nobody else has been able to do it at this level. Yeah, that brings also our position in the industry up because people recognize this, and we get clearly more demand for this now.
Thanks. A follow-up. Is it an advantage to be able to supply both the ultra-thin molding, and packaging, and die attach, for example, an integrated module in a 5G front-end module? Is that still selected on an individual supplier level?
Always at the end, it's a cost of ownership model which customers need to a final decision. The development trajectory, which is preceding that decision, can vary between two years and five years. It's clear from our product portfolio today that offering out of one hand, the die attach applications and the packaging, and to some extent, the plating as well, but for a limited product range. For the broader product range, packaging and die attach together, has improved our position as the leaders in this industry year by year.
Got it. Thanks. Final question is on high-performance compute together with memory. Seems like both are interesting markets. Thanks for the slide showing the potential for hybrid bonding. Would it be fair, TSMC has said that in a couple of years, I think it's a five-year horizon, they'll make a lot more revenue from HPC relative to mobile. Maybe I forget the numbers, but more in balance. Would it make sense to assume the same for you guys? Just a ballpark number longer term, that maybe high-performance computing together with memory could be a market that is at least similar to mobile, excluding Micro LED?
What we are seeing, I mentioned this also, for example, this edge computing is really taking off. That is one of the reasons why TSMC is explaining that their computing part will be probably comparable to their mobile part. That's why they're also investing a lot in that area, and that will have some impact on our division of shares in that market also. How that will be exactly, we have to see, but it's clearly a positive trend, and it's combined because the advanced
processors that you start to see, they also eat a lot of, we expect that HBM type of memories are really starting to take off now because for these high power computing, that's a very good option. Also those need either TCB or hybrid bond packages.
Yeah, got it.
Sorry.
Sorry. Just as an add-on to that, does it also tie into the higher prices for that type of product? I think maybe, again, roughly a number between, I think EVO is the most used machine for the mobile sector versus maybe this hybrid bonding or TCB applications for compute. Would it be fair to say that's a 5X, 10X higher price? No, I think 5X makes more sense. Price point versus what's commonly used in most mobile applications.
You're not far off there. Yeah.
All right. More value-driven rather than it's going to be huge volumes, right? Okay. Thanks a lot.
The next question is from Mr. Robert Sanders, Deutsche Bank. Go ahead, please.
Yeah. Hi, good afternoon. Thanks for taking my question. Thanks for the presentation. The first question would just be on chiplets. Is there any way of measuring a kind of penetration rate, for chiplets at the bleeding edge, like seven nanometer or five nanometer? I'm thinking in particular things like system-on-chips breaking up with SRAM going off-chip. Is that something you think can happen soon, or is that a sort of longer-term trend?
What we see in this field is that companies like in Taiwan, but also in the States, are re-looking at how to build these devices, especially to get the data transfer well done. This idea that it was a very nice presentation from TSMC a couple of days ago, where they indicated that for the first time, the idea to go to this chiplet not only is attractive from, I would say, the idea of constructing like a Lego block, also from a cost point of view for the first time, if you go to chiplet, you have to think like following. You divide the die in multiple chiplets, it means now you have to place these chiplets, you have to connect them, that balance, of course, in the past was actually higher than the gain you would get out of, say, going to chiplet.
For 5 nanometer and below, it's very clear that this chiplet is also financially a better proposition. That means that we expect that this will go faster in those areas. The idea is to split up, taking up S RAM, for example, is really happening. It also is now, and you have to follow a little bit these publications ideas that the company is bringing to stack that on top of each other in a chiplet format. Does it answer a little bit what you were looking for?
I guess if I look at the last 10 years, front-end equipment spending has outgrown back-end, but that's partly because of kind of cost inflation in EUV, et cetera. I was wondering if you thought back-end equipment could start outgrowing front-end equipment if you take a very longer-term view, because customers like Qualcomm, they do smaller chips, but they leave some of the less scaling-driven chips at older nodes. Therefore, if everyone does that at the leading edge, then your TAM would grow faster than front-end.
We believe that there's a good chance that it will happen. Even if you look at some publications from ASML, for example, they indicate already that the portion of, say, back-end and how you compose these devices is growing substantially, much more than in the earlier years. Yes, there's a good chance that this will have a larger share of the pie there.
Got it. On Micro LED, it does seem like Apple's selected their initial kind of partners already. They're going with some proprietary tool sets for some of it. I was just wondering, how do you de-risk your investment here, given there's so many different proprietary processes? Do you kind of ask for NRE, or do you think that there are steps in the process that are kind of where common tools will be used, and therefore, there's still an opportunity regardless of the technology process that is used by the big guys?
Yeah. You're right. There's an enormous amount of different technologies being proposed, and you simply have to follow the publications. I think in the mass transfer, there's 20 different ways to do it. In the monolithic ones, there are several ways to do it, and most in the mass transfer, they still depend on transferring dies with a sort of pickup mechanism. What we are interested in is supplying the mainframe to do this, but we are open to accept different heads on this machine, and that would allow for different solutions. That's a little bit the direction we're taking.
Got it. I was just wondering if that has started to kind of inflect in the past. It's been talked about, there was a cost-
Yeah
I think it was a cost issue, but is it now inflecting?
It's really starting. We start to see this with one big Korean company is already quite far with this, but the other two big ones are also starting with this now. It's happening.
Great. Okay, thanks a lot.
The next question is from Mr. Brian Chin, Stifel. Go ahead, please.
Hi there. Good afternoon, and thank you for hosting the webinar. That's very helpful information. My first question, it ties back into hybrid bonding. Again, I think you've talked about this, but what application, if you had to look between mobile handsets or high-performance computing, do you see as the biggest initial driver? What are the key issues related to both process and cost that you think could limit the early adoption? If you had to make an early prediction, how large do you think the hybrid bond market size could be in 2021?
Many questions on hybrid. Following. Technically, let's say application-wise, we know that several options are being checked. We think that the first test will be basically one or two dies connected in this method. If you take a volume of, I would say, if one element in a smartphone will be done with this, you easily talk about 25+ machines, to give you a real feeling what that means. I'm not talking all mobile phones, I'm talking mobile phones of the big guys. We are not yet completely sure which one will be the first application, but that will surface in the coming months. On the technology part, what is really new in this application, I try to mention it, you have to work with a perfectly flat surface that has to be extremely clean.
Bringing the particle level down in these machines is certainly a technical issue that we are working on. We hired some of the people that are also working with ASML in bringing particle levels down. We try to get the best people in for solving these issues. We think that for the pitches in the 10 micrometer range, particle-wise, we have it now more or less in grip. Next step is, of course, these pitches will go further down. We have to get to the next level.
Okay. No, that's very helpful color.
Yeah.
Maybe just to kind of relate that to the Foveros technology.
Yeah.
Can you talk about which of Besi's products would be most suitable for that particular 3D integration scheme?
Foveros would be, at this moment, still TCB applications. For some of the panels, we might think about large panel system.
Okay.
Also, there was, last week, a Webex conference where the top guy from TSMC, second in range with Intel and also Samsung packaging boss presented their views on where this has to go. All three of them said, for them, hybrid bonding is a key issue. Also Intel said, "We probably might have to work on this more than we initially thought.
Okay. Thank you. Maybe one last question, just going back to the Mini LED, or excuse me, the Micro LED market. I know that it's pretty fractured at the moment in terms of processing. Is there a particular mass transfer productivity level the industry is shooting for that you could share with us, and kind of how short are we of that target currently? Beyond sort of, the mass transfer is pretty critical. What are the other key areas of that ecosystem that maybe still needs development? When you look out the next couple of years, is there a particular year you could kind of circle now that says this could be sort of the year we see more rapid adoption?
Well, probably we need another two to three years at least to get this at a level where you have the productivity. What is an issue at this moment, there's different, I would say, pickup mechanisms, being, I would say, getting to some level that is working. Again, you still have to rework sometimes. One option is to place, if you have one Micro LED, to place actually two and having your controller away to shift away from the defect one. That, of course, means double placement of Micro LEDs. That's one way. Other things are repair schemes. This is quite complicated if you go to bigger screens. You have to also loosen it up then.
The other part is also good testing methods of a complete screen, how you can test as well for every pixel, because you know if you have a television screen, even if it's a normal LCD screen, if you have one pixel that's not working, especially with LCD because that's normally in the off state, the pixel actually is showing up. That is still a critical point. There's still work to be done there.
Thank you so much.
Yeah. I think also on the initial part where you make the Micro LEDs, there's still some work to be done also on the cost side.
I really-
fab processes. Yeah.
Yeah. Really appreciate the color. Thank you.
There are no further questions. Oh, one moment, sir. There's another question from Mr. Wim Gille, ABN AMRO.
Yes. Good afternoon. Just a follow-up question on the hybrid bonding application that you developed. Given where you are and where the industry is heading and where competition is heading, would you say that you would get an above-average market share in the hybrid bonding segment?
No. Our goal is a very substantial share in that market. We think at this moment, the indications we have is that we are leading in this area. We're talking here about, you have to be careful. The hybrid bonding is already known wafer to wafer, and that is mainly used in the camera field, in making the part there. That's a very limited application field. As soon as you go wafer to wafer, you are very limited in yield. Also, of course, you can only do this for products with exactly the same sizes. The die to wafer is really coming up now, and there our development goal was to really bring a tool that has productivity. Yes, what we see at this moment, we are leading, and we'll certainly try to work on that position.
If we make a comparison with our flip chip activities that we did in the past, we basically had a large part of the flip chip market because we were early successful with it. We would like to repeat that and even further expand there.
Thank you very much.
There are no further questions, Sir.
We would like to thank everyone to participate in this 2020 analyst update. Next year, we sincerely hope that we can meet many of you in person. Stay safe, stay healthy, and thank you again. Bye-bye.
Ladies and gentlemen, this concludes the Besi webcast. Thank you for attending. You may now disconnect. Have a nice day.