Third Bridge专家访谈称Aixtron或受益于InP光子集成电路设备需求，但主要信息已属行业背景与旧催化

[[00:02:37]]
Within optoelectronics, you have areas such as datacom and telecom photonics, optical sensing, VCSEL [vertical-cavity surface-emitting laser] manufacturing and microLEDs, to name a few. Among all those areas, is there one particular subsegment that stands out in terms of market size and growth potential?

[[00:03:28]]
Up until recently, I would say it was pretty diversified and as you mentioned, there are so many different topics that optoelectronics cover, but recently, there was this huge news regarding the big investment from Nvidia into this indium phosphide, especially, or into companies which work on these photonic integrated circuits and basically the data telecommunication and high-frequency, high-bandwidth technologies, especially based on indium phosphide. I think now I would say this is really a topic which stands out among the other topics which optoelectronics cover.

[[00:04:31]]
When it comes to PICs [photonic integrated circuits] and the focus on indium phosphide, is Aixtron particularly well-placed compared to the other subsegments we discussed? Is the company equally strong across all of them, or is the PIC space where it really excels?

[[00:05:03]]
I think the general philosophy of Aixtron tools is that they are modular reactors, meaning that even though I think the one which is of interest of this conversation is this G10 gallium arsenide tool, so it seems that it's meant for maybe vertical, so this VCSEL type of lasers, which gallium arsenide is used for. In fact, due to this modular nature of the reactor, it's also, in fact, very well designed to have very broad range of precursors, which can be then used and it can be adjusted towards the different technologies. I would say they are equally well placed for both gallium arsenide as well for indium phosphide technologies. Naturally, there are different problems coming for different material systems, which are unique to those material systems. As I said, the reactor is modular, so then the customer can select what works for them and maybe not use other bits which are developed as a solution for a different type of technology, but the reactor base remains the same.

[[00:06:47]]
In terms of the demand inflection point for the optoelectronics or photonics segment, do you think we're still waiting for it? Or are we already in the middle of it? Has that inflection point arrived with announcements such as Nvidia's investments into PICs, or is it still yet to come in your opinion?

[[00:07:22]]
I don't know how it's counted, but I think, yes, Nvidia is the biggest name in data centres and AI acceleration. Since not long ago, as I mentioned, there was this public announcement of, I believe it was USD 1bn investment to both Lumentum and Coherent each, and these are the two main providers of the opto parts for these data centres. I would think this is already a big moment and maybe an inflection point could be considered. We will see maybe the actual ongoing demand will still increase, but the big moment, I think, was this.

[[00:08:33]]
From the standpoint of important customers for Aixtron, should we place Coherent and Lumentum in the top 10 simply because of the incoming investments? And because the company has confirmed Nokia as well, is that an equally important customer? How should we think about the most critical accounts among Aixtron's confirmed customers?

[[00:09:01]]
I know that Lumentum already placed some multiple orders for this particular system or I think it was also mentioned that it is the G10 but they placed the order. I'm not aware of Coherent being yet the customer regarding this topic. Nokia, I think is also very, very important and big player. They also announced but only single purchase of the reactor. Maybe in some cases, maybe in Lumentum's case, they're already thinking about production size, so it's a big production type environment where those reactors can be used. I also just saw that they're building a plant dedicated to this somewhere in the United States, which should open or start the operation mid-2028. I think that already, there's a clear connection to action that they already maybe decided to go for this solution. While in case of Coherent and Nokia, maybe they are still checking for alternatives, and they need to decide when placing a big order with which technology they want to go for. As mentioned, there's one confirmed, I think, purchase regarding this six-inch topic from Nokia, but I believe maybe they would go for even bigger investments in the future. I would guess it's still up in the air, more orders which can come Aixtron's way or might go to the competitor.

I think so the important bit is the performance of these. I guess, Nokia, they are in a process or maybe finished the process of investigating how well does this tool perform, especially on six inch as you may have heard, The industry is moving from four inch to six inch technology, and so it's not just one parameter which then matters. It's the overall performance of the whole process to the final product, which needs to be investigated, and so usually it takes a long time to really tell whether this is a worthy investment that is worth to upscale this particular technology or some alternative path needs to be taken. I think there's still, in the way of investigating, the potential for this technology.

[[00:12:29]]
Given that we're only aware of a single system order placed by Nokia from Aixtron, let's assume Nokia is comparing it to a competitor's system to decide which tool to opt for when moving into production. What tend to be the key purchasing criteria? What do you think Nokia will be evaluating Aixtron against the competition on? Does it yield uniformity or throughput? What are the top factors in this scenario?

[[00:13:15]]
I think, it's a two-sided question in a sense that it might be true that you are comparing against directly competing technology, meaning, you would also go for six-inch wafers and compare two side-by-side maybe, so there is one question. Another one is the whole lines, they worked for many years in four-inch. You are also investigating whether this jump to six-inch is indeed meaningful and whether it really can pay off or maybe the technology of six-inch, for example, growth of the six-inch indium phosphide wafers, maybe at the end of the day, it's still not there yet, and so you might be conservative in that regard and you stick to a four-inch process that is well understood and so then this path could be taken. I think those two questions need to be answered, especially when it comes down to different wafer size. On paper, it's naturally bigger wafer is better in terms of throughput, in terms of how many devices at the end can come from, let's say, production run and then cost per wafer, cost per device can be significantly lowered. If you just look at the wafer area, over 2x can be better, but the problem comes with the quality of the wafer itself.

The defect level is likely higher for the six-inch wafers, and so then usual estimates would place it then the benefit of area alone considering likely the defectivity of the wafer is higher, will be then less than 2x, so maybe more conservative estimate would be 1.5x better. That's already some trade-off, and then it really comes down to your process and how well it has developed and whether you can really perform in a similar way that well-established process in four inch is performing. I think, at the end of it is then very, very convoluted problem, which then you have to test it and see whether this will work in practice. There are definitely big benefits for moving to six-inch wafers, especially also when it comes down to not only cost per device, cost per wafer, but also integration to the silicon line, which is using the larger wafers, so eight-inch or even now 12-inch wafers in production, so whatever, you need some wafer bonding steps or similar integration steps, it's naturally the closer the wafer size is the better, so there are many, many moving parts in this equation. My guess would be first is whether it's four inch or six inch and then if it's a six-inch production, which tool, which manufacturer is the best which suits the needs of the company.

[[00:17:34]]
Regarding the first factor – whether Nokia decides on four-inch or six-inch wafers – does Aixtron benefit either way, given that its G10 platform can accommodate both sizes anyway?

[[00:17:55]]
Correct. Yes, they have, as I mentioned, it's a modular design. Then, naturally, it won't be the same for Aixtron, because it's clear that the demand, at least the thinking is that demand for those devices will be high, for the indium phosphide PICs. If you would go for four inch, then likely you will need to buy more tools, but you are also limited to the fab space to accommodate. Maybe then there is a benefit of highly automated system like G10 of Aixtron's that you cannot have the night shifts or you can run it automatically without anyone. They are operating it at certain hours and long campaigns that can be fully automated. Yes, it's true that to keep this demand likely they will go for six inch just because of the fab space limitation. If it will fail, the six-inch adoption, Aixtron still, as you said, it has the solution for that. Then it's a question whether the adoption of this technology will be as planned, because the fab space is also not something that can be easily extended and easily found on a spot.

It's hard to tell, but I think especially when it comes down to the competitor, maybe, I would guess that Aixtron has more work, so it has more benefit in terms of higher wafers, because as I said, these are modular tools, and they've been using for big wafers, six inch and eight inch for, let's say, silicon carbide and gallium nitride or gallium arsenide tools. I think that they have an advantage in that regard, while the competitors such as Veeco has a huge also, maybe smaller installed base, meaning that it has a large installed base maybe over the smaller manufacturers, but if those recipes work on four inch, maybe they have also an advantage in that it's a well-understood process and usually, they stick to what they know. It's hard to tell whether sticking to four inch would benefit Aixtron. My guess would be that going fully descending on six-inch technology is beneficial at this point in time for Aixtron, because I would guess they are pretty well placed for this large wafer fully-automated processes.

[[00:21:43]]
Given how the optoelectronics and photonics demand curve is looking – especially with the major inflection point we mentioned earlier, such as the Coherent and Lumentum investments – everything seems to be heading in that direction. However, do you see any risks that could slow down the ramp-up in demand, and consequently dampen Aixtron's growth prospects?

[[00:22:28]]
I think as I may be mentioned, I think the biggest stress the way I see it but also I have this bias toward the process, toward the wafer quality because of my background. Take it with a grain of salt but yes, I think failure of this adoption of the six-inch technology, I think would be the biggest bottleneck because yes, you might have more tools maybe to try to keep up with the demand and using the four-inch technology, but I don't think that's realistic. I think there has to be solutions and it has to work on six-inch wafers since it's what is required by this particular topic. In fact, the indium phosphide being grown on four-inch wafers, but it's a rather outlier in terms of semiconductor or compound semiconductor industry, so it's been on these small wafers, relatively small wafers for a really long time compared to many others like gallium arsenide, gallium nitride and so on. That would be, from my point of view, the biggest maybe risk is that the technology for one reason or another, mainly the quality of how you grow the indium phosphides and so the starting wafer and then whether you could really win significantly in terms of the price per device, price per wafer, overall process price, whether you can reach the quality required to have the benefits of the larger area of these wafers.

[[00:24:41]]
I appreciate that regarding the four-to-six-inch transition, there is clear interest and Aixtron is trying, but there might be setbacks on that front. Is that jump inevitable? Is it essentially just a matter of when it happens?

[[00:25:11]]
I think at the end of the day, it will happen because, yes, I mentioned already, it's also for the integration of the silicon processes in general. This has been the trajectory of any other process. You have to go for the bigger wafers, especially when the demand increases. That's one of the keyways to keep the demand to provide enough devices required, so I think, yes, it's only a matter of time.

[[00:25:52]]
Given the clear demand drivers on the photonics and optoelectronics side of the industry, an analyst I came across estimated that optoelectronics will account for about 10% of Aixtron's total equipment revenue by 2029. In that forecast, SiC [silicon carbide] and GaN [gallium nitride] were assumed to remain dominant at around 35-40% each, leaving optoelectronics at 10%. Do you think that 10% figure by 2029 is a fair estimate, or is it conservative or optimistic? What key factors could dictate whether that share ends up higher or lower?

[[00:27:10]]
I think if this topic takes off, and I don't know when it was estimated by 2029. I just looked at latest quarter figures of Aixtron, it was I think already but because of this Lumentum huge order, so the optoelectronics was already on par with both gallium nitride and silicon carbide combined. In that particular quarter, the silicon carbide somehow was rather low in demand, so it's very hard to say what will happen in three years from now. All of these technologies, they come in waves, and now we are definitely riding the wave of this big news of Nvidia and a bigger picture of this data centre boom. Then the optoelectronics plays a very, very important part in the scaling up of the data-centre infrastructure. I think it's then tied to how well this AI boom continues, and again, this is something that I cannot answer, so there are people already claiming that we are in a bubble, and it's just a matter of time when this AI will slow down, the others think in different ways. I think I'm in no position to know this and to give you a really, really good answer to that. I could see this equalling, both gallium nitride and silicon carbide, if the AI data-centre related topic is here to stay, then I think this will become a very important part for Aixtron.

[[00:29:23]]
To be fair regarding the figures, you make a valid point there. The report I was referring to was from November 2025, and of course, the photonics side has really taken off over the last few months.

[[00:29:38]]
I think a year, a couple of years ago, it would have been microLED, which was the main driver for optoelectronics for Aixtron. Then there was also several different players, which they basically dropped their microLED ambitions and then it kind of almost overnight stopped this topic, so yes, very hard to predict.

[[00:30:09]]
A lot of changes always happening in this industry.

[[00:30:13]]
Yes.

[[00:30:21]]
Beyond Coherent, Lumentum and potentially Nokia, are there any other potential customers in the optoelectronics space who could serve as meaningful growth drivers for Aixtron, assuming the company lands them and all the pieces fall into place from an industry growth perspective? Are there any other notable targets it could or should go after with its G10-AsP platform?

[[00:31:15]]
Besides what you mentioned, I have seen the name Smart Photonics is another, something that can also be interested in this. Besides that, I'm not entirely… Yes, I couldn't maybe, so besides Nokia, Smart Photonics, Lumentum, Coherent, these are the another provider that could be a player for indium phosphide

[[00:33:17]]
Looking at the actual tool, when it comes to Aixtron's G10 arsenide or phosphide platform, who do you see as the company's key competitors in the market?

[[00:33:35]]
I would think, it's Veeco who has already mentioned they base it on TurboDisc technology. I would think that that's the main realistic competitor in this regard. There is also Amec or how is it pronounced, is a Chinese manufacturer. At this point of time, mainly if it's for the China market or Asia market, that could be also a competitor, but I think in this topic at the moment, it's this umbrella under Nvidia and then Lumentum, Coherent and then the smaller providers, which are North American and European based, I think Veeco and Aixtron, also given this is considered critical infrastructure, so I think that they would not go for a Chinese provider in this particular topic. Then the Veeco and Aixtron are the main competing technologies.

[[00:35:07]]
Given that Veeco is based on TurboDisc technology, as you mentioned, how does that compare to Aixtron's approach? Is the jury still out on which technology is superior, or does it simply depend on the customer's specific needs?

[[00:35:34]]
I think the latter, so I think there is no clear winner, but there is some clear advantage. There are multiple things to be considered, and then for some customers, likely Aixtron will be the better solution, since I think it's stronger, again, my opinion. I also don't know that much about the performance of Veeco's systems. I know that they have this many, many people who work on Veeco's systems working on indium phosphide. They have this on smaller wafers, so there is definitely interest in maybe continuing to use the known technology and then try to adopt it for the larger wafers rather than going to a completely different philosophy of the chamber design at Aixtron. It's a question of that. I think for automation, the more and more mature solution is Aixtron's, since it's a modular tool, which also, as I mentioned, is used for similar compound semiconductors, so they have a proven track record of going for production runs and automated cassette-to-cassette technology. I think in this regard, Veeco maybe at this stage or I don't know, they definitely should be improving on that regard and likely will have a good solution to counter this. Then it comes down to the performance or on-wafer performance or wafer-to-wafer repeatability because you have to place multiple wafers for this transition from four inch to six inch really to make sense, because you can place also multiple four-inch wafers to reach maybe the same output.

If you can place as many bigger wafers, then naturally the advantage is there going for bigger wafer. Then in terms of Aixtron's Planetary tool, each of the wafer sits on the rotating holder, or a satellite it's called. This allows to really tune each wafer, so there's additional parameters that you can tune the uniformity, which is very important, and the temperature distribution and so on for each wafer. In the case of TurboDisc technology, they all sit on one big susceptor, which holds the wafers and the wafers itself, they are stationary, and so there is no additional tuning parameters there. It comes in both as an advantage and disadvantage, because it also means more moving parts in Aixtron's case, maybe also more opportunity to have some accumulation happening or something not working. Maybe maintenance-wise, it could be both better or worse. At the end of the day, it's really what kind of parameters, output, what kind of output the customer is looking for, what kind of processes they have on these particular tools, so many, many different parameters come into play.

Also how the maintenance have to be incorporated, so how long can you also operate the tool without changing and cleaning parts, most likely Aixtron's is harder to clean because there are more parts there, because it's more easy or more simple chamber design, which again can be an advantage as well as a disadvantage. In this case, you have to consider all these factors. I think this is what big customers are doing and see what works for them best, and there is no maybe clear winner at the moment.

[[00:40:29]]
One thing I noted is that Aixtron's G10 arsenide/phosphide reactor is the first tool to feature in-situ cleaning. Is that a major differentiator against a competitor such as Veeco? Or does it simply come down to whether an individual customer values it, or is in-situ cleaning something every customer would value?

[[00:41:07]]
That's a really, really good question. It is definitely a very important addition since it allows you maybe to extend the repeatability of your process. If you can go back to the initial position after each run or after a certain amount of runs, again, it's based on what the customer process is, because it also comes in with maybe some disadvantage if you run such cycles too often, so both in time that it adds additional time to your process, but also it could give some wear and tear on the equipment. Overall, it's definitely a very positive addition that definitely helps with the process stability and maybe longer time before some bigger maintenance is required. As I mentioned, there are more moving, mechanical parts in the chamber that likely means that there's a more difficult process to clean it manually, so you would need to maybe remove the parts, clean them outside. In the case of the Veeco, it's a simpler sample design, so maybe there is more leeway. It's maybe easier to accept that without even this etch cycle that if I understand, Veeco does not provide this, so maybe even without it, you can just have a rather simple, schedule some simple maintenance after a certain amount of runs, and at the end of the day, you will get similar performance.

It's definitely a good addition, but whether it's really the most important or whether it's a really key differentiator is up to really the processes of the customers, and what are their limits of how many particles they can have, because they are running without etching might accumulate particles from deposits, which are accumulating over many runs. It really then depends on, as you mentioned, what customer values the most.

[[00:44:03]]
Was there any area you thought Aixtron needed to address with its G10 arsenide/phosphide reactor – either to improve its competitive standing or because Veeco was doing something better that Aixtron had to catch up on? We've discussed many of the tool's strengths and potential differentiators, but is there a flip side? Are there any weaknesses to this tool?

[[00:44:46]]
I don't know how much it is important, but maybe now when we talk about it, I'm thinking that it might be indeed important, is the footprint of the chamber itself. Since we mentioned it, to keep up with the demand, you might need more tools and then naturally, you have the fab space limitation. I think the modern solutions for the silicon industry is where the major footprint or major parts of their growth reactors or whether it's some cleaning reactors, so they actually are distributed over different floors rather than the whole tool sits in a fab. I think that would be one thing where there could be a task in improvement of such tools, because now it's all the gas lines sitting on a tool. You might want to distribute this actually in a space somewhere which is not considered a fab itself. At least this is the modern solution for many of the silicon fabs. The footprint, I would say, is one maybe part where it definitely can be improved. The others, hard to tell. As I said, a lot of moving parts, which is also an advantage of the planetary technology since you can have a lot of tuning for each wafer of uniformity, so I think it's also what the customer likes, what the process engineers, what they like.

This ability to have a lot of knobs to tune to get the perfect recipe or they actually want to have a very simple process where actually maybe you are not in control of many things, but if at the end, you can achieve good enough result and it's naturally preferred just to have a more simple reactor, so that's another thing, is the complexity of the tool itself, could be an advantage, could be also simplified for some.

[[00:47:34]]
I appreciate that we discussed the customers who present a large growth opportunity for the G10 arsenide/phosphide reactor earlier. However, is there also a growth opportunity within Aixtron's existing installed base? Are many customers still using legacy G4 and G5 systems for indium phosphide production, presumably on three- and four-inch lines? Is that legacy setup still widely used, and does migrating those customers to the G10 platform represent a distinct growth opportunity in itself?

[[00:48:34]]
I don't know the specifics, but yes, G4, naturally, those legacy tools, they are still in production. Certainly, yes, if the big players are moving to bigger and bigger tools, naturally maybe there is a demand to adjust their processes to the new industry standard that maybe will shift over six inch. If that's the case, certainly, there is this opportunity then to try to exchange or upgrade the tools to the newer versions.

[[00:49:34]]
Based on the report I referenced earlier regarding Aixtron's 2029 revenue breakdown predictions, the analysts had their own estimates for market share across all of the company's business areas. For optoelectronics specifically, they described it as the dominant leader with a 70% or higher market share, given that the G10-AsP is the tool of record for lasers. Do you think a 70% market share estimate on the optoelectronics side was reasonable as of late 2025? Or do you think that figure is overly optimistic or conservative, especially when comparing them to Veeco?

[[00:50:31]]
I don't have really knowledge or recent knowledge of the actual numbers, but it could be correct. I could see that. As I said, it has a distinct advantages, because of this more complex maybe architecture of the whole tool, but the automation is there. It could be if especially the customers prefer that, and the performance-wise, it's as good or even better than the competitor then. It could be reasonable as a major share then of the market of this particular topic.

[[00:51:32]]
In the long run, given that in Europe, the US and the Western world, it is primarily a head-to-head race between Aixtron and Veeco, do you think this is a market where one player can maintain a dominant share of 50% or more? Or, since customer preferences vary, is it more likely to settle closer to a 50/50 split over time rather than one player holding a supermajority – as may be the case today with Aixtron at 70%? How do you expect this market to shake out in the long term?

[[00:52:24]]
I'm certainly in no position to answer that but I think definitely it would be hard to imagine, especially if the topic really takes off, and there is, let's say, Veeco at the moment is lagging behind that they won't… I mean, if I were Veeco, naturally, I would focus then on this, and if you have a clear target and you have also, as we discussed, they have a good solution. Then it's just a matter of finding your customers. In this market, it's very also important how the customer is familiar, and familiarity with the technology, because then they tend to stick to what they know. I think Veeco has a huge installed base with indium phosphide, so I would say they also have advantages in this regard, and I could see them doing also pretty well given the importance of the topic, and also they have presence, more presence in the United States as well, and there are likely big interest there to have this technology as close to home as possible. I would say when it comes to very key topics, I think it could be harder for me to imagine that one manufacturer gets a huge share of this, but again, very hard to predict anything.

[[00:54:32]]
Is there anything particularly important on the GaN or SiC side that you think is worth mentioning?

[[00:55:24]]
In terms of just as a separate topic or…

[[00:55:30]]
Are there any key opportunities or risks on the GaN or SiC side that you think is worth mentioning?

[[00:55:50]]
Again, it's similar in this regard, it's also dominated by the planetary tools, although not really so there is also bigger wafer sizes, and they are moving to different type of reactors. There, it's a topic on its own. I think it's both silicon carbide and gallium nitride, those are even separate topics. It's a new can of worms that would be opened if we go there. Again, I haven't worked on those technologies in particular, but again, can give this rather generic overview and comparison maybe to how well one or another technology is placed with regards to also the competitors and what are the strengths and weaknesses of these tools.

[[00:57:05]]
Is there anything else you wanted to add on the optoelectronics side?

[[00:57:24]]
I saw in your questions that you gave me before this call, such as whether it's impossible for a tool like that to have a larger gross margins, for example. I think it comes down to what I mentioned as that being a modular system. Then I think this gross margin could be really flexible in that regard so that it's… Sure, it depends on the competition, but it's also if you want to increase it, especially if it's some key parts for a particular process for the customer, so you can make this configuration in a such way that naturally, you can lower or increase the price of the whole tool. At least here, from my knowledge of how these reactors are configured, I think that there is a huge flexibility in both increasing it and also adjusting to the competition if need be. The gross margin would increase over time, again, it depends on the competition, how well they are doing.

[[00:58:52]]
It's good to know there is a bit of flexibility on that side and can adjust it either way if needed.

[[00:59:00]]
Indeed, yes.

[[00:59:02]]
Thank you for adding that in as well. I appreciate it. Maybe a good note to end it on. Let me close by saying thank you so much for your time today. I really, really enjoyed that conversation. I found it incredibly helpful, and yes, it was a pleasure speaking with you. I hope you enjoyed it, too.

Transcription ends at 00:59:33 of the recorded material