Ladies and gentlemen, welcome to the ams OSRAM conference call on third quarter 2026 results conference call and webcast. I am Moritz, your Chorus Call operator. I would like to remind you that all participants will be in a listen-only mode and the conference is being recorded. The presentation will be followed by a question and answer session. You can register for questions at any time by pressing star and one on your telephone. For operator assistance, please press star and zero. The conference must not be recorded for publication or broadcast. At this time, it is my pleasure to hand over to Jürgen Rebel, Head of Investor Relations. Please go ahead, sir.
Hey, good morning, everyone. This is Jürgen speaking. Welcome to our third quarter 2026 earnings call, despite the global Microsoft server issues. Some of you might not have received all the emails, but as we have a large group of participants, we believe that most of you have gotten the information. With us today is Aldo, our CEO, who will comment on the business performance and strategic progress. Rainer, our CFO, will walk you through the financials. Please refer to the Q3 earnings call presentation available on our website. That should still be accessible. We checked it. Aldo, please take us through the quarter.
Thank you, Jürgen, and good morning, everyone. Q3 marks another step ahead in our transformation. The core business is growing despite mixed end market demand. The financial profile is improving, and our AI Photonics development is moving closer to commercialization. Let us turn to slide three and start with the quarter. Our semiconductor core portfolio grew 3% year-on-year on a like-for-like basis, led by 8% comparable growth in Opto Semiconductors. Group revenue reached EUR 829 million. Adjusted EBITDA margin came in at 16.3%, above the midpoint of our guidance. We secured approximately EUR 1.3 billion of new semiconductor design wins in the quarter. One nice example of this is the flagship Android smartphone design win, combining three of our most advanced optical sensors. In AI Photonics, our 850 nanometer thin-film microVCSEL platform achieved the key performance and reliability milestones.
The Credo-led wide-and-slow initiative in advancing ecosystem validation and industry standardization, while our micro-emitters featured in Credo's active light cable demonstrator. In summary, the core portfolio is growing, design momentum is strong, and in AI Photonics, our innovation pipeline is translating into tangible customer and ecosystem validation. Let me now ask Rainer to walk you through some of the financials.
Thank you, Aldo. Good morning from my side as well, and I am on slide four. Q3 was another strong quarter against the backdrop of all the portfolio changes. Group revenue reached EUR 829 million, up 3% sequentially. The increase was driven by continued growth in OS, Opto Semiconductors, and the seasonal upswing in the traditional automotive Lamps and Systems business. Reported revenue declined 3% year-on-year due to divestments. Portfolio changes weighed on the year-on-year comparison with approximately EUR 60 million. On a like-for-like basis and at constant currencies, the semiconductor core portfolio grew around 3%, led by the 8% growth in Opto Semiconductors, while the consumer business was down. Including Lamps and Systems, the group core was up by 4% on a like-for-like basis. Adjusted EBITDA was EUR 135 million, corresponding to a margin of 16.3%. That is again above the midpoint of our guidance.
Sequentially, adjusted EBITDA was broadly stable despite the changed portfolio mix and the transition to low-margin manufacturing services for the divested non-optical sensor business. Year-on-year, adjusted EBITDA declined by 19%. The main driver was the deconsolidation of non-core businesses, with additional headwinds from FX and raw material costs. Reported EBITDA was much higher at EUR 503 million. The difference to adjusted EBITDA primarily reflects the gain from the sale of the non-optical sensor business. Now let us turn to the segment performance on slide five. OS delivered another strong quarter. Revenue increased 9% sequentially and 9% year-on-year to EUR 399 million. Adjusted EBITDA rose 19% sequentially to EUR 77 million, and the margin improved to 19.2%. Higher volumes and better factory performance drove the sequential improvement.
With the closing of the sale of the non-optical sensor business, the division CSA has been renamed to Light Sensors and Photonics, in short, LSP. LSP reflects the portfolio transition most visibly. Revenue was EUR 227 million, down 12% sequentially and 16% year-on-year. Excluding the portfolio effect from the divested business to Infineon, the year-over-year decline was approximately 6%, mainly reflecting the weakness in the Chinese smartphone market. Adjusted EBITDA was EUR 21 million, with a margin of 9.1%. The margin reflects the manufacturing services provided to Infineon at close to zero margin. At H1 pricing, the LSP margin would have been around five percentage points higher. Lamps and Systems delivered a strong seasonal performance. Revenue increased 11% sequentially to EUR 204 million. The auto Lamps and Systems business alone grew 9% year-on-year, demonstrating continued momentum and market share gains.
Adjusted EBITDA increased 40% sequentially to EUR 47 million, and the margin expanded to 23.1% on higher volumes and improved factory performance. We certainly benefit from the fact that main customers ask us to jump in after the insolvency of our key competitor, First Brands. Overall, the last quarter shows that our streamlined core portfolio performs in difficult or mixed market situations, such as a weak Chinese consumer market and despite sluggish auto OEM sales. Now turning to slide six. Total semiconductor revenue was EUR 625 million, up 1% sequentially and down 2% year-on-year on a reported basis. On a like-for-like basis and a constant currencies, however, our semiconductor core portfolio grew approximately 3% year-on-year. This gives a better view on the underlying performance of the business following the portfolio change. Automotive revenue increased 2% sequentially.
On the same underlying basis, automotive grew by approximately 4% year-on-year, as continued content and market share gains helped more than offset softer market conditions. Order patterns remain short and the market competitive, but our design wins continue to translate into share and content growth. Industrial medical revenue was EUR 196 million, down 4% sequentially due to the deconsolidation of the non-optical sensor business. Underlying growth remains strong at approximately 23% year-on-year, driven by continued industrial recovery, particularly in horticulture and professional light. Consumer revenue increased 5% sequentially in line with the seasonal smartphone ramp. Year-on-year, revenue declined 13%, primarily reflecting weakness in the China Android market. In summary, the streamlined core portfolio performs well even amid mixed market conditions.
Growth in Opto Semiconductors and industrial medical, continued design win conversion automotive, and strength in consumer custom products balancing the weakness in China to a certain extent. With that, let me ask Aldo to comment on design wins and latest progress in Digital Photonics.
Thank you, Rainer, and let's turn to slide seven. Now let us look ahead. In Q3, we secured approximately EUR 1.3 billion of new semiconductor design wins, bringing the nine-month total to approximately EUR 3.8 billion. This keeps us on track towards our approximately EUR 5 billion objective for the year. The momentum is broad-based across our end markets and regions. In automotive, we continue to win across forward lighting, signaling, and interior applications, with particular strength in advanced high-definition forward lighting. For the first time, we also secured new platform design wins for the new upcoming panoramic head-up displays. We come to that in a minute. In industrial, we saw continued traction in professional lighting, particularly in Europe and Americas. In consumer, we won significant new business in display management, camera enhancement, and ambient light sensing. Importantly, we are also seeing tangible progress across our Digital Photonics growth factors.
EVIYOS continues to scale with more than EUR 50 million of additional design wins in Q3. In advanced optical sensing, our multi-zone direct time-of-flight platform is moving further into commercialization, with the first flagship Android smartphone design win ramping soon. In AI Photonics, we are advancing our 850-nanometer thin-film pixel technology and developing the wide-and-slow optical interconnect system. We come to that later. Together, these wins strengthen our future revenue visibility across established platforms while building the foundation for growth from our emerging differentiated Digital Photonics opportunities. Now let's take a closer look at one of the very tangible examples of growth in our core automotive business just mentioned. Time on slide eight now. The windshield is becoming a widescreen display, and we are the key technology supplier enabling it.
Instead of looking down at multiple screens, navigation, safety alerts, and other relevant information can be brought directly into the driver's natural field of view. Keeps the driver's eyes on the road, put information into the context where it's needed, enables a much more intuitive cockpit experience. This is where technology comes in. These much larger projection areas require high brightness, high contrast, and very efficient light sources. Our advanced automotive LEDs are a key building block and enabling bright projection or bright illumination across a significantly wider area of the windshield. We're working with leading tier ones and OEMs. The market is still in early stages, but according to Omdia, panoramic head-up display volumes are expected to grow from less than one million units today to more than eight million units in 2031 and beyond.
For us, this is a great example of what we mean by growth in the core, taking our established leadership in automotive LEDs into emerging higher value applications and increasing our content per vehicle. Slide nine brings us to a commercial product in the market today. The vivo X500 Pro Max is the pinnacle of camera-centric smartphones and hailed by many tech reviewers. You even get an accessory 400-millimeter telephoto lens for it. To enable this kind of professional-grade photo performance, it combines three ams OSRAM sensing technologies: a behind-OLED spectral sensor, color sensor, the multi-zone direct time-of-flight sensor, and a multi-channel spectral sensor. Together, it supports smooth display adaption, faster and more accurate autofocus, reliable subject tracking, natural color reproduction, and stable image capture under challenging lighting conditions. This is a good example of our system relevance.
Our value is not one discrete component, it's the combination of emitters, sensors, algorithms, and integration know-how that improves the customer experience. Let me now turn to AI Photonics on slide 10. AI computers scaling faster than copper can keep up. Increasingly, the next bottleneck is connectivity, moving enormous amounts of data between accelerators with low latency, low power, and acceptable cost. Copper remains attractive, but its practical reach becomes more constrained as bandwidth and system density rise. Traditional optical solutions provide reach, but can add cost, power, and complexity when multiplied across the much larger number of scale-out connections, as illustrated in a drawing of a data hall. Our answer is wide-and-slow. Instead of pushing a small number of optical lanes to ever-higher speeds, we distribute bandwidth across many parallel optical channels, operating at lower rates per channel.
The target is to remember: optical reach, copper-like latency, and a path towards copper-like economics. This is not about replacing every existing optical architecture. Silicon photonics and indium phosphide will remain important, especially in scale-out. Wide-and-slow is designed for a different optimization point: massive parallelism in tightly coupled AI systems. Slide 11 illustrates the strategic opportunity we see in AI Photonics. Our ambition is to move from being a supplier of individual components to providing an increasingly integrated photonics platform for high-volume applications. We combine micro-emitters, photodetectors, CMOS know-how, thin-film processing, through-silicon via integration, advanced wafer-level packaging, and test within one industrial platform, all from under one roof. The key differentiator is our thin-film technology, which enables top-emitting devices and is designed for dense, highly parallel architectures.
Combined with our industrial manufacturing experience, this gives us a technology platform that can address not only device performance, but also integration and manufacturability at scale. On the left you see our technology stack. In simple terms, electrons enter from the bottom, photons leave and enter from the top, and heat can be managed laterally. During the quarter, our 850-nanometer thin-film microVCSEL platform demonstrated error-free 32-gigabit-per-second NRZ operation at approximately 0.25 picojoules per bit on pixel device level, with a pitch below 50 micrometers. We believe that 850 nanometer will be a sweet spot, as it's based on readily available gallium arsenide, even on eight-inch, and it allows using standard multimode optical fibers that are very cheap. Key elements of the technology stack have been validated, but it's of course not yet commercial revenue. For that, our route to market will be pragmatic.
Start with pluggables or active optical cables to accumulate field hours, then move closer to the compute package as reliability, manufacturing, and system integration mature. Turning to slide 12. Technology alone is not enough. A new interconnect architecture needs an ecosystem and an interoperable path to adoption. That ecosystem is now taking shape. We joined the Credo-led OCP workstream for interoperable short-reach micro-emitters, micro-emitter optical links. This brings together participants across emitters, fiber, connectors, integration, and AI infrastructure, as you can see from Credo's press release. Our micro-emitters were also featured in Credo's active light cable demonstrator at the OCP trade show, moving the architecture from an individual component proof point into a functional system environment. We are working with BizLink on the critical fiber and connector interface. This is essential because high core count alignment and scalable connector architecture are central to commercialization.
Wide-and-slow is developing from a compelling architecture into an interoperable industry platform. We remain early in the commercialization journey, but the direction is clear. The technology has been demonstrated, the ecosystem is forming, and the commercialization is advancing. With that, let me hand back to Rainer for more on financials.
Thank you, Aldo. Let's return to the financial profile on slide 13. The end of September cash on hand stood at EUR 1.39 billion. Available liquidity was approximately EUR 1.9 billion. Net debt, excluding the Kulim 2 sale and leaseback, declined to EUR 883 million. Including the sale and leaseback, net debt was EUR 1.34 billion. During the quarter, we continued to optimize the maturity profile. We announced the tender offer for the 2027 convertible bonds with maximum acceptance amount EUR 143 million, and we just extended this as we expect the closing of the sale of the CMOS imaging sensor business to indie soon. Our maturity profile is now staggered, and the strong liquidity position provides substantial coverage for near-term obligations. A look at the cash flow on slide 14.
Free cash flow was -EUR 62 million in Q3, driven by the EUR 569 million of divestment proceeds from the sale of the non-optical sensors to Infineon. Operating cash flow was -EUR 61 million. This mainly reflects the continued reduction of factoring and reverse factoring, EUR 25 million repayment of customer prepayments, as well as the quarterly interest payment from our senior notes. Cash flow related to the CapEx was -EUR 45 million, fully in line with our disciplined investment approach, though for Q4 we expect a bit more CapEx. Important to note, the divestment proceeds significantly strengthened the balance sheet. Underlying cash generation was also burdened by transition effects during the quarter. Execution of the Simplify program continued. We have delivered more than EUR 20 million of run rate savings to date toward the EUR 200 million target by fiscal 2028.
This program remains a central lever for lowering the cost base, increasing speed, and improving agility. Now let me hand back to Aldo for the summary and outlook.
Thank you, Rainer. Now turning to slide 16. Let me summarize the quarter in three points. First, the underlying core is growing, also in mixed market situations. Opto Semiconductors grew 8% like-for-like. We delivered revenue of EUR 829 million and adjusted EBITDA margin of 16.3%, above the midpoint of the guidance. Semiconductor design wins reached approximately EUR 1.3 billion in the quarter. Second, we have materially strengthened the financial profile. The divestment proceeds reduce net debt and support a strong liquidity position. Third, new Digital Photonics growth factors are moving closer to commercialization. In AI Photonics, our next-generation thin-film microVCSEL platform delivered important proof points. And the Credo OCP initiative, the active light cable demonstrator with our micro-emitters, and our work at BizLink show that the ecosystem is forming around wide-and-slow. There is still a lot of hard work ahead of us. These are early commercialization steps.
We are building the next growth chapter of ams OSRAM, a stronger core, a healthier financial profile, and a focused Digital Photonics portfolio designed for profitable growth with differentiated technology. With that, let me close with the outlook. For Q4, we expect revenue between EUR 750 million and EUR 850 million and an adjusted EBITDA margin of 16% plus-minus one half percentage points. The guidance assumes an euro-to-dollar exchange rate of 1.12. In semiconductors, we expect a typical seasonal decline. In Lamps and Systems, we expect a usual seasonal peak in the traditional automotive lamp business. For fiscal 2027, we continue to see positive free cash flow.
We are entering this next phase with a reshaped portfolio, a stronger liquidity position, and lower debt burden. Our focus now lies on execution, convert core design wins into revenue, deliver the Simplify savings, and move our new Digital Photonics opportunities towards commercial scale.
With that, we are at the end and happy to take your questions.
Ladies and gentlemen, we will now begin the question-and-answer session. Anyone who wishes to ask a question may press star and one on their telephone. You will hear a tone to confirm that you have entered the queue. If you wish to remove yourself from the question queue, you may press star and two. Questioners on the phone are requested to disable the loudspeaker mode and eventually turn off the volume from the webcast while asking a question. Anyone who has a question may press star and one at this time. One moment for the first question, please. The first question comes from Sébastien Sztabowicz from Kepler Cheuvreux. Please go ahead.
Yeah. Hi, everyone. Thanks for taking my question. First one is on the order intake over the past few months. Could you comment a little bit on the dynamic there and the kind of backlog you have today? Talking about a bit the visibility you can have for the coming quarters. That would be the first question. The second one is on AI Photonics. Based on the current roadmap that seems to be accelerating on microVCSEL, I'm just curious about when do you expect to record the first revenue from, I would say, cloud optical interconnect, and do you believe there will be first revenue from microVCSEL before seeing microLED coming in, or both could finally ramp up simultaneously? Thank you.
Okay, Sébastien. Yeah, happy to answer those questions. Yeah, order intake over the last month was okay, I would say, around one for the semiconductor business. We actually see a good order book for Q1. It is, however, at the same time still the case that order behavior is quite short-term. Also there's still some external effects as well that are moving around, and we're not that clear yet whether there will be an impact or not. For example, the last days and weeks, I think the PCB shortages became much more visible. Talking to a number of my customers at Electronica, they see the impact and are struggling to get enough PCBs to build their products.
So at the moment, nothing of huge magnitude yet, but you can see the supply chains are being strained by the huge demand of AI in many ways, not only through memory. That, of course, also has an impact. And we see that impact mainly in the consumer space. That's why the Android phone business is down so much, because less phones are being built because of availability and price of memory there. On the automotive side, also there, of course, some limitation because of memory. But this PCB topic seems to be increasing, and with that, we also hear some first rumbles of some of our OEM customers closing down a little longer over Christmas. So, nothing big yet, but you can see how strained the supply chains are.
At the same time, as the supply chain is the limitation, it shows that demand overall is not that bad, but it's just hard for everybody to work through these supply chain topics. On your second question, revenue, I think if you look at the technologies and the applications, we see that our micropixel approach is really attractive for this, let's say, one to, whatever, 75 meters distance play. So, very much in this active light cable environment, where anyway it's good to start because that's a safe environment to try out new technologies. You get hours on the clock in terms of reliability and performance in an environment where the cable is still easily replaceable, and that will be our first product. MicroLED, however, also remains very relevant, but that's really for the short distance, below a meter or so.
If you think about memory disaggregation, for example, you're talking about a few 10 centimeters, 30, 40 centimeters already would be a huge relief in terms of being able to put more memory a little further away in a second or third row, and with that, really increase your memory availability around the GPU. And that's really, in our mind, mainly microLED territory. However, you get much closer to the basically co-packaged optics that you're then working, or co-package area that you're working now on, and that requires much more integration work, and that therefore also will take longer to materialize.
The next question comes from Amelia Banks from Bank of America. Please go ahead.
Thank you for taking my question. My first question is just on automotive. I was wondering if you could just help me understand maybe why you could be underperforming some of your peers. I know that it's sort of a different exposure, but it's been cited as quite a strong market in Q3, and just wondering maybe where you're seeing the weakness with an automotive. Thank you.
Well, it very much depends on the specific domains that you're in. I think in the LED space, we're definitely not underperforming our peers. We are continuing to gain market share slowly but steadily, especially on new products are ramping where there's very few companies that can provide it, especially in this high-density forward lighting. It's only us and our key Japanese competitor that can service this, and this market is growing. So, we feel quite confident that our market share is stable to slightly increasing at the moment in the LED space. So that comparison, I think, is okay. Versus some of the other peers, it depends a bit on where you are exactly in car platforms with what kind of technology, so that is a little bit more difficult to compare.
As I was saying before, also we see a pretty decent automotive market at the moment, and also the outlook for next year is quite okay. I think the basic underlying trends are pretty much the same, but it is just the specific homes that your product land in that might be different from semiconductor group to semiconductor group. In our LED space, I think we are performing well.
Okay, brilliant. Thank you. Just my last question on the adjusted EBITDA margin. Just wondering if you could maybe help me bridge where we are today, sort of 16% going into Q4, over the long run, where you sort of see the gain getting to the 25% target. Is it just from operational leverage, or is there a cost-saving program, or is there other elements you see improvement on? Thank you.
Yeah, I mean, that's obviously a combination of quite a few things. I mean, what happened was with the disposal, we lost quite a few percentage points of margins of a key business, the sensor business, while on the Opto Semiconductors side, the trend is intact. I mean, we have had wins certainly this year from the gold price, also from FX rates. But underlying, we see that the business is improving its margin. We will see over the next years a combination of massive cost reductions, and it is not only the Simplify program that is taking costs out. It's a lot of value-added measures in the production. It's procurement savings, productivity improvements. So we will see massive savings next year. The pipeline is really full.
And that, in combination with the volume growth that dilutes the fixed cost, that will give us a margin boost over the next three, four years. And with that, on the OS side, we certainly will reach the 25% earlier than on the LSP side, where margins are a bit diluted from that low-margin business we have now with Infineon. But within the next years, I'm quite positive that we will get there.
Okay, perfect. Thank you very much.
Then the next question comes from Harry Blaiklock from UBS. Please go ahead.
Hi there. Thanks for taking my questions. I have two quick ones. The first is just on the Chinese auto market. It sounds like you're seeing relatively kind of weak trends there, which again, to an earlier question, is kind of a different dynamic to what you're seeing at peers, but it's actually more in line with what you can see going on in the kind of domestic auto unit sales that are being reported. I'm wondering whether you can just give a bit of comment on, yeah, what you're seeing in the Chinese auto market.
Yeah, I think you're right. The domestic demand is not strong in China. There's, of course, a lot of pressure or push for export. For us, China is a very important market, as you know, so that might also be one of the reasons why some of the comparisons might be a little bit worse. China is very meaningful, and we see that, yeah, overall build volumes don't increase anymore like in the earlier years, and the export growth, which actually is helpful for us because, of course, on export vehicles, the car makers pay even more attention to quality and IP safety, and that's, of course, where we also shine. Overall, we are keeping our share in China also stable.
There is some pressure, understandably, on the lower end of the product portfolio, but we're gaining quite a bit of share against other international players across the whole portfolio, plus what I outlined before, this high-density lighting. This is actually a big thing in China. This is also for us probably going to be next year the biggest single market for our EVIYOS product. So you can see that they are very willing to accept new technology, and as you know, we play a major role in that.
Got it. Super clear. My other question is just on the optical interconnect side. I know over the last maybe three, four months, there had been some news flow and commentary coming out about some challenges that microLED optical interconnects faced just around crosstalk and dispersion. But I think there's been some commentary more recently around kind of technology being able to address some of these concerns. I don't know whether that's something that you've seen at all, and yeah, whether you're looking kind of more favorably at the microLED solution.
Yeah, well, for the microLED solution, like I outlined before, it is still a little earlier days, I would say, the per-pixel solution. Here, we really need to work closely with the guys who make the GPUs and give us access to their shoreline. That's a journey that we'll have to take on together with, for example, a hyperscaler that makes his own chips, things like that. That is still early days. We do believe, and also our customer interactions continue to underline that we think we can make this work, and that it is a very attractive way because the microLED is, on the short distance, just a lot more energy efficient than a micro-pixel. We continue to think that this is for the memory disaggregation, for example, a very interesting use case.
For the longer distances, as I said before, we believe that micro-pixel will probably win the race because it's just easier also to couple in the focus light from a micro-pixel to a fiber that then has to carry tens of meters. That's a topic that is, yeah, already much closer to realization than the other one.
Great. Thanks, Aldo.
Welcome.
The next question comes from Janardan Menon from Jefferies. Please go ahead.
Hi. Good morning. Thanks for taking my question. Just going back to timing, you have said in the past that you have got the stabilization in your core business, and then you have the potential for the smart glass in a more nearer timeframe, and then the optical is more further out. Is that the way you would still characterize it, or is there a possibility here that the micro-pixel that you are saying is moving closer to commercialization could come in the same timeframe or perhaps even earlier than the smart glass opportunity?
No, I would still see it in that sequence. The smart glasses is really in revenue quite close by. As you can see also from ECOC and other shows, there is still work to be done, not only by us, but also by the ecosystem as a whole to make the micro-pixel become a product reality. It is clear that the industry is working towards that. However, the other part is already much further progressed than this opportunity. The sequence is still the same, but fortunately, we are moving forward, and with that, everything gets closer.
Understood. On page 11, you have got the diagram of what the ams OSRAM sort of full solution is, which includes things like the TIA. Are all those components already in existence? I mean, are you in a position to provide that full optical engine solution that you are showing there, or is that something that you are still having to work on?
Well, these are all capabilities that we have, and they are all in development, with different parts being in different stages of maturity. Although I think that if you look at the driver and TIA and the interposer, that is stuff that we do in the millions for our, for example, ALS products, right? I mean, that is all not that strange to us. Also photodiodes, we make in millions, but we need to tune them, of course. That work is going on. So on the micro photodiodes array, the micro emitter array, that is very active development work where we still sometimes need to even truly develop new things. On the driver and TIA, I mean, this is standard technology building blocks that we, in brackets, just have to focus on this application.
The capability is there, and the integration depends really a lot now on how our first customers, how much they want this integrated from the beginning. Some of our customers already have their own driver and TIA, so they might be willing to supply this themselves and just, in brackets, want the top part in terms of the emitter array and the photodetector array. Some others might want the full stack as we outline here. I think the uniqueness is we have the full stack, and over time, we are quite optimistic that also by having capability for this full stack, we can optimize that. Once you get to a bit more stable design, you can optimize the cost of these components and get them either better performance or similar performance at lower cost. Both, of course, is helpful.
My last question is, when Marvell bought, or in the process of buying Lumentum, I think they have said that this is a scale-up, scale-in opportunity that they see, which they can work together and make that happen. They have talked about the scale-in potentially being closer than previously expected, which could even be 2029 kind of timeframe. Your comment on microLED would suggest that you are seeing that opportunity a little bit more further out. Is that the way I should read it?
No, I think roughly speaking, I mean, I think it is everybody's guess whether it is 2029 or 2030, but that timeframe around could be the first commercialization of this microLED CPO topic. Much earlier, I don't think it will be, because neither we nor the other players have the technology ready to go right now. I think on this very narrow or short reach, in my mind, it will go microLED, and I think there we are quite unique with what we can offer. I think still, as I said before, the microVCSEL will be before that, but that will be more a scale-out game or scale-up game, sorry, than a scale-in game. The chip-to-chip is in the longer term, in my mind, the holy grail in terms of volume and there are two things to think about there.
The one is the optical engine, and at the moment, there are glasses out there with an LCOS imager, and they still need, as an illumination source, high-power LEDs. We provide those too, but these LCOS glasses are sold in pretty small quantities, so that is not really that much a topic. That will change in terms of also the EUR per pair of glasses. It will go up significantly once the microLED solution is introduced, where you basically replace the imager as well with LEDs. Not only you light up the imager, but you are the imager. You can imagine that comes at a very different price point, and that is then really the step up to this EUR 50 to EUR 100 revenue per pair of glasses. The other part is that we also supply sensors to these glasses.
For example, they also need an ambient light sensor. They also, in a number of cases, are looking at your eyes to see basically eye tracking, where you are looking at. Things like that are also there already today, and we sell some components out of our LSP portfolio into glasses as well. But these are, relatively speaking, inexpensive parts. There will be already today low single-digit euros in a number of the glasses out there for us, but this will increase both in volume and in content per glasses dramatically fairly soon.
Fairly soon means in the first quarter of next year, then?
That is very soon.
Very soon.
Fairly soon is a bit after that, but not too far away.
Okay. Thank you.
As a reminder, if you would like to ask a question, you may press star and one. The next question comes from Craig McDowell from JP Morgan. Please go ahead.
Hi, good morning. Thanks for taking my question. The first one I had was on the consumer piece, and obviously quite weak during this quarter. I think you flagged mobile and Android being weak. Just wondering to what extent you agree with comments made by Qualcomm and Skyworks that they are beginning to see the bottoming out of the market in terms of low-end Android, beginning to see a pickup. Is that something you are beginning to see also? I have got a follow-up as well. Thanks.
Well, I would tend to agree that that is probably the reductions are stopping, but we are now still at a very low build volume. Before that truly picks up, the memory needs to be, I think, much more accessible. Of course, our customers are working around it as good as they can, and they have had some time now to do that. I think basic principle problem of memory being just not available and very expensive still makes it hard for mid and low-end phones. The real low end, I cannot comment too well on because that is not so much the space we play in. We are in mid and high end. There, I would say, we probably have seen the end of the reductions, but I would not talk about growth yet.
Understood. Thank you. My second question was on margins, and in particular, the OS. Just would be grateful to help us understand whether that sort of 9% that you posted is a good run rate, whether there is any kind of seasonality with the new quarter that we should be aware of. Maybe linked to that, I think going into 2026, you are flagging lots of input cost headwinds. Just in terms of how your inventory works, four to six months of inventory on hand, should we see the Q3 margin as largely reflecting the peak of the input cost headwinds? That is aside, obviously, from the restructuring and so on you are undertaking. Thank you.
Yeah. I mean, you could say that for LSP, the 9% are kind of the new reality. I mean, there are two effects. I mean, we sold the profitable business, but now also we are supplying to Infineon at close to zero margin. That is certainly dilutive. That is roughly a 5% effect. So without that, we would be at 14% now.
Over time, that business will slow come down, and we will replace this with a normal margin richer business, and then replace basically 0% with something that has a decent margin, and that would have a major improvement. Will take a few years. At the same time, we take a lot of cost growth and cost outs, and then we have the effect from the low consumer mobiles, particularly in China, which is kind of also kind of running the factories a bit emptier, and that is also not good for the margin. So kind of assuming there will be a pickup in the, particularly in the Android consumer, and then we are replacing kind of the foundry services with some real business, and then over time, the margins will start to increase.
That's nothing that I would expect yet for next year, but then after year, we will see a continuous improvement. Again, at the same time, we will take a lot of costs out.
Super. Thanks, guys, very much.
Then the next question comes from Robert Sanders from Deutsche Bank. Please go ahead.
Yeah. Hi, good morning. I just had a question on Kulim 2. If you choose to keep that facility in the end, what does that facility look like today? Are there idle MOCVD tools there? Have you sold a lot of the tools? I'm trying to understand in a scenario where you were to keep that open, perhaps via a JV, what is actually in there and what needs to be spent on it to get it back up to production? Thank you.
Yeah, this is still in great shape, obviously, and we still have a number of MOCVDs there that we can unpack if needed. So in that sense, there is capacity to be had. Also from the other line, of course, a lot has been sold. Before we start to manufacture LEDs again there, we might have to add a few pieces here and there to get up to a full line, but that depends very much on the customer application that you then would put in there anyway to then probably tune and tweak the line. So, yeah, there's still quite a bit available, but there's also quite a bit to tweak on before you reuse that then in full swing in such a scenario.
Got it. What is the fungibility of those tools between different compound semiconductor types? Obviously, if you worked in that microVCSEL, it would be gallium arsenide. I mean, how should we think about that, or is it relatively easy? There seems to be a big debate about how easy it is to convert tools from one to the other.
Well, for us, these tools, there's some specificity to it, but basically, the fortunate part for us, our pixels are not in the phosphide laser, so it is a gallium arsenide-based system that we anyway use also for our red LEDs. So in that sense, a lot of equipment is the same, a lot of the processes are the same. That's also, I think, what's the attractiveness also of this solution, that it is based on a LED technology that we have scaled in volume since many years on gallium arsenide with the thin-film approach that we are also now proposing for the pixel part. So in that sense, there is not a big hurdle. Once we get there to scale, we know how to do this.
Got it. Just last question. In terms of the 850 nanometer thin-film pixel and the relationship with Credo, they did talk about a second source. Presumably, that would be on 850 as well, which would exclude Lumentum and Coherent as a potential second source, but Broadcom seems to be on 850. What's your view of second sourcing? Is that something at a later stage, or is that something that could happen in the early stages of this ramp for ALC? Thanks.
I think that is a question you have to ask Credo. I think the picture that you saw was how they now started and envision an overall ecosystem for wide-and-slow. Of course, it is good for an ecosystem to not have one player in one modality. So, obviously, we are listed there on top and clearly noticeable, but others will and should follow in my mind even, because for this technology to be widely adopted, probably we need one or two players more so that the industry feels comfortable moving this way at scale.
850 is definitely the way forward in your view?
There are different ways to roam, I would say. It depends on how you optimize the overall system, and 1060 is also a possibility. It is not that that is dead. We just feel that 850 is more advantageous, and the systems are different, but not that different, I would say. The benefits for 850 in our minds is, on the one hand, for us, the scalability, also for our customers, and secondly, also the ecosystem. You can use cheaper 850 nanometers OM2, OM3 fibers that are already readily available. So there is, I think, an appeal to this story, but does it now completely exclude 1060? I think that would be also too black and white.
Got it. Thanks very much.
You are welcome, Robert.
As a reminder, if you would like to ask a question, you may press star and one at this time. It looks like there are no further questions at this time, so I would now like to turn the conference back over to Jürgen Rebel for any closing remarks.
Yeah, thank you, operator. Thanks, everyone, for dialing in today and asking questions. If there are any follow-on questions, please do not hesitate to reach out to investor relations. With that, I would like to wish you a great day, great weekend, and speaking to you latest next quarter. Thank you and goodbye.
Ladies and gentlemen, the conference is now over. Thank you for choosing Chorus Call, and thank you for participating in the conference. You may now disconnect your lines. Goodbye.