AIXA.DE FY2026 Q3 Earnings Call Transcript
Date: 2026-10-29
Source: Financial Modeling Prep

Operator: Ladies and gentlemen, welcome to AIXTRON's analyst conference call for the third quarter of 2026. Please note that today's call is being recorded. Let me now hand you over to Mr. Christian Ludwig, Vice President, Investor Relations and Corporate Communications at AIXTRON, for opening remarks and introductions.
Christian Ludwig: Thank you very much. A warm welcome also from my side to AIXTRON's Q3 2026 results call. My name is Christian Ludwig. I am the Vice President of Investor Relations Corporate Communications at AIXTRON. With me in the room today are our CEO, Dr. Felix Grawert, and our CFO, Dr. Christian Danninger, who will guide you through today's presentation and then take your questions. This call is being recorded by AIXTRON, and is considered copyright material. As such, it cannot be recorded or rebroadcast without permission. Your participation in this call implies your consent to this recording. Please take note of the disclaimer that you find on page 1 of the presentation document as it applies throughout the conference call. This call is not being broadcast by webcast or any other medium. However, we will make a transcript available on our website after the call. I would now like to hand you over to our CEO for his opening remarks. Felix, the floor is yours.
Felix Grawert: Thank you, Christian. Let me also welcome you all to our Q3 2026 results presentation. I will start with an overview of the highlights of the quarter and then hand over to our CFO, Christian, for more details on our financials. Finally, I will give you an update on the development of our business and our guidance. Let me start by giving you an update on the business development of the third quarter of 2026 on slide two. The important messages for Q3 are: We have received strong new orders of EUR 275 million, more than double the prior year level, and 40% above Q2, predominantly driven by optoelectronics and increasingly also Power Electronics. Revenues came out at EUR 184 million within our guidance of EUR 180 million ±20 for the quarter, reflecting the ongoing ramp-up of the Photonics business. We also continue to generate strong cash flow. Operating cash flow reached EUR 138 million in Q3, and free cash flow came in at EUR 128 million. This was primarily driven by higher customer advance payments, reflecting the strong order momentum and supporting the upcoming production ramp. We are fully on track to achieve the full-year guidance with revenues of EUR 560 million ±30, as we target a very strong Q4 with a further increase in output beyond the level of Q3. Finally, construction of our new site in Malaysia continues and is fully on track. Christian will now provide a detailed look at our financials on the following pages before I take over with an update on our market. Christian?
Christian Danninger: Thanks, Felix, and hello everyone. Let me start with the highlights of our revenue development on slide four. Q3 revenues marked the next step in our production ramp, increasing to EUR 184 million from EUR 59 million in Q1 and EUR 115 million in Q2. This was within our quarterly guidance of EUR 180 million ±20 million and 54% above the prior year quarter. For the first nine months, revenues amounted to EUR 358 million, 3% below the prior year period. Optoelectronics accounted for 61% of equipment revenues, GaN and SiC Power Electronics for 19%, and LED and Micro LED for 18%. After-sales contributed EUR 88 million, up 10% year-on-year, and represented 25% of group revenues compared with 22% a year ago. Now let's take a look at the financial KPIs of the income statement on slide five. Gross profit in Q3 increased by 88% year-on-year to EUR 87 million. The increase reflected both the higher revenue volume and the favorable product and regional mix. The gross margin improved to 48% from 39% in the prior year quarter, primarily due to the favorable product and regional mix. We do not consider the Q3 margin level to be representative of Q4, and have therefore maintained our full-year gross margin outlook. For the first nine months, gross profit amounted to EUR 145 million, and the gross margin was 40%, three percentage points above the prior year period. The nine-month result also includes a mid-single-digit euro million, one-off expense related to the personnel reduction implemented in operations in Q1. Operating expenses in Q3 were broadly stable year-on-year at EUR 31 million. For the first nine months, operating expenses amount to EUR 96 million, up 2% year-on-year, mainly reflecting the 17% increase in R&D costs. EBIT in Q3 increased to EUR 57 million from EUR 15 million in the prior year quarter, benefiting from the higher gross margin and operating leverage from the higher revenue volume. The EBIT margin improves to 31% from 13% accordingly. For the first nine months, EBIT amounted to EUR 49 million, corresponding to an EBIT margin of 14%, three percentage points above the prior year period. Let me now turn to the key balance sheet and cash flow indicators on slide six. Working capital decreased by EUR 254 million compared with year-end 2025, primarily due to higher customer advance payments and the collection of receivables related to last year's fourth quarter revenues. Trade receivables declined to EUR 110 million at the end of September from EUR 131 million at year-end. Customer advance payments increased substantially, reflecting the high order intake, the composition of current orders, and individually agreed payment terms. At the end of September, advanced payments stood at EUR 309 million, compared with EUR 44 million at year-end 2025. Inventories increased to EUR 323 million from EUR 284 million at end, mainly reflecting higher work in progress for shipments scheduled in the coming quarters. Trade payables rose to EUR 41 million from EUR 34 million as purchasing activity increased to support the ramp. Both developments are consistent with the planned increase in output. Operating cash flow reached EUR 311 million in the first nine months, compared with EUR 128 million in the prior year period. In Q3 alone, operating cash flow amounted to EUR 138 million. The main driver was the increase in customer advance payments. Free cash flow amounted to EUR 290 million in the first nine months, compared with EUR 110 million a year ago. In Q3, free cash flow was EUR 128 million. Capital expenditure amounted to approximately EUR 22 million in the first nine months. For the full year, we now expect capital expenditures of around EUR 40 million. This includes baseline investments and approximately one-third of the announced EUR 40 million investment in the Malaysia expansion, with the remainder shifting to next year due to timing. Liquidity, comprising cash equivalents, and other current financial assets increased to EUR 945 million as of September 30th. After deducting the liability component of the convertible bond, net financial assets amounted to EUR 591 million, compared with EUR 225 million at the year-end 2025. The equity ratio was 58%, primarily reflecting the increase in total assets following the bond issue. With that, let me hand you back over to Felix.
Felix Grawert: Thank you, Christian. I would like to continue with an update on the key trends across our end markets. Overall, optoelectronics remained exce ptionally strong in the third quarter. In power electronics, customer utilization continues to improve. Order activity is gradually returning, and customers are preparing new capacity. These developments provide increasing evidence of a beginning recovery in GaN and SiC. Let me start with optoelectronics, which was once again the most dynamic and the most important market for our business during the quarter. Before we briefly discuss GaN, SiC, and MicroLED. Optoelectronics remained our largest business and accounted for approximately 68% of equipment order intake in the third quarter. Equipment for indium phosphide-based optical applications represents the majority of our backlog, which provides revenue visibility well into 2027. The structural demand driver remains unchanged. The rapid expansion of AI and data center infrastructure continues to fuel demand for lasers used in high-speed optical interconnects. Underlying device demand forecasts have increased materially year over year. For example, in its May 26th report, LightCounting increased its 2030 demand forecast for 800 gig and above transceivers, as well as LPO and CPO ports from approximately 198 million units to 279 million units annually, representing an increase of around 40%. This corresponds to the customer demand, which is well ahead of the initial estimate that we made earlier this year. In particular, we see several acceleration factors supporting demand in the near TAM for movement. The first factor is the rapid AI infrastructure build-out. Communications laser architectures are becoming increasingly sophisticated. High bandwidth requirements require higher laser power, and that, again, requires larger areas, which mean more wafers. In many cases, this also requires more complex device structures and as a result, additional epitaxy steps. This increases the demand for advanced high-performance MOCVD deposition capabilities, which mean more tools. It is also important to point out that we see demand from a broadening base of customers worldwide. The first large orders initially came from the U.S. and Europe. We are now also seeing broad and strong demand from Asia. This includes Japan, Taiwan, Singapore, and China. Finally, we see the industry's transition from 100 millimeter to 150 millimeter taking much longer than initially expected. The vast majority of orders today is on 100-millimeter wafer diameter, even on 75-millimeter wafers, driven by limited substrate supply of 150 millimeters. Based on this, we expect the migration to 150-millimeter manufacturing to take longer than previously anticipated. As a result, customers are likely to remain on 75 millimeter and 100 millimeter production platforms for longer, which increases tool demand. Taken together, these structural growth drivers and near-TAM acceleration factors give us increasing confidence that datacom lasers will remain one of the most attractive and fastest-growing end markets for our industry over the coming two years. With this, let me move to Power Electronics, starting with gallium nitride. In gallium nitride, utilization rates are back to normal. First customers start planning expansions and new tool purchases. We expect this trend to continue through 2027. The main driver for the increased demand now comes from the power conversion in AI data centers. With the new 800-volt HVDC architecture, NVIDIA has created a new power standard together with its partners. In this standard, the majority of down conversion steps from 800 volts to the supply of GPUs and CPUs is made with GaN power devices. Here, we look at medium voltage to low voltage devices in voltages such as 200 volt, 100 volt, 40 volt, and 20 volt. In AI data centers, the overall opportunity extends beyond traditional power supplies. GaN is being considered for intermediate bus converters, battery backup units, and auxiliary power supplies. The focus is on improving power efficiency in the data center infrastructure. This shifts the focus of new capacity investments in power GaN towards the ICT infrastructure at lower power classes. In contrast to the early GaN investments of the 2021 to 2023 cycle, which focused on 650-volt consumer devices such as fast chargers. These broad market applications of power GaN remain an important pillar, but are not the main focus of the recent activity that we currently see. We currently see the coexistence of 200 millimeter and 300 millimeter wafer sizes confirmed that we had projected earlier. Customers with large 200 millimeter fabs, fab capacities stay on 200 millimeter, while customers with existing 300 millimeter fabs focus on 300 millimeter fabs for their new investments. We are discussing volume orders for both wafer sizes with customers. With this, I come to silicon carbide. Demand for silicon carbide has been soft throughout the year. Yet customer utilization rates have been continuously improving, and now the first new tool orders have been coming in. Also, selected silicon carbide substrate vendors have mentioned capacity constraints for SiC substrates, and we see the growing volumes in our after-sales business. Where does this all come from? Our customers tell us that electric vehicles remain the primary demand driver for SiC power devices for the time being. The shift to 800-volt platforms in EVs and adoption beyond traction inverters significantly increases the SiC content per vehicle. A main driver of this is the sharp reduction in substrate prices. Overall, the number of produced silicon carbide units is increasing. At the same time, industrial applications, grid infrastructure, and AI data centers expand the SiC application space beyond automotive. In AI data centers, a big part of the conversion from the high voltage of the power grid to 800-volt intermediate voltage is done with SiC devices. Here, a new class of very high voltage devices at 2 kilovolts, 3 kilovolts, and even higher voltages is in the making. Also, in the grid itself, a new class of devices is emerging. With an increasing affordability of SiC power devices, solid-state transformers, SSTs, have emerged as an alternative to traditional transformers. We closely work with customers on these applications. According to multiple customers, our Planetary G10-SiC system delivers outstanding results in this new class of SiC epi-layers. This makes us more optimistic on a gradual recovery of silicon carbide throughout 2027. Let me now come to MicroLED. Here, the demand remains soft, and the market has not yet reached broad volume deployment. However, we are seeing an increasing level of activity in the industry. These developments are taking place in two areas, in particular, augmented reality applications and optical interconnects for AI data centers. Nevertheless, we do not expect this market to become a major volume contributor in the immediate TAM. Nevertheless, the increasing level of activity is encouraging and could see future demand for MOCVD systems in the coming years. With this, let me now continue with our footprint expansion. Our new site in Malaysia will become an important element of our global manufacturing footprint and strengthen our manufacturing flexibility, supply chain, and long-term cost competitiveness. The project remains fully on track. The groundwork and piling has been fully completed, and we have almost finished the ground floor and expect to be able to use the building from Fall 2027 onwards. With that, let me now move to our guidance. We confirm our full year 2026 guidance as published in April. We expect full year revenues of EUR 560 million within a range of plus or minus EUR 30 million. We expect a gross margin of about 42% and an EBIT margin between 17% and 20%. The gross margin and EBIT margin guidance includes one-off expenses in the mid-single digit million euro range related to the personnel reduction in operations earlier this year. Let me also add a few words on our expectations for next year. Based on the current sales pipeline, AIXTRON expects to maintain or further grow the current strong level of order intake over the coming quarters, supported by demand for optoelectronics and increasingly by growing momentum in power electronics. This lets us take a very positive look on the prospects for 2027. As always, we will give you formal guidance for 2027 when we release our full year results on March 4, 2027. With that, I'll pass it back to Christian before we take questions.
Christian Ludwig: Thank you very much, Felix. Thank you very much, Christian. Operator, we will now take the questions, please.
Operator: Thank you very much. Ladies and gentlemen, if you would like to ask a question, please press star nine and pound key on your telephone keypad. If you would like to revoke your question, press star three and pound key. You can also use the dial-in function in the webcast and raise your hand if you would like to ask a question by phone. I repeat, to ask a question, please press star nine and pound key. We have a lot of questions listing out. I am just going to start from Adithya Metuku from HSBC. Please go ahead.
Adithya Metuku: Good afternoon, guys. Thank you for taking my questions. Firstly, Christian, you are talking about EUR 275 million plus in order intake over the coming quarters. If I annualize that, I am looking at roughly EUR 1.1 billion in revenue. When I think about 2027 revenues, Street is at EUR 800 million. That is clearly not particularly demanding. Is there any reason why you might not be able to do EUR 1.1 billion in revenue next year? If you do not think it is possible, then what could detract you from hitting those levels in revenue, I suppose? That is my first question, and I have got a follow-up.
Felix Grawert: Well, let me take this one. First of all, EUR 275 million in order intake is now what we have achieved in Q3. Please take into account that we are currently looking at lead times on the order of 12 months, plus or minus. Such EUR 275 million is rather translating than not starting from Q1, but rather from Q3. Please take the time shift into account. I think that is a very important element. Second one, again, we would not want to go now in the discussion of a precise guidance for 2027. You see we have given you some indications on the order pipeline, with the remarks that we see EUR 275 million, and potentially a bit higher as order intake for the coming quarters. This was not an indication toward the revenue. To the second part of your question, for the revenue, we have to monitor how we can ramp. As we have discussed in earlier earnings calls in the previous two quarters in this place, we have given you the indication that we are ramping up our supply chain. We are ramping up our production, means our own people. Also, we work very closely together with our suppliers. Many suppliers are helping us a lot. We have a great group of suppliers around us. This whole ecosystem of AIXTRON plus partners, production partners, is hiring people. Some of our suppliers have rented additional facilities for production. Some have bought additional machines for production. As we have indicated earlier, it is all about the ramping speed. We do not have any limit in terms of production capacity. This is unlike our customers who have a fab, and at some point, the clean room is full, then you need to build the next clean room. We always have indicated it is about the ramping speed, and I think we have done a good job over the year, Q1, Q2, Q3, and we have given some indication for Q4 to ramp up, and this will continue also for the next quarters. This ramping speed is really what limits the increase. Again, no formal guidance for 2027.
Adithya Metuku: Understood.
Felix Grawert: Yep, go ahead.
Adithya Metuku: Got it. Just as a follow-up, you are talking about power starting to recover. As far as I know, silicon carbide is largely G10 reactors. GaN, I think, is also more skewed to G10, especially if 300 millimeters is coming in, or at least, your newer reactor in the GaN space. I believe these should be more profitable than your older reactors, where the wafer sizes are smaller. If you do see power picking up into 2027, could that help your gross margins, and help profitability trends into 2027? Am I thinking along the right lines?
Felix Grawert: I think you think along the right lines. Let me shine a little bit of light on that. As you indicate, as you rightfully ask, the G10 generation is more innovative, has a higher productivity, has a greater value for our customers, and hence, of course, part of the value also stays with us in terms of higher margins. That is clear. That is correct. Please, nevertheless, also take into account that it depends a lot on the regional mix. We have some regions, Taiwan and China, who likes to continue using the older generations, G5+ and G4, G5+ especially also in the power electronics. The legacy products continue to run at a lower margin, just because it is older products, not yet on the high productivity. Those ones take a drag on the margin compared to the G10. But overall, looking in 2027, and looking at the drive for the market segment that you have rightfully indicated, a lot of the opto business on the G10, some Western power electronics customers expanding on the G10 generation, we can clearly expect an uptick in gross margin as well.
Adithya Metuku: Understood. Thank you.
Christian Ludwig: We take the next question, please.
Ruben Devos: Yes. Good afternoon.
Operator: Thank you very much. We can hear you, Ruben. Please go ahead.
Ruben Devos: Oh, okay. Great. All right. Okay. Yeah. Thank you. One question related to the 6-inch, I think you talked about 6-inch indium phosphide wafers, as one reason the opto market is bigger than you expected. I think in the previous call, you talked about 70%-80% of systems shipped in 4-inch. Has that share changed in the Q3 orders? At this stage, what is mainly holding customers back on 6-inch? Is it still the substrates, the yield? At what point would you expect maybe the larger customers to convert to 6-inch? Thank you.
Felix Grawert: Very good question, yes. It is funny, we actually looked at data for this one. Interestingly, for 2027, the share of 6-inch systems is going down. Normally, you would expect the more modern wafer size as time progresses, increases. In fact, in 2027, the 6-inch goes down from around 30% down to 25%. As you indicate with your question, this is in fact due to a shortage of substrates. We see even towards the orders that have been already placed or in discussion with customers for 2027, we see even 3-inch, very small, increasing and doubling from 15% to 30%. This is literally because the industry is ramping so fast, the substrates are short, and the customers literally take every wafer from every substrate that they can get a hold of because there is such a demand in the industry.
Ruben Devos: Okay. If, let us say, it seems far out still, but if they can convert eventually and they buy the new tools for 6-inch, sorry, would they actually buy the new tools for 6-inch, or would they reconfigure the installed G10, you think?
Felix Grawert: Customers can reconfigure the G10s, which they are now buying and use for three and 4-inch. They can reconfigure these tools for 6-inch. I think the real question is, customers have existing production lines. It is not only that you just buy some new wafers and convert a tool, but then you have to convert your entire production line, all the other tools in the line as well. Normally, those things do not go so fast. Look what we have seen in the recent years with silicon carbide. Everybody expected 200-millimeter wafers are coming, and they take volume around 2023, 2024. Now looking toward 2027, we gradually now see the switch towards 200-millimeter wafer, four years later. You may have also seen that we had, in the Q3, quite some sizable after-sales business, if you look at the bar chart in our presentation, because that big part of that was now in preparation of the next silicon carbide ramp, customers converting from 150 to 200 millimeters. I would say and think we've seen something that was expected for 2023, 2024, now finally happening for 2027. I have no reason to believe why on the optoelectronics this would go much, much faster now.
Ruben Devos: Okay, great. Just the second one, still staying on opto. Could you give us a sense of the mix between the EML lasers, and how much maybe for the continuous wave lasers? To what degree is demand already linked to 1.6T? Thank you.
Felix Grawert: I can unfortunately not, because our customers take the tools, and we make sure when we install the tools at our customer factories, that the tool delivers all the specifications and is capable to perform according to that. But then in the end, the customer closes the door when we leave, and the customer keeps their secrets, both on the recipes they run on the tool as well as on their product mix. Unfortunately, I have no transparency into that.
Ruben Devos: All right. Fair enough. Thank you.
Operator: Thank you very much. The next question is from Om Bakhda from Jefferies. Please go ahead.
Om Bakhda: Thanks for letting me on. I had two questions. Firstly, on pricing. If we take the EUR 275 million order intake, and you mentioned that you could maintain or grow that, it seems as though you could get towards the EUR 1 billion revenue mark in 2027, and that would take you to being fully utilized before Malaysia capacity comes online in 2028. Does that then change the pricing dynamic for 2027? Could we see some areas where there could be some opportunistic pricing? Have you already been in discussions with customers for pricing in the coming quarters?
Felix Grawert: We have a very clear strategy on the pricing. We maintain our price points. We neither give volume discounts as customers are now ordering very big quantities, nor are we pushing for excessive price increases. As your question is suggesting. Because we have seen that by keeping a stable price level, customers see us as a reliable supplier. It's not necessarily a negotiation up and down. It's a long-term partnership. This long-term partnership in the very long run is then beneficial because it also secures that when the reorders are coming, they kind of by default go to AIXTRON and it secures our long-term market position. We don't take, and I think this may be different from us than from customers who have a hard limit on fab capacity, right? I always say my customers, right, they have a clean room, and at some point the fab is full. Then, they have no possibility to extend. While on our side, we don't have such a hard limit. A stop of something, capacity full. We can always rent an additional building or some manufacturing space. So we are not limited by a capacity. A number of tools or certain upper revenue number in a year or in a quarter. It's rather, we have to manage and bring the supply chain up and running higher, enough people, make sure our suppliers are ramping up, and then kind of, it's just a timing question of how much we can then produce.
Om Bakhda: Thanks. That's very helpful. Just to follow up, coming back to the lead times that you mentioned earlier. I think previously you talked to sort of six to nine-month lead times, and now that's sort of increased to around 12 months. Is this predominantly because of supply chain related? If it is, could you give some colors to where there could be some constraints in areas of sensitivity? If there are 12-month lead times now, does that then mean that the orders that you'd get in Q1 and Q2 of next year, you wouldn't be able to deliver those tools in 2027?
Felix Grawert: That's a very good question. The limit I would say is overall the ramping of ourselves and our supply chain. I think it goes very well hand in hand. It's not that there is one thing short and like one component short or two components short is limiting us to grow. It's rather, we hire people, our suppliers are hiring people and then, we together run as fast as we can. So that's I would say. That gives us the quarter over quarter, how much every quarter can we increase. This is how you have to think about this. Now in terms of what orders translate into revenue in which quarter, it's not very straightforward to answer. It's not like, you take the order intake from one quarter, and you shift it by four columns in Excel to the right. Each column being one quarter. Because with some customers, we have capacity reservations. I spoke about the long-TAM partnerships. And they've slot reservations. So I think it's more the general picture. If you come as a new customer, not having made any reservations and want, let's say, five tools or want 10 tools, then I would say, "Well, the slot I can give you is four quarters into the future." But it does not necessarily mean that there might not be a shipment of a tool in January or like in April, 2027, where the order might not come, I don't know, in December. That could be the case if customers have made reservations. So it's not such a black and white picture, but it's just how the reality works.
Om Bakhda: Sorry, just following up on that, does that mean that essentially your prepayments become more important than your orders in terms of customers if they're willing to put down prepayments now, then that even though they're not translating into orders that could sort of be enough to then be ready for 2027? How should we sort of think about that order and prepayment relationship?
Felix Grawert: Oh, that's a whole mix. Unfortunately, it's a Sorry. Sorry for not being clear. I like to be more clear normally.
Om Bakhda: Yeah.
Felix Grawert: But it's literally a mix of models, right? There's models about capacity reservations. There's models about prepayment staggered over the order lifetime. Because many of these orders, as you can imagine, is not an order of three or five tools, but rather whatever a dozen of tools and then stretched over two or three quarters. And so on. So, it's a mix of different types of contracts and various shapes of grid.
Om Bakhda: Great. Thank you.
Operator: The next question is from Oliver Wong from Bank of America. Your line's open.
Oliver Wong: Hey, guys. Good afternoon. Thanks for taking my questions. First question, just wanted to ask a bit more on the capacity and revenue run rate. Right now it's around EUR 200 million a quarter. Should we expect, as you work on improving the supply chain, should we expect this number to kind of increase in a linear fashion throughout the year next year? I was wondering what kind of ambition you have in terms of, if things go well in terms of what you're doing with the supply chain, what kind of runway we could be looking at, let's say by the end of next year?
Felix Grawert: Oliver, you ask the billion-dollar question. Hello. I wish I could answer this already. I'm not at this stage ready to answer this. I think this is the key question. We are working with our suppliers. We are working internally right now. We work on the ramp. But it's a bit too early to give you a definite answer. We will come out with our guidance in March, when we have the full year results. By then we will have an answer. At this point in time, there are multiple factors. As you see, it's really operational topics. Hiring people, working together with our suppliers. In very close alignment, identifying which of the suppliers are the bottlenecks, what kind of limit the capacity rates. Then see, and I think we've done a pretty good job about wherever there was a bottleneck to eliminate the bottlenecks. In many cases, this was with one supplier, and we have literally together with that supplier eliminated the bottleneck, and the one supplier stayed one. In other cases, the supplier said, "Guys, we just can't be one." We are not unhappy if another second or third supplier is added. We need a bit more time as this goes deep into the operation to work out, and then by the time we give our guidance to come up with an answer. It has multi levels of facets below it. It's a bit more complex than the orders and the customer orders and the market dynamics.
Oliver Wong: Makes sense. That's helpful. Thank you. My other question is regarding SiC. We're pretty much at the bottom this year. You mentioned demand kind of picking up. Looking at historically, 2023 was, I believe, your last peak at around EUR 200 million. Before this year, you were at EUR 100 something million a year in SiC. I was just wondering, how fast do you think SiC demand could really ramp based on what you're discussing with your customers, et cetera?
Felix Grawert: Very good question. I see SiC coming slower than GaN. You see, I've also changed the order and put GaN upfront. With GaN, we clearly see now a ramp starting. We see the market very clear capacity being utilized in full. I think I've indicated in the prepared remarks. Silicon carbide, we see some customers ordering. We clearly see the utilization rates going up, also in our after-sales business. I've seen selected customers plan a new expansion, while I've seen other customers, I would say undecided, maybe because they got burned so hard. There was a Chapter 11 in the SiC market, if we recall just 2 years back. Very painful. I haven't seen the broad sentiment in the SiC yet where I would say that they go all in. I think my guess is SiC will be a bit of a softer recovery.
Oliver Wong: Okay. Got it. Makes sense. Thank you.
Operator: The next question is from Veysel Taze from Metzler. Please go ahead.
Veysel Taze: Yeah. Hi, thank you for taking my questions. Maybe the first one on Oliver Wong's question, I understand you cannot really give full visibility how your capacities will ramp up throughout the year. But nevertheless, asked differently, what keeps you away that first quarter will be not a similar revenue size in terms of capacities like Q4? Q4 will be probably around EUR 210 million-EUR 220 million, right?
Felix Grawert: I think your reasoning makes sense to say if we reach a certain capacity level, we should be able to maintain that capacity level also in subsequent quarters. I think that logic is right. Because in the end, as I've indicated, it's about hiring the people, getting things to run, working together with our suppliers, help them ramp the people. Sometimes we work with suppliers also helping them on operational processes. It goes very deep into operations, what we are doing. And when things are running and up and running, normally they don't fall back or collapse, right? It rather stays. I think your logic is reasonable.
Veysel Taze: Great. Then the second one, Felix, a little bit on opto technological understanding. There has been now a lot of rumors about VCSEL lasers making into the AI scale-up networking. I think few vendors have pushed forward their timelines, particularly for up to 15 meters as a sweet spot for VCSEL laser adoption. VCSEL is a gallium arsenide product, and you are quite strong in gallium arsenide. But nevertheless, VCSEL has a much more mature supply chain, bigger wafers. I think the epi complexity is not as high as with EML lasers. Can you give a little bit color if VCSEL takes off a little bit stronger in scale-up in 2027, but particularly 2028? What would that mean for your TAM or for your SAM expectations?
Felix Grawert: We are involved in a number of VCSEL projects for the data communication. We see the activities that we have seen so far focusing mostly on very short-range activities. Sometimes literally on a card. Nevertheless, the impression I have is those topics are still under the development. So in a completely different maturity stage than what we've seen on the indium phosphide, more R&D and exploratory and focusing on a complementary sub-segment of the stack. This is how I have seen it from my technical team. In other words, I would rather translate that it may turn a top point in an upside for us, rather than in a threat or a replacement of the indium phosphide-based market. This is how I see it.
Veysel Taze: Great. If I may, can squeeze in one more for Christian Danninger on his comment for Q gross margin. I think you said that you don't see the 48% gross margin for Q3, not an indication for Q4. I was wondering why that should not be the case. I mean, at the end of the day, your order backlog, the mix is still towards opto. If you now have in Q4 a little bit GaN shipments on top, why should not gross margin stay at the same level in Q4?
Christian Danninger: Yeah. We wanted to make that clear that you do not draw wrong conclusions. We have a good view on what we will ship. That is why we also highlighted it was in Q3, we had a very good product and regional mix. So the combination was just very good. In Q4, we will have a different regional mix and also different product mix, more older product generation, G4, G5, then also more China mix in. We do not know exactly how it will come out, but it will come out lower than the Q3.
Veysel Taze: Thank you very much.
Felix Grawert: Welcome.
Operator: Thank you. The next question is from Gustav Froberg from Berenberg.
Gustav Fröberg: Good afternoon, everyone. Thank you for taking mine also. Just two, please. First, a question around the momentum comments that you made, that you expect more momentum going forward, either at the same level or even a little bit higher. What gives you this comfort? Or have you received letters of intent? Are there shadow orders? Is there anything more concrete to point to, or is this more just based on customer commentary and your general feeling from discussions with customers? Then the second question is on opto-
Felix Grawert: First one.
Gustav Fröberg: Oh, I can do the first one first. That's fine. I'll pause.
Felix Grawert: Yeah. Very short. It's discussions with customers, pipeline preparations, quotation activity. So no paperwork in the stack.
Gustav Fröberg: Okay, great. Thank you. On opto, I was just wondering if you dare, rather, give a new assessment for addressable demand of optoelectronics tools at this stage, and maybe some words on what new assumptions you would make around that addressable demand. I know it's exceeded expectations so far this year, so just curious what your sort of new estimates or assumptions would be around sustainable opto demand.
Felix Grawert: Yeah. That's a very good question. I keep asking my marketing team, and I keep not trusting the numbers anymore. Honest answer. Our first estimate was incorrect. As you have rightfully spotted, as we have indicated. The demand is currently much larger than we had initially indicated. Also, we see that market analysts and predictions are putting up their demand forecast. The numbers I've quoted earlier in my prepared remarks, plus 40% from what they had estimated. I had many discussions just in the last three, four weeks, I was around SEMICON West meeting many of my customers. The honest answer is nobody really My customers also don't really know. So I refrain from giving out a new number. It would not be serious. What I can tell you, nevertheless, is that customers give a very clear indication and a very strong confidence. They see the demands of tools that they order from us backed by solid orders from their customers. So it's not just to build out some capacity and then see what happens. What we've seen earlier years in silicon carbide. I think everybody was there just following a dream, but in the end, the dream didn't materialize. I hear from my customers that what they order at us is backed by solid order from their end customers. So I hear a very strong confidence of my customers, that the pipeline they are placing out, and the forecast they are giving us is solid and increasing. This is what we have reflected in our remark, that we see the strong order momentum that we had now in Q3 continuing also in the next, I don't know, two, three quarters.
Gustav Fröberg: Perfect. Thank you very much.
Operator: The next question is from Craig McDowell from JPMorgan. The floor is yours.
Craig McDowell: Good afternoon. Thanks for taking my question. My first one was on gross margin. I wanted to ask how we should be thinking about 2027. From the order book that's been building through this year, it seems like heavily weighted to G10, opto, all for delivery in 2027. The revenue mix seems like if we're delivering in 2027, should be similar to what you've reported in Q3, and indeed at larger volumes. The question is, should we use Q3 2026 gross margin as a good proxy for 2027 gross margin? I've got a follow-up. Thanks.
Felix Grawert: That's a difficult one. It's really too early to give a precise guidance here. We will give a precise guidance when we publish our full year results. But overall, I would be a little bit careful in just taking the Q3 gross margin as a run rate throughout the next year. As I explained, that was really a constellation of stars. A very good product mix, very good regional mix, but you cannot expect that to continue into the next year. But overall, of course, the higher volume will help with some operating leverage in the COGS area, also in the gross margin and overall the mix with the opto and some GaN coming in should naturally give a few percentage points over this year's gross margin into the next year. But we cannot be more precise at this point in time. I think, Christian, this makes sense.
Christian Danninger: Let me build on that one. I think, this time of the year, everybody looks into the next year. I want to give also one more indication for next year just to make sure that, I think you all see that we don't know yet. We give you as much transparency as we can, as we had it. To add to this one, I think all of us clearly see 2027 will be clearly ahead of 2026. It's a no-brainer. Also we will use this good financial situation in 2027 for another push on the product generation with also a significant uptick on the R&D. If you guys start modeling, a very significant uptick on the R&D because we see we have a good experience now with the G10 generation. It's in the field for quite some time. Customers have experience. We spent some time to really make it mature and get all the topics done that customers wanted over the last years, and quite successfully, I would say. Now next year will be the one to take on the next generation product. So we will use the financial situation also then for a good uptick on the R&D. We will not just run it flat. It would mean all the additional gross profit would fall down to the EBIT, but also we are planning to really up the R&D rate again to prepare for further growth, to secure our market position, because now with good feedback from many customers. We've collected their wishes, what they want for the next generation product, and then, of course, we'll work hard on getting the things done. Make sure what, I don't know, 2029, 2030, or whenever it is the next generation comes and, it stays as healthy and as strong for us in the market as we are today.
Craig McDowell: Very helpful. Thank you. My follow-up was on GaN. We've seen Infineon this week announce a large dedicated 300 mil GaN facility. Can you comment on your position here? When do you expect a decision on whether you or Veeco will tool this facility? Maybe just your degree of confidence on winning that position there. Thank you.
Felix Grawert: Yeah. On 300 millimeter GaN, first of all, we see our initial hypothesis that we have shared also with the market. I think we have somewhere in the backup of our investor presentation, even a slide on that since several quarters. We see that there will be a coexistence of 200 millimeter GaN and 300 millimeter GaN. We said from the beginning that we expect the customers who have large 200 millimeter fab capacity today, silicon power, 200 millimeter fab capacity will stay on 200 millimeter. This is what we see. We see that the customers who have large 300 millimeter fab capacity, silicon power devices, MOSFET, IGBT, and so on and so forth, that those guys who have existing 300 millimeter, they will gradually convert the existing fabs towards 300 millimeter GaN. I think here we have a first evidence now that this is just happening. So our hypothesis of the coexistence of the two wafer sizes is first of all confirmed. Secondly, as you know, we never comment on individual customers. Also not on this one. What I can tell you is that we're preparing for a sizable double-digit number of tools, next year on 300 millimeter. So our 300 millimeter will go into very large volume next year. Not only from one customer, but from more than one customer.
Craig McDowell: Excellent. Very good to hear. Thank you.
Operator: Thank you. The next question is from Martin Jungfleisch from BNP Paribas. The floor is yours.
Martin Jungfleisch: Yeah. Hi, good afternoon. I just have two quick follow-ups. The first one is really on the order comment you made in the beginning. Would you expect this potential step up that you were talking about in the orders in the next couple of quarters being more driven by Power Semi? Is it more by opto, or should we assume this 65% share opto, 25% share Power to be the norm for the next, I would say, one or two quarters?
Felix Grawert: I would expect on the first shot that this mix that we have seen will continue. So a good level of optoelectronics, clearly the dominant one. Nevertheless, the Power also continuing to be a healthy portion here.
Martin Jungfleisch: Okay, that's helpful. Thank you. The other question is really just on the use of cash. I mean, obviously you have, sitting on a quite substantial amount of cash with the cash generation year to date, also the convertibles adding to that. Can you just give an update really what your preference is? Would a buyback make sense here? Maybe if you can share some plans, some potential smaller bolt-ons and M&A, that would be helpful. Thank you.
Christian Danninger: Yeah. Somehow expected that question. Let me put that a little bit in perspective. Yeah. The overall cash position, of course, is getting quite high. But please consider also the high number of contract liabilities for advanced payments, where we've received that money from customers, but we've not earned it yet. We need to first get the equipment deliver. I cannot do a share buyback based on that. Yeah. Also, with increased liability, of course, from the convertible bond, if you take all of that out, if you take the non-current liabilities out and the contract liabilities for advanced payments, well, then you end up at around the EUR 275 million, which is still very good. Yeah. But we are conservative in our financial policy and are not taking share buyback now. It's too early. Yeah. But we promise you we will make good use of that cash to generate value for the future.
Martin Jungfleisch: Sounds good. Thank you.
Operator: The next question is from Martin Marandon-Carlhian from Oddo BHF. Please go ahead.
Martin Marandon-Carlhian: Hi. Thanks for taking my question. My first one is on photonics. How should we think about the respective contribution of pluggable transceivers and co-packaged optics to the current order pipeline? It seems like the inflection of CPO volumes is for NVIDIA Feynman GPUs in 2028. I was wondering if these orders are more expected next year for co-packaged optics high volume, or do you see already the current order intake addressing that trend for larger CPO volumes?
Felix Grawert: I expect that the capacity, as mentioned, right? We do not know in the very details what our customers ultimately do because the machines are very versatile and can be used for multiple use cases. From what I infer from customer discussions, this is all the preparations for what we see for the Rubin Ultra and later on for the Feynman. This is really the preparation for these optical architectures, which then have the benefit of connecting many more GPU racks with each other and further building up the power of the AI. So it is really in preparation of that ramp, what NVIDIA is driving with their architectures, and together with their partners.
Martin Marandon-Carlhian: Okay, very clear. The second one is on power. You said momentum in GaN and SiC has improved. Do you consider Q3 order intake in power, so I guess it was around EUR 60 million or so, to be already the sign of a recovery? Do you actually see more meaningful orders in Q4 and the following quarters for power? How much of these orders do you think are directly related to GaN adoption in data centers?
Felix Grawert: That's a very good question. I expect the GaN and the SiC, and I don't have the detailed breakdown here, to be in about the same proportion as we have seen it in the Q3. We expect this to stay on a good and healthy level or to further increase. I can clearly comment on the gallium nitride. For the GaN, I attribute the biggest momentum or the biggest part of these orders clearly to the low and medium voltage part, and the AI power build-out. We clearly hear that from the customers. 650 volt, which was the initial wave, around consumer devices also is adding a little bit, but the majority of additional capacity goes into the lower voltage devices. That all goes into the AI build-out.
Martin Marandon-Carlhian: Okay. Thank you very much.
Operator: The last question goes to Adithya Metuku from HSBC. Please go ahead.
Adithya Metuku: Thank you for letting me on, guys. Two additional questions. Firstly, just one of your peers recently talked about a group 4 material being used in photodetectors and them winning epitaxy layers for that step. I believe they were talking about germanium, so I just wanted to see what you're seeing in terms of use of germanium in photodetectors. Could you also talk a bit about your positioning there? Secondly, I also wondered if you could talk a bit about how you see MicroLED slow and wide adoption happening in data centers. Any clarity you can give around that would be great. Thank you.
Felix Grawert: Yeah, very happy to. Germanium as a photodetector material is widespread and very well established. Yeah. That's kind of a well-known material. We are not addressing that. This is not our focus. I think it makes a lot of sense. What I mean is, it's the right band gap. It's well-proven. It's very nice, compatible with the silicon amplification circuits, that low noise amplifiers that you build right behind it, in terms of silicon photonics and integrated optics. This is the way to go. This is all what I can say from my general physics knowledge. As I said, we are not in this market. Regarding the MicroLEDs for the AI, it goes a bit in the same direction as we had earlier in this call, a question regarding the VCSELs. There's quite some heavy work on this research work. We are working together with multiple customers on that one. Again, it would be complementary and an additional market segment to the indium phosphide, what we see. What customers of ours are playing with, or what the vision is, let's put it like this, is to have gallium nitride MicroLEDs. With gallium nitride, we are involved. Gallium nitride MicroLEDs bonded to either the GPU or to the CPU, and then to use, for example, a silicon carbide or a sapphire interposer on which you place both the GPU die, the dies for high bandwidth memory. The vision is to make also the interconnect between the GPU and the high bandwidth memory an optical interconnect. Today, in all the architectures we've been discussing in the last 45 minutes, we spoke about copper interconnects between the GPU and the high bandwidth memory. But again, the same limitations hold, which is why they get for the longer haul or within rack or within the data center replaced by indium phosphide. The same bandwidth limitations hold. The dream would be to make the interconnect between the memory and the GPU also an optical interconnect. You could do that with gallium nitride MicroLEDs, making little MicroLEDs, and then kind of contact print them on the GPU and the backside of the high bandwidth memory and put them on a substrate into which you write some waveguides, and then have optical connects between the GPU and the memory. That would be the vision, and there is some work ongoing.
Operator: Great.
Christian Ludwig: Operator, if there are no more questions, then we would close the call, please.
Operator: We actually have Ruben Devos on the line again with a follow-up question from Kepler Cheuvreux. Ruben, please go ahead.
Ruben Devos: Yeah. Thanks for the final question. I just had one on Penang. I think you are adding 25%-30% capacity by 2028. Is that mainly for Opto? How could the cost structure change once that site is up and running? I think you are also referring to that you could potentially expand beyond that initial build-out. Of course, curious when that decision could happen. I mean, looks like certainly customers are committing based upon the sort of customer advances, the lead time stretching beyond 12 months, so your visibility gets much better. So yeah, curious on your thoughts here on Penang.
Felix Grawert: Let me try. Very good question. In fact, we are currently, as we built the plant, we built for the full building, we built the shell. That means the outer part of the building, the walls and the roofs, kind of, and we prepare everything for the infrastructure. While from the possible clean room space in the stage one, we are only equipping about one-third with clean room space, because as always, when you build clean room, the expensive part is not the building, the shell. But the expensive part is all the ventilation, the clean room systems and the technical apparatus that you put inside of the building. This is the reason why in stage one, we only fill one-third of the building. While, of course, it is economical, the heavy construction ongoing to make the shell and the roof and the kind of stuff, to do that in one shot. So that means, in other words, as we only equip one-third, it is clear you could multiply the additional capacity that we build out there by a factor of three, stage two and stage three also. We estimate that each phase will take us six, seven months to build, so relatively fast, because it is just inside of the building, the technical infrastructure, so we could respond and react quite fast to do that. We see that for the starting point, we are well with those additional capacity that we are creating here. In terms of the answer regarding the cost structure, which was also part of your question, it is a bit too early to tell. Because we are just building the plant now. We are ramping our team. We are learning about the supply chains in Southeast Asia. We get quite some good indications. We are quite happy with our decision that we made on the strategic direction, but it's too early to put that in quantitative numbers because we are just starting to engage with the local supplier ecosystem. Before I can give you a quantitative number, I think I need to have 60%, 70%, 80% of my parts, and get prices and quotes on them, such that then at that stage, I would feel well to extrapolate and share the numbers with you.
Ruben Devos: Okay, great. Very briefly, I think in the presentations historically, you talked about the silicon carbide could be doubling the midterm tool demand, and then for GaN, three times. Those figures aren't in today's slide deck, I think, but should we read anything into that, or do those figures still stand?
Felix Grawert: No, I think for the midterm, that all still holds. We've tried in this version of the slide deck, you've seen it, you caught it right. We've updated the detailed slide on the GaN and the SiC. We thought it's more important now to shine some light, as we've tried also in this call, about the near TAM. Means the trend that we see for 2027, rather than the complete timeframe extending towards also 2028, 2029, 2030. So we really have focused here with the update of the slides just on the next year to start giving an indication what we see just now happen.
Ruben Devos: Okay, great. Thanks very much.
Operator: That was it for today. I would like to thank everyone for your participation. With that, would like to hand over to your host, Christian Ludwig, for the closing remarks.
Christian Ludwig: Thank you very much. Thank you all for the excellent questions. Hope to see you before end of the year at one of the conference or road show. If not, go back and talk to you at the latest on March 4, when we will publish our full-year results for 2026, and then also have a more formal guidance for next year. Have a great day. Goodbye to you all.