← 返回

2026.09.30 | TB | Power | Infineon: Powering AI Server Megawatt Racks

原标题：Infineon – Powering the AI Server Grid & High-voltage Megawatt Racks

#power #semis #datacenter

Third Bridge · 2026-09-30

Analyst-led • 23 Sep 2026

Companies: IFX DE, NVDA US, TXN US

Paul Arts (PA) Head, Research Unit Integrated Photonics Technologies at Silicon Austria Labs GmbH (SAL)

Saad Maqsood (SM) — Third Bridge Sector Analyst

Agenda

Key Insights

Introduction

SM [00:00:02]: Welcome to Third Bridge's Interview entitled Infineon – Powering the AI Server Grid & High-voltage Megawatt Racks. I'm Saad, and today, we have with us Paul Arts, former Senior Director of Product Development and Industrial ICs [integrated circuits] at Infineon.

Just before we get started on the call today, could you please state I agree or I disagree to the following statement: You understand the definition of material non-public information and agree not to disclose any such information, or any other information which is confidential, during this Interview.

PA [00:00:42]: I agree.

SM [00:00:43]: Could you start us off with an introduction to your background, please?

PA [00:00:49]: Yes. I'm Paul Arts. I have been working for Infineon for eight years. Now since one year, I'm leading Photonics Technology team for Research. Before that, I was eight years in Infineon, leading for four years, the IC development for industrial applications, of which, of course, power supplies for data centres was one of them. This was mainly on the high-voltage IC part like gate drivers and controllers for high voltage. That matches well with the topic of today. Yes, before that, I have also been working in supercomputers, developing supercomputers. Also from that domain, I have quite good knowledge of the application. I think also there, we were developing the first liquid-cooled supercomputers getting up to a few hundred kilowatt per rack already, which one of the customers there was Nvidia.

Questions

00:01:57] Infineon has been benefiting from the AI boom. The company's AI server business is growing tremendously, achieving EUR 700m at the end of FY25. What is Infineon's positioning in this value chain relating to the AI boom? Who are the company's potential customers?

PA [00:02:45]: Infineon has always been delivering components to the power supplies for servers. In the recent years, they also started really targeting the new development in AI data centres, delivering point-of-load devices, IBCs as a system and of course, still also delivering components. Their strength is mainly in power devices, which is then also for data centre more and more moving from MOSFETs or IGBTs to more advanced switches like silicon carbide. This business was brought into focus roughly five years ago and where they made a plan to ramp up, and that plan is actually going very great for them, even exceeding expectations, I would say.

00:03:44] Infineon talked about being well-positioned in grids, core and beyond. Across those stages, where do you think the company is the strongest?

PA [00:04:04]: They are good in power conversion in general, regardless of what it is, but also in the past, they have dominant positions in solar converters, in chargers, etc. That market, the higher the power, the more unique their proposals are. They are usually leading by technology and especially where the technology is moving fast and power density becomes a problem, they have an easy way to excel. Their drawback has often been speed and also the way to dive into these topics. The other one was they stayed usually away from systems with all that for data centres, they now have really grabbed with two hands, and they have been able to really turn this around, not only creating a very good business for their MOSFETs and for the medium- and low-voltage part, but also more for high voltage and of course, also for the complete systems where they are then taking the position that they are not completely taking over the market if they have partners that are delivering similar parts, but they are even selling their topology (inaudible 05.29) patents to their partners and/or producing these devices in-house themselves.

00:05:35] How would you break down the EUR 700m growth in Infineon's AI server business in FY25 across the areas the company plays in or the products it provides in each area? What is driving most of the revenue today?

PA [00:06:13]: This is for them, mainly the switching business, so their MOSFETs. Component business more than the system business still. Most of these systems are new or they are currently in development and especially now where they take over the complete the power supply for Nvidia GPUs and also where they do similar things with other competitors, they are really ramping up fast. In the past, they were mainly delivering components, like I said before, to power supplies, so like the AC to 12V supplies. Now they also delivered a lot of components to their to their IBCs, so they are integrating these bus converters. Also, they started making some point of load. For the point of load, I think that's still not too big in '25, but that will be definitely driving a lot of it now in '26 and beyond. In percentages, I don't want to say. I think that's too much of sensitive.

00:07:37] Most of Infineon's revenue is still centred around the first- and second-stage bus converters, MOSFET [metal-oxide-semiconductor field-effect transistor] switches, etc. Might the company get more involved in the core stage or start realising more revenue from the core stage with the point-of-load converters in 2027?

PA [00:08:09]: I think the AC to DC converter business, I am not so very sure that this is seen as data centre revenues, but this has been always, especially with these high energy AC to DC converters. They have been quite active in that as component supply. They have had long relations with a few of these world-leading power supply suppliers. I'm not so very sure if the first stage, most probably if you consider this point of load or you're counting from the other side. The medium voltage, so let's say, now also these IBCs, that definitely makes a bit of sense. I think actually, from the past, their main was into these AC/DC power supply units. This is, of course, then not for these GPU servers, which are now driving the AI or hardly any of it. Most of these racks, they have centralised power distribution now most of them are 50V or 48V.

00:09:35] You mentioned that Infineon has historically stuck to the component side, but now, the company is shifting towards the full system side. Do you think Infineon is ready to undertake that transition, or are there changes that need to be brought in before addressing that? How is the company going about that shift?

PA [00:10:06]: They build up a quite big R&D force now in the US to be close to their customers. That they have done quite aggressively. They also have taken over some managers from their competitor in the US. From TI to name one. They are definitely buying themselves into this with competence and growth close to the customer. The other one is indeed that it has been defined as a separate target market for them. Infineon does this more often now also, for example, for humanoids, and then they target this with a long vision view on how to develop this. Data centres was then started up by one of the fellows actually here in Austria in Villach, and he has worked out a plan.

He was doing research with a small team like 10-15 people to really tackle this to bring up the bright ideas and then to start executing on it with a very clear business focus and buying into this with very strong sales, let's say, they don't really have sales accounts like for Nvidia in the past, but they now have really target this market and also having a special sales force for this with high skilled business unit manager or product line managers that are then in constant contact with customers in this field. Yes, they have reorganised it, they do it differently to make this a target area and to do R&D and to develop the technology. It has been done before. Like I said, that has been definitely paid out for them in the data centres.

00:12:09] Does Infineon's shift from products to systems imply something about the margins it can command or the BOM [bill of materials] per server rack it can get? Is it naturally higher, lower or the same?

PA [00:12:37]: First systems, these are point of load, so direct DC-DC converters. Where a component is a switch just or an IC, they are now packaging this together, so that you can make converters, these are sold. The thing there is that the speed in which this has to be developed is very high. To do this with delivering components and having somebody else doing the system is lengthening time to market a lot. They have been very successful in shortening the time to market for new developments that are required for the next generation like, for example, GPUs. This speed, they could keep up where… With that one, they took over many of these big key accounts. That was not driven by margin per se. Now since they have now such a dominant position for AI, they are definitely also cashing in on that and are finally able to really drive up their margins, although for power switches, they have been always leading by technology, and that has in also many other occasions allowed them to have very high margins. For this market, I think the market… The margins are driven up by a few things. They are able to introduce their new technology and drive it to volume much faster. That is the main contribution to healthy margins. The return on their investments is much shorter. Of course, you're selling the new instead of the old.

00:14:47] In terms of Infineon's transition from products or components to systems, what might be the impact on customer stickiness?

PA [00:15:14]: The stickiness, yes, that's very relative because you have actually with servers every three years, you have to completely redesign your whole architecture and with that also your power supply. Actually, this market by such is not very sticky. You have to continuously innovate and continuously develop. If you are looking into delivering into these customers, so not delivering directly to the server manufacturers, but to deliver this to the power supply manufacturers, then this is becoming very sticky because indeed, there, they have to, of course, besides delivering these systems themselves, they also deliver IP and the components. Of course, with that one, they are trying to keep their customer base and help… Basically to divide the tremendous increase in production also between Infineon and then their customers, which are then the ones that are delivering these power supply systems, these subsystems that are then assembled on the server boards. Actually, the stickiness towards the end customer like Nvidia or AMD, yes, that is just as long as you can keep up with their speed. If you miss one or you have problems in the delivery chain, they will definitely not be so loyal to you. This is actually new for them, and that's also why they have what you asked before this reorganisation. They have really tried to reorganise themselves to deal with this.

00:17:15] Where do you think Infineon's go-to-market approach might be better suited? Is the company better off going after merchant GPU suppliers such as Nvidia, or can it equally go after the hyperscalers that are going after their own ASICs [application-specific integrated circuits], etc? Is there a customer category that Infineon is better-suited for, or do you think the company can cater to anyone?

PA [00:17:58]: They are definitely asked to take more. Of course, let's take Nvidia. They would love that Infineon would also use their capabilities to produce other parts. That from Infineon perspective, they are usually quite hesitant. They keep themselves to their core business, and they are not looking into developing other parts that are a bit outside the core business to satisfy these big customers. The same maybe even for automotive, if you take Volkswagen or any of their other, they were always a bit like sticking to the components they had and not developing a complete new set of chips or set of devices for this customer if it was not part of their core business. They will fit, like I said, the servers, they also want to have as little layers between them and Infineon because of this speed and of course, getting technically advanced products as fast as they can. They also want to team up directly with them. Then the board manufacturing, yes, so the server builders or if you look at the people that are building these systems, yes, that's more complex, so that market is completely changed.

If Nvidia wants to build a complete rack and servers and cables and everything on it, I think they are doing this because they don't see any other way to keep up with the development speeds and they can control over the market like this. If that's really that they did this out of business perspective that I don't know. I think with Infineon, they also are keeping up with the speed of Nvidia for what they are now already doing this power supply, then of course, they could also make controllers, so other components for the server board and that's going on and that will definitely also… They will increase their chipset. To really start doing something that they don't have in their portfolio to develop for them, that's like let's take optical links to talk about my new business, that they are hesitant and that you see like others. That might well be that they will start doing or at least taking that in the coming years. For now, they focus on the things they are already doing. It's a mixed answer.

00:21:09] Infineon increased its guidance, targeting EUR 1.6bn in full-year FY26 for the company's data centre power segment. Do you see any risks to meeting that target? Do you see any opportunities to go beyond that target?

PA [00:21:54]: First, Infineon works from September till September, so we're at the end almost. We're talking about this EUR 1.5bn that they definitely will reach in my view, but I have no evidence for that. I think they even will surpass that. Then the growth to the future, they will adjust that upwards, I'm pretty sure. They are going to also reorganise even further to even get more focus, and that will even help them to do more and more business in this domain and even speed up. Maybe we'll also talk about it, but high voltage is, of course, their core domain has always been and now with high voltage supplied to the megawatt racks that will even increase their chipset much more and especially on these high voltages, there's hardly any competition. The margins are huge for them. That will definitely help them to even get more and more successful. I would not be surprised if next year will be a doubling again. Like I said, I have no evidence. I don't talk about that with my ex-colleagues anymore about numbers.

00:23:07] Infineon's Q3 FY26 earnings mentioned that the company is planning for a material upgrade to its FY27 AI revenue estimates when it officially releases its full-year FY26 results in November 2026. Do you think that's more related to demand visibility? Why didn't Infineon upgrade it in Q3 FY26, rather than waiting until November 2026, knowing that there will be an upgrade?

PA [00:23:57]: No, they do the yearly forecast, and they keep to themselves. They are German. They will keep themselves to procedures. They have this and that they will do. This is normal procedure. They would not see any reason to do this intermediately. I think everything you can already plan out from meeting the investors to publishing the figures, to setting the forecast, to upwards adjusted, etc, etc. Yes, you can see it maybe more in the… They have a few factories that will come online now. Then you, of course, also know that silicon carbide is one of their new babies. This is really now finally ramping very good. They have capabilities of producing this on 300mm wafers. They have their facilities in Asia running. They are now because I think very soon, this will be capacity limited. Their sales will be capacity limited, but they are quite well suited to make a big jump in their volumes that they can produce. Especially if this is then silicon carbide because we will… I don't know if you will have separate questions on this high voltage to 10kV delivery into 1MW of racks for Nvidia that will definitely need all these new technologies that I was just talking about. Of course, ramping these, yes, like I said, there's hardly anybody that can really do that. There are only a few, maybe ST, but Onsemi and maybe some others, they are not really capable of doing these kind of things, I guess.

00:26:06] Nvidia said that its 800 VDC [volts direct current] data centre architecture is intended to support 1MW IT racks and beyond from 2027. Is that shift a tailwind for Infineon? Is the company ready to tackle that?

PA [00:26:32]: Yes. They always have been delivering high voltage. That's also where their technology pays off. The lower the voltage, more competition you will get also from China or any other… Mainly from cheap labour countries. Where you do high voltage, then the processing and the technology becomes very much important than the process control. That's definitely their strength. Like I said, that's then on the power switches where they will have much more is then on the IC development to speed that up and to keep track with that. Then especially if, in the future, you will get more to smart switches, so you will build in intelligence into these switches. That kind of things, they will have definitely to ramp their gears and to change something in the way they do these kind of projects. If it comes to… Because 800V is one thing, but also the other thing is that you need then to deliver 10kV into your rack. You cannot do your 400V AC lines for the US or even the three-phase AC into your racks, but you will do directly high voltage, medium voltage into your rack and then convert it at the rack from these high-voltage lines to the 800V distributed voltage.

That needs then, of course, isolation. That comes then back together to all these patents that they have now on the new topologies for IBCs. They have established a unit for IBCs for these bus converters already now two years ago, and they are already working on this. They are ready for this. In conclusion, it's not only the 800V, but it especially to get the power into the rack that is very interesting for them. Then you have to see also like on these EV chargers for cars, which also have the same, and they also go to roughly 1MW. They also have exactly the same. They would like to come in with a high-voltage AC and then convert it to in your… Also 800V for your battery of your electric car. Already, they have a lot of these technologies. Also there, they have a quite dominant position. Of course, they can translate these components and technologies directly to the data centre market.

SM [00:29:41]: If that does kick off in 2027, should that result in a margin uplift for Infineon?

PA [00:29:56]: Yes. The higher the voltage, the higher margins, I would say.

SM [00:30:03]: You mentioned STMicroelectronics and Onsemi as competitors here.

PA [00:30:14]: I would say, like I was already saying, the IC might be something that might be their weak point. There, they have, of course, also Texas Instruments. Texas Instruments is also especially on low voltage, they are quite strong. I see them as more competing, but of course, Texas Instruments doesn't have switches. They don't have silicon carbide switches nor MOSFET production. High voltage is then something they can keep quite easily, but to keep the whole chipset might be then under threat from the IC part. For the ICs, then you also get other components like hot-swap controllers, etc. That is definitely one. Yes, ST, as I already said, they are quite strong as well in their technology and also they have silicon carbide. It's not yet completely sure if that will be massively entering the data centres. In my view, that's definitely going to create quite some wind in the sails for ST. Also there, I think for lately from the chipsets that Infineon can deliver and especially when it comes to driving these high-power switches, so the high-voltage ICs and also the microprocessor, so the small microprocessor business that Infineon has that I think that ST will not be able to beat.

Then Onsemi, I think they were before Infineon, the ones that were the main suppliers for Nvidia. I think they are now quite out. I don't know exactly why I didn't analyse that. I don't think they will be coming back, although the market is now so big that even for these guys that don't have a dominant market share, they will still have quite healthy business for server boards, the same as also for some others like Vishay and maybe even also the Würth, so you have also magnetic component, passive component suppliers that also take active components or they even source them from Infineon and then they will build the systems. That's one thing. It might be that these companies catch up. The same is for the power supply. Of course, these guys that were developing the AC/DC supplies, they still want to be in business somewhere. They will also, together with Infineon or with ST or maybe somebody else take part of this market again. I don't think they will have themselves beaten so easily.

00:33:23] What do you think is the biggest risk to Infineon or its AI or power systems business? Is it high targets or accelerating growth? Do you see any risks to the company achieving its targets? Is it market- or Infineon-related?

PA [00:33:57]: No, I think the opposite. I think there will be a supply shortage, and that will call in competition and especially from low-risk countries, China has been not really able to take a dominant position in many of these big growth areas from power supply point of view. They definitely will be able to catch up when there starts to be shortages in the market. That, I think, is the biggest moment of threat for Infineon. There will be production capacity limited and also maybe the investments and not doing enough investments to keep up with the growth to take risks. That might be another one that might, in the end, bite them in the tail. Then the market could collapse, but I don't see this happening. I know that you cannot buy any machine to produce chips to be delivered in the coming two years. For the coming two years, we are safe. Everything else is a glass ball. Then indeed, so you also see that even if the markets completely collapse, like automotive, they are quite resilient as a company, have proven to be quite resilient against that and maybe even taking a lot of market share. If we compare to their competition like NXP and ST, they have outgrown them by far now, just exactly in the time where maybe it was more difficult, of course, partly by acquisition, that's to be said as well.

00:36:01] Infineon's management talked about the demand outstripping the supply. Do you think this is more related to the company's capacity and ability to get capacity online in time, or is it more related to the supply chain side in terms of what it can secure? When do you think supply might meet up with demand? What would be the primary factor?

PA [00:36:40]: I think the supply is mainly front-end production, so to get enough wafers through your fab. The other definitely risk point for them is the back end, so the packaging, which they… Actually the whole world does in China, and that's, of course, politically also a difficult one. The other one is the raw material, silicon carbide is okay. If now soon we also need gallium nitride, gallium is, of course, in shortage. That's also something to look out for in the years to come. They are ramping up, of course, gallium nitride, but that is percentage-wise of sales, it's nothing. Of course, that will change in the coming five years. Now how that plays out, yes, that… In the supply, like I said, you need to do big investments in factories, and they have done now. They definitely improved, especially in the last years, Infineon has massively invested in more capacity in production. Of course, they have also done some acquisition, but their weak point there is that they always have their own technology not produced in any foundry. Also that makes them vulnerable for this because in times where their own capacity is lacking, it's not so easy for them to partner up. I think they will try to cover that risk in the coming years. They have done that in the past as well, but then the political situation was different.

I don't know now how they want to do that. Of course, China would be an easy one. That might be very sensitive because, of course, their investments have been a lot also sponsored by Europe. Of course, the job market in Germany is screaming for more jobs with Volkswagen and Bosch and all the others firing a lot of people. I think it will be for them politically very difficult to outsource a lot of work to outside Germany to basically cover this problem. Then it's more likely that they would drive up the investments. I didn't read anything about that lately. I know the Dresden investments that's in the East Germany is now finally coming online. I would then expect that they immediately would go for the next plan, but that I don't see happening. They are just already happy that they got this investment that they did in the last 3-4 years starting to pay off. They are very risk avoidant also because the cost of underutilisation has definitely been a headache for them in the years before the CEO took the seat. This CEO has changed a lot of things, but one of them is that, of course, in the downturn, you are not going to be loss giving anymore.

00:40:28] Infineon's Smart Power Fab in Dresden was officially opened in July 2026. In terms of filling it out and accelerating the ramp-up there, the original plan included a 3-7-year capacity fill, but now, management is suggesting that the fab can be fully equipped in under three years due to the sheer allocation demand. Does that seem realistic?

PA [00:41:08]: They have tried it on a smaller scale in Villach, where we did EUR 1.5bn or EUR 2bn invest. That worked out fine. They even took a technology lead already in that, so the 300mm production that I was talking about before. They have done that now, but that was definitely not in their genes before. I am too far away now to see how Dresden is really coming. They also wanted to scale up there in their IC technology. Like I already stated a few times, I think there, there might be a bit slow. They have still ramped up or doubled, tripled, whatever, they have magnified the production capacity for their current technology. Of course, they should have worked also on the next one already. They do. Of course, the factory should be ready to ramp immediately with new technology as well. I don't know how that will play out. Definitely, they have the right people in Dresden, also technology knowledge. There is almost unlimited availability now of resources, but there is a big competition coming, and that's the TSMC factory that they are building together with NXP and Bosch and TSMC also in Dresden. In that area, there will be also maybe a shortage of personnel.

00:42:48] From a capacity perspective, how easy or difficult would it be for Infineon to convert any more of its legacy lines and reallocate capacity from standard automotive IGBTs [insulated-gate bipolar transistors] – which aren't in strong demand right now – and repurpose that for AI-dedicated power stages and MOSFETs? Do you think that's feasible, or would it take too much effort?

PA [00:43:31]: They have already done that. That's not a problem. The only thing is there for these very high power and very high voltage switches, they were usually making them for trains, etc. That's far from mass manufacturing. That might be more difficult. To make a switch for 2kV or whatever you need, that's usually done for a train and then delivered with almost one per box. This kind of production lines, they have already since years. I don't think they are used to the speed of data centres. That will be challenging for them to convert that part of the business, but that has not started yet. That comes with this 800V rack supply for 10kV line. For the ICs, I think they have not to do massive migration. I think they are too slow in adapting to the next generation of their technology in any case. They have also started some, especially for their microcontrollers, they already do a lot of foundry supply. I think that they have much better under control.

The IGBT business, the other one that's very important is the scale up to 300mm. With that one, you quadruple your capacity in your factories. In the same factory, you really have a big increase in your volume and that they are really leading. That is a technology problem for these big power switches like silicon carbide even. That one, like I said, they were suffering with a very slow ramp of their silicon carbide, but they have used that time to scale up and also, of course, to bring their cost down. One of the things they do is this getting more wafers out of their raw material and that one also came online. They can get factors more of wafers for silicon carbide out of one ingot piece of silicon carbide delivered to them. Also that one. In the raw material for silicon carbide, I think they are quite well sorted. That will change, of course, when this step has to be made to GaN. That's not only a step in the material, but also in the way to process that. I don't know that I cannot foresee. That might give them difficulties if there now will be a massive adoption to GaN in the data centres.

00:46:37] Infineon's Q3 FY26 earnings call mentioned that there might have been an acceleration in GaN [gallium nitride] adoption in stage 2 that's closest to the XPU. If that accelerates, would it increase the gallium shortage? Is that a risk that Infineon needs to be wary of?

PA [00:47:01]: Yes, they are. Gallium is not like the problem with rare earth and motors. It's definitely quite a complex supply chain. They are definitely also doing technology research. Yes, this problem cannot be solved very easily for them. If it comes to the link to China, Siemens, which is, of course, Infineon times before, has always had a good footprint in China. Their relation with the Chinese was actually quite good and is still quite good there. I have been there for many, many years already. Maybe that will help them. Yes, there are other suppliers of gallium, so that's supply chain. I think Europe will try to tackle. There's a lot of support from Europe to get control of the supply chain, especially for new materials. Yes, that one, I think Infineon will just join the European initiatives if that would be required.

00:48:46] What places Infineon in a strong position vs competitors such as STMicroelectronics and Texas Instruments? What is Infineon's differentiator? What puts the company at the top of the market? Is it the breadth of its offerings or portfolio that can cover many different areas, or is it the high-voltage side? What stands out to you?

PA [00:49:27]: There are a few things. Their strength, I think, really plays in this. That's already their technology development, especially when the power density is getting very high. They also really reorganised the way they do things. They were starting on the return on their investments on R&D just to shorten the time to market for their R&D and to, of course, also have R&D focusing on cost of the products. That has helped a lot. They have completely reorganised the way they do the business side, so the marketing and sales. They have moved a lot of people from R&D into technical marketing, so technical support for customers. That has helped them to be very fast in following the GPU market. They had always the technology leadership. They have reduced their resources in research, and they have moved them into technical marketing, putting them local. You see also building up in India for other markets. You see the building up of R&D and resources in America. You will see it also in other parts coming up to be close to the customer and putting people with very high skills that we were doing before research on power devices or on ICs now supporting customers directly.

If you look to the competition, like, of course, the invest into new devices, especially in the… Silicon carbide has been in development by ST and by Infineon for 20-25 years. That's why the two of them are now ready for it and hardly anybody is. Then for the MOSFET, they always had this technology leadership with their vertical CoolMOS. Like I said, that already was always that with a smaller device, you can deliver more power. This was, of course, but for a more expensive price. Now this becomes a requirement to have more and more concentration of power. Then, of course, this plays in their advantage. Where China mainly was making less efficient, bigger chips, but for a cheaper price, you could just choose for your application just to put a bigger device in it, and that would do the job in any case. For servers, this is not possible anymore. There, you need to have that high power concentration.

Then like I said, for intelligence and ICs, you need to be also good in understanding your customer. They do this now with moving their guys close to the customer, but definitely, that was not their strength, and I still think it's not. Also there, if you look to TI, how they do it, they are much better in that. That's also why I think, especially in that direction, where you have to add intelligence, where you have to work on adding intelligence in the form of ICs or even integrating it in the switch. That's definitely something where the competition might have a chance to break back quite a big part of the chipset.

00:53:37] What market share do you think Infineon has in the AI data centre market? Is it north of 50%, higher or lower?

PA [00:54:14]: No, I think that will not happen. I think the big players will not allow that or they will not like that. For some components, they will definitely have a too dominant position. Definitely, I think there will be actions to keep the market share of Infineon under control. I think it's becoming too dominant already for most of them. How much in percentage that I don't know or I cannot tell. If you take the whole market and you have EUR 2.5bn, then you know already that it's 25% or so. That kind of things. I don't think that's the whole story here because, of course, it then depends on what you count under this. They need Infineon to drive the innovation. Then you will see that, of course, Nvidia will not sell only their latest generation. I think for the other generations, the market share of Infineon will not be too high. I think they will keep on focusing on keeping pace with the latest development, and then they will give up or give away their old generations to mainly then Chinese. In total, I don't think that that market share will really increase too much, but I think they will more and more get to high-end devices. The market share not, but their margins definitely will have a positive upswing. These are already very healthy for this domain. Not still as Nvidia themselves, so maybe.

00:56:43] Is there anything else that you'd like to discuss around Infineon or the broader AI data centre market?

PA [00:57:04]: Yes. Very quickly, you need microprocessors that is, I don't know… Of course, Infineon is, in volume of microprocessors, the biggest supplier at the moment in the world. This is going great for them. You need all these small microprocessors to add intelligence to this. I think that one might be a nice thing to analyse. As I already said, I think the big guys will also force Infineon to go into other things like indeed, I hope for optical links, for example, or for the other components that you need in the data centre. That might be also an interesting one to watch how they will play with that. Exciting times. Indeed, the customers are now forcing Infineon to change. That's also nice to see.

SM [00:58:15]: Perfect. I think that would be a good note to end on. Let me close by saying thank you so much for your time today. I really enjoyed that conversation and it was incredibly helpful. Yes, it was a pleasure speaking with you. I hope you enjoyed it, too.

PA [00:58:30]: Thank you. Thanks.

Transcription ends at 00:58:37 of the recorded material