Peter and I, we're very much looking forward to today's session. We'd like to show some ambition around the areas of Infineon powering AI, and where we wish to become a strong growth driver for Infineon. O n top of that, really, the introduction of gallium nitride, where we believe is a disruptive technology and reinforcing our global leadership in power systems. Peter will be talking about the enablement of green energy, as well as the leader in silicon carbide modules. T his is a pictorial view of a lot of us live in today and some of the applications that Infineon are very proud in servicing.
Here, later on, where you see the gallium nitride, I will provide a little bit more insight in how GaN is supporting mobile charger, as well as topics around onboard charger and as well as even data centers. Let me get going now for the Power & Sensor Systems. Let's start really talking about the phenomenal exponential growth in global data, especially by the year 2030, with predictions of a further 30% CAGR growth by the year 2030. There's been already a give or take, a 145-fold increase of data since the year 2010 to the year 2025, the expectations. Of course, data itself is driven through the areas of AI, ADAS, for example, autonomous vehicles, AI factories, among many of those areas.
Now, by the year 2030, we're saying here that it could be growing exponentially to 600 ZB worth of data that will be generated by us. That is equivalent with roughly around 600 smartphones at 120 GB per person on the planet. T hat is just to represent the amount of data that is being generated today. The other area s over 95% of the data is actually stored without active use. A s we know now, of course, with data, especially with AI, with training, then hopefully that data will be able to advance us further moving forward, especially in the area of generative AI. Now, let me just walk through a little bit more about how Infineon is very proud about powering AI today, and more to the point, moving forward.
The average Infineon BOM per AI server is approximately $850-$1,800. Let me explain in a moment why the broad range of BOM, so when I break down a typical server rack. H ere what we've used as a data point is one server rack, which will include 4 AI servers. F or each AI server, we actually have a GPU tray of 8 AI cards. Now, those AI cards will include GPUs or TPUs, and here the range of the bill of material will, for us, go from $50-$200. I'm sure you're probably asking the question, why such a broad scope of BOM?
It depends upon whether you are doing a down solution, where you can use more industry-standard power stages, which is less advancement in technology, or whether you actually go for better power efficiency, which will can be actually attained by using vertical power modules, which I'll introduce in a moment. In each of the AI cards also, there will need to be a DC to DC converter, where today the industry has standardized more or less on the 48 V to 12 V, where again, Infineon has content. On top of that, we have the server motherboard, and here depends upon the architecture, but again, approximately $50 of content of Infineon technology available to us.
Then we come down to the power supplies, and as we talked about earlier on, a lot of the new servers are coming out, and they're a lot more hungry to obviously for the power levels to compute the amount of data that is needed, especially with the GPU and TPU technologies, which I'll explain about the wattage in a moment. The power supplies here, again, we have a range between $20 and $100. I n the power supplies today, we have silicon power supplies, we have silicon carbide power supplies, as well as gallium nitride. A gain, it depends on whether they're using the power ranges of 3.2 kW or even up to 5.2 kW. We'll obviously highlight on which the USD content we will be able to support.
This is really for us a breakdown of the bill of materials. As you can see, it is rather compelling, and now let me just go through why we believe we'll be powering AI. Firs t of all, AI enables obviously the system demand, because ultimately the higher power that of course will need an increase of semiconductor content. Down the bottom, you've got what basically the release date and on the vertical axis, you've also got the power requirements. H ere for the server CPUs, you can see that goes to roughly 500 W, and then today in the GPU and the TPU, you can go up to 1000 W.
Especially now for the AI training, you can see here the unit CAGR growth of 40%, as well as the BOM CAGR growth of greater than 10%. W e are very happy to, to be announcing that we are, of course, working and we've got design wins with a number of, leaders in the GPU and TPU, and we can talk a little bit more about how we are offering those solutions, shortly. From the market model, what we're seeing at the moment, we see around 218 K servers, today for AI-enabled, which are then really for training purposes, I must stress. T hese are really, translates to around 1.7 million cards.
This year that we see that that number could go to around 500k which goes to approximately 4 million cards. We've seen some other data that is even more bullish than that. B y the year 2027, we see numbers around 850k servers, and then we will see, of course, that translates into, give or take, 8 million or 7 million cards moving forward. N ow the actual topologies, as it were. O f course, we have the 48 V architecture, and now this, combined with the vertical power solution, delivers, we believe, the best-in-class total cost of ownership. T his is very important because ultimately, the more power that we need to compute, the more electricity that will be generated.
We believe Infineon, of course, with the decarbonization vision that we have, we want to support a new green power for servers, and this is how we are looking to do it. Today, you have the traditional lateral flow. This is where you have the DC to DC Converter, the 48 V to 12 V. You have the power stage from the 12 V down to core, and then, of course, that is laterally put next to the AI or effectively the GPU or the TPU. Now, just to put it into context, where you say the power for one of these GPUs or TPUs are 1000 W, say you take it down to the core voltage, let's just use for simple sake 0.5 V. That is then 2000 A on this board.
That is a lot of power, and then, of course, you have to look at the power losses. O ur system architects have been working with the GPU vendors and the TPU vendors, and then we came up with supporting the industry, the vertical power, and this is where we actually put the power stages under the GPU and TPU. W hy is that so important? Two major dimensions. First of all, the power loss. It brings the power loss significantly down. Every single millimeter makes a major difference on the power losses, and we believe that actually there's a 12% total cost of ownership saving compared to the lateral power delivery in an AI topology. Then, of course, it increases the power density. W hy is that so important?
Because if you make it smaller, as you can see here with the vertical power, you can then enable further increase in compute power, so then you can fit effectively more compute power into a rack. Now, Infineon also is very proud because we've obviously been having a legacy within servers. We've been supporting many of the processor companies and continue to support the processor companies in CPUs and advanced CPUs, as now, of course, in the AI. We have an abundance of technology in the power semiconductors, in silicon domain, in silicon carbide for the power supplies, as well as in gallium nitride topologies, and we're also proud to have the analog and mixed signal capability. On top of that, advanced packaging is critical in these applications. We have chip embedding, for example, in the power stage.
This is some very advanced technologies that gives us significant savings versus some of our nearest competitors in the same footprint. On top of that, the v ertical power module. In saying that, we also need to have the integrated magnetics, where we can have the magnetics close there to the power stage that obviously benefits the system performance. System architects as well, we have a novel 48 V to 12 V solution for the converter topologies. We have then again, this vertical power delivery, and again, we are working with the magnetics designs to make sure we optimize the performance. We also are very proud to be talking about the digital capabilities of our controllers. We have this, what we call Hyper Transient digital controller technologies, which is very well received in the industry.
We also do Hot Swap, eFuse controllers, and we also have the point of load capabilities there as well. A s you can see here, we have a rich ecosystems of not just hardware, but also software, including the latest bleeding-edge packaging technology. W hat does that translate into? W e, of course, are stating that we are powering AI moving forward with the Infineon Technologies for the server business. H ere's just some examples of six design wins that we've actually been working with the customers. Here, for example, on cloud providers or of course, with the CPU or GPU manufacturers itself. In FY 2024, we have low triple-digit million dollar amounts, and we say that there's a revenue CAGR of greater than 50%.
In simple terms, this will be greater than EUR 1 billion of revenue in years to come. W e're very excited about these opportunities. Let me move over to now gallium nitride. W e're very happy to be announcing here that we've now completed the acquisition of GaN Systems on October 24, 2023. F or those of you who have already been watching our progress, we genuinely believe that GaN has this disruptive nature of performance. We wish to make sure that we become one of the leaders in gallium nitride-based solutions. H ere now, reinforcing our leadership in GaN, we'll be addressing the fast-growing applications, which are, of course, highly complementary to the strengths. We now will have a new IP that comes into the Infineon family.
On top of that, amazing application understanding with our combined teams and further customer access and project pipelines, especially in the areas of automotive, onboard charging and DC to DC, which I will talk about in one moment. W e can now really look at the roadmap acceleration, and of course, we have unmatched R&D resourcing and application expertise and know-how. Now, this really complements what we wish to do within Infineon. We wish, and we want to be, and we are a leader in power systems, but we need to do that through mastery of all relevant power technologies. We're very proud to be walking to the customers where we have the silicon-based solutions, silicon carbide solutions, as well as gallium nitride, and in some applications, we actually have the ability to do a hybrid of those technologies.
We can really listen to the customers, and we can then offer the best solution for each of the customer. W hat does this mean in tangible numbers? F irst of all, what we've decided to do is actually put all of the GaN activities into one dedicated business line. We want to make sure we do that to give them the entrepreneurial ownership, to make sure that we can be fast and agile, to listen to the customers and respond to the customers, and learn from the different application use cases. We will now have an incredible 350 patent families of GaN IP, and on top of that, we will actually increase our family of products by a factor of 2, and today we are publicly talking about 100 V- 650 V.
We also have what we believe the world's best GaN experts now really serving the 2,000 active customers that we are working with, and they are around 450 GaN experts under this dedicated business line. O n top of that, we now have dual sites for in-house production, but for 200 mm. W e also, thanks to the acquisition of GaN Systems, we will have a strong foundry partnership opportunities there as well. O n the right-hand side, look, at the end of the day, we have a pipeline which is ever-growing, which it currently stands at EUR 3 billion, and we feel that, now there's a tipping point for the industry to really accept, learn how to use gallium nitride, because it's not a drop-in replacement like silicon carbide. W hen customers lean into the technology, understand the technology, you then unleash the capability of GaN.
The market will grow. Okay, these are the public, sort of resource information here from Yole. O f course, we see different adoptions of the technologies by application. F or time purposes, let me just jump to three examples I'd like to really stress. The first one is the mobile charger. A s you can see here, that, we believe by using gallium nitride, you can actually get 2x less size and weight within the charger, and also increase the power capabilities of the charger. Here is just one example of a 100 W charger. I'm not sure if you can see that on the camera. A 100 W charger here, and here that there is obviously significant bill of content from Infineon, not just on the gallium nitride, but also in the controllers, which are silicon-based and other primary side as well as secondary side.
GaN really has this ability here for this charger, can actually charge three devices at the same time, and this is thanks to new topologies, hybrid flyback topologies, that Infineon has been working with, with the industry. The second example is the switch mode power supply. I n front of me here, again, on the picture and also in front of me, this is a 2,000 W silicon design that goes into the data centers or into the racks that I was just highlighting earlier on. T hen moving forward. Thank you, Peter. Moving forward, you can see now this is 3,000 W, so this is a 100 W per inch cubed of power density. Y ou can see this is actually based on gallium nitride. H ere you increase the power, and on top of that, you can see. Thank you, Peter. Doing a good job.
You can see the size difference, and again, what I mentioned earlier on, by reducing that size, then the rack vendors and the suppliers can then put more compute power into the same rack, but on top of that, saving energy and ultimately for total cost of ownership as well as decarbonization. The last point here is onboard charger. Here there is a huge content for us. Today, we're very happy to be working with a number of onboard charger companies. Ultimately, there are a number of designs going from silicon to silicon carbide, but we are also now working on Gallium Nitride solutions for the onboard charger.
Ultimately, we see a keen interest of taking the DC to DC and embedding it into one solution, including the onboard charger, which of course will reduce the size and the weight for those applications. A gain, as you can see down the bottom, a lot of BOM content, a lot of customers, which for us is very, very important 'cause it's a learning curve, and we're very motivated and excited in translating that 3 billion design funnel into revenue moving forward. W ith that, that concludes, and we'll take to Q&A, and now I hand over to Peter.
Yeah, thanks a lot, Adam, for this very insightful presentation of what's going on at PSS. In the end, it's about ultimately enabling our customers to bringing down costs on the system side, either in reducing CapEx, meaning the investments needed for the systems and/or reducing the OpEx. I think the same story holds for GIP, where we want to be the undisputed master in the area of green industrial power systems, while our focus is clearly on the berry-colored area regarding driver ICs and, of course, power switches. Regarding the power switches, we do not exclude any material coming, of course, from our stronghold, namely silicon, IGBT, diodes. Of course, we address silicon carbide and gallium nitride. You will see in a few moments, a very similar example as the nice thing shown by Adam.
What I would like you to remind from this presentation is that there are basically now three things which are instrumental for GIP's success looking forward. The growth from the market has accelerated. While GIP was used to single-digit growing markets, predominantly coming from the industrial automation and home appliance space, now, thanks to the decarbonization, renewable energy, and power infrastructure, including EV charging, this opens up huge additional market, markets with growth momentum above 20% CAGR for the years to come. If you do the math, as you can see on the right-hand side in the graph, that accounts to more than 10% market growth for GIP looking forward. In addition to that growing market, inside this, there's specific peculiarity regarding penetration of silicon carbide.
S ilicon carbide additionally fuels this growth because silicon carbide captures additional value from the bill of material from the systems, since silicon carbide provides additional performance for selected applications. T his additional growth, plus silicon carbide, drives profitability for GIP. We consider SiC as a differentiating point, which helps us to increase profitability, and SiC is margin accretive for GIP as we speak. W hat drives this growth? H ere you see a nice example regarding the trend of electricity costs, sources here, investment bank Lazard, for the different sources of how you generate electricity. I t's a bit of a complicated graph, but I would like to draw your attention to the lowest cost lines of electricity generation.
That is the levelized cost of electricity over time, and you see the lower ones being yellow and green, namely solar panels and wind, which already today achieve the lowest cost of electricity generation, according to this source. If we believe that this is true, then we should also see that the investment into these sources, namely solar and wind, outperforms investment into other sources. We try to find some proof of this, and here you see the numbers from S&P Global Market Intelligence, where you see in this pie chart the electricity generation additions in the U.S. in last year. They total to close to 25 GW overall, and you see here the split.
No surprise, and very much in line with the left chart, you see that the lion's share was taken by solar, close to 38%, followed by wind, 27%, yellow and green s o dark green. That means these both amount to roughly 2/3 of overall electricity generation additions in the U.S. last year, followed then by gas, the light green, coming in around 26% and 9% for the others. T hat's a nice proof that the left-hand graph is not only ambition, but has already turned into reality because the investment of the utilities into these new sources of energy generation move on with light speed. T his is an impressive trend, and we are somehow proud that we supported our customers to support this cost reduction regarding green energy production.
You see here the example how innovation enabled us to reduce system size and hence, of course, also cost for the, in this case, PV-generated electricity. T he story starts in 2008. On the left-hand side, this kind of system had a weight of more than a ton and provided 100 kW output power. I f you do the ratio, then you come up with 19 W / kg at an efficiency around 97%. The subsequent progress in power density, now here calculated in watts per kg, with the relentless innovation on the IGBT side, brought the system then over the 8 years to something like 700+ W/kg . Now, see what happened when the customers, 3 years ago, introduced our silicon carbide-based solutions.
From already 700+ W/kg , the density tripled to something around 2,100 W/kg. Note that already also the efficiency increased to more than 99%, and with the recent high performance, high power system, we talk about a 350 W/kg string inverter, which comes up with around 3,000 W/kg, having also an efficiency around 99%. I think that is a very impressive example how, with help of innovation, we are able to bring costs down for renewable generated electricity. T hat now nicely, of course, adds to the picture Adam has shown for the big in the beginning for gallium nitride.
The energy flow from generation via transmission to consumption is, of course, being enabled by silicon, gallium nitride, and now also silicon carbide is sneaking in a variety of applications. Here I would like to highlight that a couple of applications are predominantly suited for the use of silicon carbide, like PV inverters, like EV charging, whilst the others, for example, industrial drives, due to the cost pressure and the cost competitiveness of the incumbent IGBTs solutions, do not see this kind of development. I t's a big advantage for us, being the market leader in power, that we are able to provide all these various material, high-end silicon, silicon carbide, and gallium nitride, because in all these applications, there they find a reasonable playing field. Now let's have a look, again, from the GIP industrial perspective, how we have developed our share. Starting with silicon.
On the discrete side, we nicely were able to keep the share, even slightly increase it. Also, on a high level, we lost a bit of a share, but remaining the dominant market leader also in the area of IGBT modules in the middle, while of course, head-to-head also being on top for gate driver ICs. That's the IGBT perspective. Now, if we have a look into the numbers for silicon carbide, then you see that, of course, still STMicroelectronics has done a very good job leading the pack regarding also their close engagement in the automotive area. I nfineon now comes in already as a second.
I would like to draw your attention to the graph in the middle, where now for silicon carbide-based modules, we massively gained share in the recent two years, namely doubling our market share, having this strong portfolio, that we, of course, inherited from silicon times and that we now can leverage also based on silicon carbide devices. You might be a bit irritated regarding the silicon carbide discrete share loss on the right-hand side, and there's a simple explanation. Since we are limited on the supply side, we could sell quite some more. We decided to allocate the volume into that area where we can differentiate and generate more profitability, and that is, as we speak, the area of SiC modules. The picture on the right-hand side is simply limited by supply constraints.
With that, I would like to sum up a bit the most compelling silicon carbide offering that we have. We decided not to invest into silicon carbide wafer manufacturing, but as we speak, we have more than five qualified silicon carbide wafer and boule suppliers all around the globe, where we feel very comfortable with and have also quite competitive pricing. We invested into Cold Split technology, which enables very efficient splitting of silicon carbide boules. We meanwhile also licensed this technology to one of our silicon carbide material suppliers. The beauty of this technology is simply to reduce the losses while splitting the wafer, having much, much less kerf loss compared to the traditional wire saw.
In addition, we started our 200 mm conversion project, where we are targeting a qualified silicon carbide technology in two years from now. F or 2025, we assume that we will be able to manufacture silicon carbide starting on 200 mm wafer diameters. T he superior trench technology is a topic that we are proud of because it enables more chips per wafer, meaning a very good cost position, together with unmatched reliability and very good DPM rates, which I'm gonna comment separately in a second. The thing which enabled the massive increase regarding module market share is, of course, our packaging portfolio and the differentiation of technology we can provide, like, for example, .XT technology. This, combined with deep system understanding, provides additional service and perceived value to our customers.
Now, regarding the reliability and performance of silicon carbide in general and the trench technology versus planar, there has been quite a debate for some time, and I think it's now time also based on solid statistics to disclose here a bit the DPM rate. W e're proud that so far in the industrial space, we never caused any spill or severe incident on the customer side. B y now doing the math and gathering the data, the silicon carbide numbers are here based off more than 23 million SiC power switches sold into the market, discretes and modules. A s you can see, we even beat here the DPM rate of silicon. Of course, the absolute number is here not shown. It's well below 100 DPM that I am able to share with you.
Of course, the overall statistics on silicon side is much higher than on silicon carbide. N evertheless, all the precautionary measures we have, we took to prevent any issues regarding silicon carbide reliability topics to slip to the customer has been a big success, as these numbers nicely show. With that, a brief outlook on how we see our revenue develop looking forward. In the already ended fiscal year, we achieved EUR 100 million +, so very strong growth continues. F or GIP, we were able to grow with a CAGR of 50% for the last five years. F or the foreseeable future, we believe that we continue this growth rates. That would bring us to more than EUR 1 billion revenue potential in the next two years.
While this revenue so far has been generated of our brownfield side, converting existing silicon capacity in Villach towards silicon carbide. Of course, cost is of utmost importance looking forward. Therefore, the big chunk of the additional revenue potential, which is up to EUR 7 billion end of the decade, will now come out of Kulim. Kulim, as you can see, module three is under construction and will be ready for manufacturing end of next year. W e will expect first wafers coming out end of next year, maybe beginning 2025, and that will help us to drive significantly additional revenue looking forward, while of course, then serving the market with a very competitive cost position. Ending my short overview and presentation, I just would like to share with you that there's much, much more potential in the market.
That is based on the IEA net zero scenario, the additions that are expected from the market and from the applications until end of this decade. H uge numbers on the generation side, dominated by PV, followed by wind, then tremendous investments that we need to undertake on the infrastructure side. Grid network alone, EUR 600 billion annual investments. That's a huge number, but already in the past years, past decades, typical investments were in the range of EUR 300 billion globally. T hese numbers now will double and grid storage plus EV charging, and then, of course, also for decarbonized heavy industries, electrolyzers, are the topics that we foresee now kicking in for the second half on this decade on the infrastructure side, followed by consumption, heat pump, H2 fuel cell, and e-aviation, e-marine, looking forward.
With that, I think we're pretty well in time, and we jump to the Q&A.
Okay.
Okay, great. Thank you so much for that very informative presentation, Adam and Peter. N ow to the Q&A. Anyone on the webcast wish to ask a question, please use the raise your hand function. M aybe just to kick things off, question on my side, maybe just for you, Adam. You spoke very positively on AI and the drivers of the PSS business. Just wondering how you should reconcile that with the current down by single digit outlook for fiscal 2024. M aybe same question to you, Peter, as well, for the GIP business. Very positive for some of the renewable stuff, but again, flat outlook for GIP for FY 2024. How should investors think about those two dynamics to reconcile that?
Yeah. T hank you very much, David. L et me just take the question related to servers and general servers. O f course, there has been at beginning or the end of last year, as well as the beginning of this year, a little bit of a correction on the inventory for generic servers for the non-AI-enabled. We believe that there's still a CAGR growth of servers between 6%-8% growth, but of course, the significant growth will come from AI. W e are projecting at the moment, low triple-digit revenue for this year, and with then the 50% CAGR growth up until FY 2029. That will obviously result into 1 billion dollars or euros of revenue in the next years to come. Peter?
Yeah. Looking into 2024, of course, it's much more cloudier compared to 2023 because all the big areas driving our business growth, like, the renewables, but also like the industrial space, which is still determining something like 30% of GIP revenue did well. What we now clearly see, that is, the renewable space is cooling down. Coming from a very high level, we still expect growth, while definitely we expect declining a negative market environment inside on the industrial side. T hat, of course, also is reflected by increasing inventories at our customers. Since we, I would say, have a flexible manufacturing setup and also took precautionary measures on the supply side, our inventory levels are very healthy. That is very positive.
While we have to take into account that this is not the case everywhere on our customer side, we are careful and still expect a single-digit growth for 2024.
Okay, next question, please.
Joshua, you can see.
Hey, guys, this is Josh Buchalter from TD Cowen. Can you hear me okay?
Yes, we can.
Great.
Okay, great. T hank you for the informative presentation. I wanted to ask about the mix in silicon carbide device market. I f we compare silicon carbide to IGBTs, modules are currently a much smaller portion of the mix, but one where Infineon is overrepresented. Can you maybe help us better understand why the market thus far has overindexed to discretes, and how you expect modules versus discretes to trend over the next couple of years? You know, is the overindexing to discretes more because of company-specific decisions from early adopters in silicon carbide or something about technology maturity, and how should we expect that to trend over time? Thank you.
We would, I would love to do more on silicon carbide discretes as well, but, as, as said before, we had to take a decision how to allocate the volume. S ince we saw also from a margin perspective and from the aspect of differentiation, much more, arguments to allocate the volume towards modules, we decided to go this way. O f course, now having, the capacity and the output of manufacturing increase, looking forward, significantly increase, including kicking in then Malaysia. End of next year, we, definitely want, of course, also to gain market share on, on the discrete side. Because also the .XT technology is not limited only for the modules, but we also apply it on on the discretes.
That means with silicon carbide and the ever increased power density, the challenge for the customer and the system is to get the heat out of the device and out of the system. H ere, . XT technology provides an advantage not only for modules but also for discretes. Our customers love this performance gains because so far, this kind of feature, I consider us as a sole source. Looking forward, definitely much more volume growth is expected to be also on the discrete side for the industrial space.
Thank you, Joshua. Next question, please.
Sorry. Alexander, it's your turn.
Alexander, can you hear us?
Hello, Alexandre Peterc.
We can't hear you.
Okay, let's try the next question then, please.
Yeah. Then, Jonathan, please give us your question, and we try Alexander later on again.
Sure, yes.
Just a second. We have to unmute you. Thank you for your patience online. Just give us one moment. We're just having some technical issues in unmuting. Just one moment, please. Okay.
Jonathan, now it should work. Give it a try, please. I think you have to unmute yourself.
Yeah. Yeah. Hi, can you hear me?
Yes, we can. Thank you.
Hi there. Thanks, thanks for taking my question. Two of them, both actually on the gallium nitride side. Y ou gave us your views on the SMPS power supply and the onboard charger for gallium nitride. I n past presentations, you've also talked of smaller sort of rooftop solar panel inverters as a big driver of gallium nitride revenue. I was just wondering why you omitted that in this presentation and what the outlook there is. M y second question is also on the onboard charger side. You've said you've got these 25 wins on the onboard charger, and seems to be that they're starting to ship from 2025.
How do you see the onboard charger share splitting over the next five years between silicon carbide and gallium nitride? Would it move quite fast to gallium nitride given the advantages of gallium nitride over silicon carbide, or would it be both will coexist for some time? Thanks.
Okay, great. L et me take the second one first on the onboard charger. W e believe, and we're modeling at the moment, that actually the both technologies will coexist in the silicon carbide solutions, as well as in the GaN based solution. Why do we say that? In some designs and OEMs and tier ones, clearly, weight and density is the key factor, and that's where we are seeing now the traction on the GaN solution. I n other cases, some of the architects and the designs are very happy with the silicon-based solution. W e believe that both will coexist. O f course, in some of the architectures, we're actually also seeing now the DC to DC block for the Aux also being GaN-based, which can also then be housed into even a smaller, more dense environment.
Back on the solar. Look, honestly, we have a lot of applications now showing of interest in GaN. That does not neglect anything that we previously said in solar. We do, and we are engaged with a number of optimizers who are making optimizers for the residential solar, for example, where they clearly see the benefits in gallium nitride for a variety of value propositions in their use case. W e do see that adoption happening. Of course, we see a little bit of a slowdown at the moment in overarching residential. W e are still believing in the fundamental drivers of the build-out of residential solar. T he answer is yes, we just physically couldn't put it into the presentation, which is good news because we have a lot of wonderful examples of now customers really seeing the benefits of gallium nitride. Thank you for the question. Sorry, Peter.
No, no, all fine. I think it's fair to say, Adam, right, that the order of appearance for the onboard charger is start with silicon-
Yep
-then silicon carbide, now gallium nitride.
That is correct.
As you said, right, it will coexist down the road, but what we now see ramping and introduction is still based on silicon and silicon carbide. Next thing upcoming then, in addition, gallium nitride.
That is correct. Okay, next question, please. Hi.
Hi. Hi . Can you hear me all right now? Is it working?
Yes.
Wonderful.
Yes.
Thanks so much. I just have a question on your capacity in silicon carbide. H ow much of your Villach plant is currently dedicated to silicon carbide versus silicon? D o you still have any more capacity to grow in that location?
Yes. B asically, we have the ability to continue to support our growth with kind of 40% CAGR. T he thing of Villach is the following: of course, we are now converting the legacy area and the legacy manufacturing, which used to run and still partly is running on 6-in silicon, right? There are legacy technologies on 6-in that we are transferring or obsoleting as fast as we can, either transferring to 8-in, 12-in, and those are then being dedicated to silicon carbide until end of next year. G rowth for Villach will support our overall revenue growth until end of next year, when then finally Kulim will kick in. Villach will remain very important for variety of small volume technologies, technological variants of silicon carbide and of course, for R&D, right?
That will continue, is, and will continue to be our key R&D hub, while we then transfer to Kulim and ramp the volume in Kulim. O f course, not to forget, we of course also ramp and have a huge footprint in Villach on silicon 12-in, but this does not affect the strategy on wide bandgap.
Okay. W ith that, the next question from Sebastian. In a second, you are unmuted. Now, okay, that should work now. Hello, Sebastian.
Hi, everyone. One question on silicon carbide and the visibility you have for 2024. One of your competitor in the U.S. made a warning on silicon carbide, partly due some specific issue with their leading U.S. customer, but also blaming some slowdown among European carmaker. How do you see silicon carbide moving into 2024 visibility? The second one is regarding your target for 2030 on silicon carbide, EUR 7 billion of revenue. That's quite huge. What is the assumption behind this EUR 7 billion target? Thank you.
Yeah, I would say, silicon carbide for 2024, personally, I'm very optimistic. T here's still a significant backlog while the market overall is softening. Still, there are customer requests for more, in the industrial space. That is true for the module side, and that is also true for the industrial side. I f we would not be able to consume it, which is not the case, I also know from my dear friends from automotive, which are not on the table today, that they would love to take it. A ll in all, we remain bullish and very optimistic regarding silicon carbide, that we will be supply limited even throughout 2024. Of course, the, the huge number you're correctly highlighting, it's a revenue potential, right? EUR 7 billion is a huge number, but that is simply the potential.
That we will put or are able to put in place, thanks to the expansion in Kulim. M arkets and customers in the end will tell if we will make use of it. O f course, we have also flexibility to allocate volume to gallium nitride or even to silicon, right? Depending what the markets request. I t shall indicate that we definitely also strive here for leadership. We want to remain the leader in power semiconductors. That is true not only for silicon, but for silicon carbide and for gallium nitride as well.
Thank you.
Okay, and with that, the next question comes from Didier Scemama.
Yeah. Cool. Thanks very much. Good afternoon, everyone. Thanks for hosting this call. Just would like to come back to a point you made, Adam, on the AI server opportunity. I think if my math is right, you had about EUR 650 million of data center-related revenues in fiscal year 2023. I think you said 4% of group sales. A couple of questions related to that. How much of that was AI? Can you remind us also the content in sort of traditional servers that you can target? I think you said roughly twice as much in AI servers versus, let's say, normal servers. Is the math right?
Then, the final question is, when you look at your design wins in AI-based servers on both GPU and TPU, what's your sort of best guess of the market share you will achieve in overall AI servers for the power management section, of course. Thanks very much, and sorry for the multiple questions.
No worries. No, all good. Very, very relevant. Y es, you are correct. We have publicly stated here that 4% of our revenue last year is obviously indicative of servers. Previously to that, it was at the peak of around 6%, if I'm not mistaken, on the previous years. We see then approximately the same coming in to this fiscal year, meaning really a similar trend of around the 4%. On the content, and again, this is why we gave a big range regarding the bill of material. It really depends upon whether you do the down solution, where you use the power stage solution, which is more of a standard type of a product solution, or whether you use the vertical power modules.
A s the power of the GPU, TPU goes up, and those who are very much focused on total cost of ownership, then, of course, we will see the ASP content definitely increase. A s a number, you can say between 2x-3x, we believe, content-wise, of course, is a reasonable number, but it could even be higher. It depends upon what power supplies are used as well, as well as what the GPUs, TPU requires, vertical power modules or on the down solution. T hen on the market share question, the final question, this is a very good question, but honestly, we're not in the state now to provide market shares. It's a very dynamic world, as you could imagine, that is happening. W ith our estimates at the moment, we are talking about 220K servers at the moment, 1.7 million cards.
It's pretty dynamic, and the architecture designs are still being finalized. O ver the next years to come, we will get more public reports on market share, and then we will have ourselves a better insight like yourselves as well. V ery clearly from the Infineon Technologies perspective, we also wish to be one of the leaders in powering AI. Thank you for your questions. Okay, David, yeah, your brief introduction got a little lost due to technical issues. I'm, I think so. Thanks again to BNP for hosting us, and I think you have some additional questions, so if you like, that will be the moment to-
Yeah, absolutely. M aybe if I could squeeze in just one or two more, gentlemen. Firstly, just on China, could you kind of each talk about the competition from China for your individual businesses, what competition you're seeing, and also maybe touch on the, the, the trade, regulations from the U.S. and China, where do you see the impact there?
Yeah. F irst of all, you know, what we see in China at the moment is that this isn't Infineon Technologies against China, right? We also believe in, in scale. We also believe in our product to system by bringing innovation to, to our customers. U ltimately, we, we, we believe and we've engaged heavily in a number of China-based customers that very much appreciate our capability in their, in, in innovation. On the export laws, so again, we, we of course, apply and, and adhere to all of the regulations and regulatory requirements that are given to us. We also believe in open, fair ability to, of course, open markets. We, we believe that that is a, a fundamental backbone for, for society.
At the same time, at the moment, we don't see any major material impacts with the latest regulatory changes, but we continue to monitor it very closely. W e don't see the immediate impact on the U.S. Bureau of Industry and Security related to front end or back end or manufacturing at this time. W e continue to monitor it, and where we see innovation and we see customers very much appreciating our technology and obviously our supply security, we're very happy to be servicing all markets and all applications. Peter, from your side?
Yeah, I think very important, and we simply take the competition always serious, regardless if it's coming from China or elsewhere. Of course, we know that China is a specific topic due to the huge market and, of course, also due to the huge amount of competitors. I think as in the past, for silicon, competition is not something new for us. A s things get a bit more complicated, namely silicon markets remain, and they remain also to grow, of course, with a lower CAGR, single-digit CAGR. In addition, we have the opportunity to grow and to differentiate in the areas of silicon carbide plus gallium nitride. This kind of bifurcation, so to say, is rather an advantage. You have to master and you have to manage the complexity and also, of course, the opportunities coming out of this additional complexity.
As in the past, our justification to exist in growing our business is differentiation and having, of course, a very competitive footprint. Y ou alluded to it, Adam. The topic of regional resilience is of very high importance. We all became aware of it due to the virus topic, that our supply chains have to be resilient towards whatever might happen and occur looking forward. On the political landscape, we're aware that we do not encounter another virus upcoming, but I think we have learned certain lessons, and therefore, we improved our supply chain that it is much, much more resilient than it used to be a couple of years ago. T hat's also the way to go looking forward, including, of course, all our important Chinese customers and suppliers, to make simply best use of it.
Okay. Any other questions? Nobody's raising their hand here at 5 P.M., but I think, good opportunity for you to-
Yeah, maybe, just one last question, just on pricing.
Mm-hmm.
Can you just talk about the trends expected for pricing throughout your businesses for next year, for FY 2024? I know you've given an indication that it's kind of more muted, maybe flat to down pricing, but just general, it's given again, through the lens of the Chinese competition, what are you seeing from a pricing perspective?
Yeah. I would say then, just at a group level, as we've stated, it's minor, single digit impact, but of course, a lot of that will be supported by volume, especially in the areas, for example, in automotive. I will hand over to Peter in a moment on GIP. In the areas of where we have more consumer exposure, especially in PSS as well as CSS divisions, we won't be able to make up, of course, on the pricing versus volume, till the markets, of course, really do come back in a stronger capacity, which we're anticipating in our second half of our fiscal year. I n China, look, again, we see the competition, as Peter stated, is not just in China, but in all parts of the world.
We thank goodness we also have the capabilities of scale and manufacturing, and we will then, of course, figure out where we need to be remaining competitive and make sure that we continue to look at the market share through the cycle. T hat's the PSS perspective. GIP?
Yeah. I think no surprise that it's a similar situation. We still see pockets and areas for differentiation where we might be able to keep or maybe even slightly raise prices, pricing. The times for significant price increases are over, no doubt about that. T here are, of course, more commoditized areas of products where we see competition heating up, where we have to give in on the pricing side not to lose share. T his is an overall combination, and the average temperature of the portfolio, which comes out as a result, is similar, as Adam says, right? I would say, keeping the pricing as stable as possible, realistically having low single-digit price decline is what I would expect for 2024.
Okay. Thank you very much.
Fantastic. I want to thank Infineon for joining us today and everyone on the webcast for dialing in. Any questions? The IR team is forever and always available, and thank you very much.
Thank you for your interest.
Thank you.
In Infineon Technologies. Thank you, everybody.
See you next time. Have a nice evening. Bye-bye.