The power requirements and that requires innovations The architecture level of the power flow in the data center, We are even innovating on quantum which could be the next big thing in AI. Hello and welcome. This is the podcast for engineers, You just have to listen to if you're interested in what's going on I'm your host, Peter Balint and today who oversees the business line responsible for power supply intermediate bus converters So welcome. Thank you Peter.
It's great to be here on your podcast. First question I have for which is needed for powering AI look at a server rack in a data center, and it consumed maybe 60 kW of power. Today, it's a different story. We're talking So what does this energy, explosion mean in terms of the trends You're right.
You're referring to increasing power And I'm I will be happy to, but I would like to give it because we here at Infineon, we believe AI adoption and and in particular, and computing capabilities needed for What we see here is really a convergence It's software on one side But it's even, you know, And all these three things combined what I would call a exponential growth And I believe that this exponential growth will really and truly transform our society within the next few years, Yeah.
And what we see right now with respect to, the AI that every one of us can use opportunities to use AI assistants like, for example, if you want to help yourself But the potential of AI is much broader. Intelligence to humans, to us as human brains. And now it's becoming available. It's becoming abundantly available And the AI models are improving Right.
Large language models have now tested up to with an IQ of up to 150, while just maybe nine months ago, So you see this, And now just add and you can see where basically this is, exponential age is, is taking us. Yeah. It's fascinating, but That's a good question, Peter, and intelligence are very closely linked.
On one side, smarter better into energy sources, in particular And this is important to bring also the And on the other side, more energy allows us to build up you know, train bigger models, But with the advancements of AI, is that the demand for electricity grows, And that is a bottleneck right now for, for the build out of AI. And so there are multiple ways but three of them are mentioned, One is more and more decentralized renewable energy That's point number one.
Point number two is working on really the and and here DeepSeek with smart preparation of, of the data, the amount of energy And the third one in a data center, in particular, in the data center can really make a great contribution, at the complete power flow from grid. And then with all the conversion So the to the AI chip. So we are on one side, enabling better power supplies And on the other side, we're even thinking architecture approaches.
So the best conversion steps the whole power management more efficient. So then let's dive into this idea To the server rack. Why do you think it's necessary moving forward? To put it very simple, On one side, the GPUs. So the chips are getting more We are today standing at numbers.
Head up to 2000 watts per chip, And on the other side, the, architects of this racks want to put more So that means in the future, don't have enough space to also house So you want to move the power supplies challenges, that has its own challenges, you need to bring the power back in Typically they use bus bars because they now need to handle which means a lot of current needs to flow And in addition, you have all these that you already use water cooling Basically all the power of the rack
So then how does this translate So we see that everything like the AC, DC But then you need And today So 50 volt for example. But that means lower voltages Right. So if you want to improve That would But once you start to use higher voltages, of different safety and isolation Right. And the next logical step, would be to go to a voltage class of 400 volt or maybe even plus and -400 volt. But once you do this, you get a lot of So tell us about the advantages first.
So the first and most obvious advantages that you can bring in energy and power with much thinner wires into the core. The second is related which is the DC DC conversion. And this can be done in a much more power And the third advantage to use 400 volt that is today So you can reuse a lot of the product that has been developed So those were the advantages.
The disadvantages are that you really need to prepare for a different infrastructure And some of the older That is, it is not compatible with the safety and isolation So I assume that this 400 volt DC something used in new installations around Stargate, where everything And you can already prepare for that. There's one more disclaimer the work that we do today around really is still based on the hypothesis that that is a system requirement, right?
And the system the GPUs are connected via copper cables. And because they need to, the length of these cables So you really want to have all the GPUs In the future. You could also imagine photonics and laser light connecting, the GPUs. But at this point in time, that's at least 3 to 5 years out. So for the time being, this physical proximity of GPUs is necessary needs to adopt to these requirements Okay.
Interesting to look at the advantages But if we go to this 400 volt architecture, then what would this mean Well Peter, such a change in architecture really requires innovation And for us and we would like to use really starting from silicon, silicon carbide Yeah. So let's start which would be the AC/DC rectifiers. So the front end power supply to 600V of AC voltage In this case we would very likely use I'm thinking about five level, topologies piece of it.
And we would use silicon carbide products on the on the use case And then of course also price performance. So in particular efficiency is important for this AC/DC rectifiers So there's enough space. But efficiency is important that needs to be converted. Right. And we are targeting For that, a piece of conversion, volt DC down to 12 volt, And here we are We are moving close to the GPU. So here It's all about power density.
The real estate around the GPU is scarce So we want to come up And for that we use our latest We use, for example, bidirectional gallium But we also use silicon. And we use depending on the sweet spot we use these components But as a blend we believe density for these DC/DC converters So 400V. This seems like a possibility But what lies beyond that. Yeah. So the the described solution this would bring us to power levels Right. And that assumes that the power supplies That is good.
But that's not good enough. The power levels will continue to go They will reach half a megawatt, So for that further architecture are needed. And the ultimate goal is to have only IT supplied by a 400 volt DC, busbar across the data center.
So you would have centrally produced 400 volt, maybe plus minus 400 volt and then distributed But that comes of course, with a complete own set of safety and isolation You need to have dedicated personnel And you also have the challenge backup units that operate on that 400 You need to have protection features potentially e-fuses that are able to to be used in 440 environment and Infineon is innovating in order to be ready with a portfolio of of products for the 400 volt ecosystem.
So as we mentioned in the beginning the increased need And you've gave us a couple of solutions that are maybe right around the corner, What else can we apply here? Yeah, in fact, I do see a lot of potential in quantum You know, for certain use cases, over traditional brute force approaches a lot of transistors on a chip, and you then use 100,000 of these Right.
There are certain use cases, like, and one of the most famous optimization And I briefly described So it's it's and they live in different cities of this customers, all the options If you have a customer count of ten, with a classical computer, you have already or more than 181,000 of options So knowing the right sequence.
But once the customer count increases, It's impossible It's impossible because there are more options than there This is where quantum computers This is where you can really use them And you have this kind of optimization in DNA sequencing, you can also find it And in the semiconductor industry, we have a pretty complex So even if AI advances big time, there will be certain limitations And even if the times are still early, We in the early days of quantum computing, quantum hardware to
train AI models for some use cases will be significantly and potentially might even open up where you cannot go So, okay, you introduced this topic of But we also have traditional computing. What are the benefits Yeah. So in indeed, I see that I see quantum computers excel challenges, So a small dataset. But that dataset has an enormous and vast Examples for Right.
So where a handful of different proteins in a huge amount of possibilities complex molecules, describe in a quantum environment, computing science on the other side. Ever. You have vast amount of data, for example, pictures in the internet This is where traditionally I think, state of the art So I believe for different challenges highly specialized approaches Okay. And it makes sense.
And quantum computing But give us an idea of what Infineon s Infineon is innovating on trapped-ion quantum computing. What we are doing that's really the core processor And we have been investing in this combining our expertise in high volume with some specialties or integrated control We are working together with three industry leading partners. And in the meantime, we have more than 100 patents So at this moment we see another technology This could be really big.
So it looks like quantum computing And this is a whole field by itself I will ask you to give a final few words Yeah. Thank you very much. You know, even if it's sometimes exponential evolution of AI, to stay optimistic and forward looking And we should embrace this future Because with the smart energy we are not only embracing but we are also making sure Perfect. Thank you so much for coming in today Thank you very much for having me. And thank you for our listeners for dropping in today.
If you have any questions please don't hesitate to contact us Thanks a lot and see you soon.