Powering tomorrow: The global infrastructure behind AI

0:00 (gentle electronic music)

0:02- It's not the energy generation.

0:04 So we are going to have enough energy, at least for some time, but to get it into the grid and to manage it properly from generation to transmission, to distribution, to storage, to conversion, and all these things you need along the entire energy power infrastructure, this is where I see the main bottleneck.

0:28 With the AI data centers, the loads are very concentrated. They are very focused on certain locations, and that makes it somehow special. The grids as we have it today are simply not prepared for this. (gentle electronic music continues) (screen swooshing) (objects clattering)

0:50- Welcome to the "Podcast4Engineers"

0:53 and to the eighth and final episode of our "We Power AI" series. Over the past episodes, we've explored the technologies, the infrastructure, and all of the semiconductor innovations making AI possible, from efficient power delivery, to the data centers, and to the energy systems behind today's most demanding workloads.

1:12 To close our series, we're looking at the bigger picture, how power infrastructure is evolving globally to meet the demand from power-hungry applications such as data centers and what we can learn from past technology transformations. We're also gonna look at what comes next as AI moves from the cloud into the physical world.

1:33 I'm very pleased to welcome Andreas Weisl, who is the CSO of the Industrial and Infrastructure Sales segment. Andreas brings a broad perspective on power semiconductors, infrastructure, and the major market shifts that's shaping our industry. Andreas, welcome to the "Podcast4Engineers." It's great having you.

1:52- Thank you for the invitation.

1:53 Happy to be here.

1:55- So, Andreas, this is actually your first time

1:57 on the "Podcast4Engineers," but you've been at Infineon for a while, right? Can you tell me a little bit about your journey at Infineon?

2:04- Yeah, sure.

2:05 So I joined Infineon back in 2018 at this time for the corporate account management, and then roughly three years later, I took over the role as the head of Automotive EMEA. And that was a very exciting time, I would say, from today's perspective because that was for me the starting point of the automotive transformation.

2:27 We already felt that this transformation is accelerating. We saw already first changes in the customer landscape, and that was a unique opportunity for my team and myself to shape the organization in a way to meet the new market and new customer requirements. So that means basically we looked into all the roles.

2:47 We reorganized sales, technical marketing. We introduced new marketing roles also. Then we expanded our presence in East Europe, focus Romania and Poland. We expanded our own team in Portugal. And we also discovered new channels like OEM, which was not a real sales channel before but became very important as part of this value chain changes.

3:14 Yeah, and then just beginning of this year, I took over the new sales segment, what we call I2, or industrial and infrastructure. And now it's all about strengthening the customer relationship in this industry and accelerating growth, of course, and put a big focus also on collaboration to drive the success forward.

3:37- And you mentioned in your intro transformation.

3:41 So we are in the midst of a big AI transformation right now and you mentioned with electromobility being another transformation. Have you seen any other shifts or big, yeah, movements in the industry during your career?

3:57- Indeed, yes.

3:58 I would say that I have been witnessed now firsthand three transformations, industrial transformations. The first one is lighting. My first company was also opto semiconductors, and then later I was eight years with a Korean company called Seoul Semiconductor. Both companies were in the field of optoelectronics.

4:19 And at this time, these new technologies basically replaced all the traditional lighting technologies. And what started as a niche technology basically changed the entire industry within a fairly short time period. So, big market-leading companies were challenged and, yeah, challenged in a serious way by game-changers and new entrants.

4:46 Then at my time at Infineon, I experienced a similar pattern with automotive at a much larger scale because the first one was optoelectronics only, but at Infineon Automotive, it was a much broader scale or broader portfolio of semiconductors. So here we were talking about, and we still talk about because it's still ongoing, this transformation, electrification, e-mobility, but also software-defined vehicles.

5:15 And when I say broader portfolio, it's about increasing computing power, MCUs. It's about battery management, sensing, connectivity, software-enabled features. This is all based on semiconductors. And now after taking over the infrastructure segment, I see the same pattern again, that we have certain challenges in this industry.

5:43 New technologies which have been there for some time are now reaching this inflection point, and they make solutions more efficient, more reliable, and they also enable new features. So I see similar patterns now here, and I believe that we are ahead of the next transformation, which could be even bigger than the other two.

6:09 And there's one common thing, it's always driven by semiconductors. So I strongly believe in the meantime, if a traditional technology can be replaced by semiconductors, it will happen.

6:22- And yeah, history often repeats itself,

6:26 and you mentioned being there for the transformation of Infineon during these major milestones, but what can our customers learn from these transformations on how to best deal with them?

6:37- Yeah, I mean, first of all, I would say

6:39 that breakthrough technologies are never enough. So there's a lot more necessary to start a real transformation. I think what is important to know is that at the very beginning, the transformation takes a little bit longer than people estimate. So everybody's excited about it. They talk about it. The whole industry talks about it, but then, you know, everybody's waiting, and it takes more or longer than people expect.

7:14 But then when these inflection points are reached, then it's really accelerating, and that's also then faster than people anticipate. Second pattern I would mention is that it's never the product or the technology alone. You need to have the ecosystem. You need to have technology providers like Infineon for semiconductors, but then you need system integrators.

7:40 You need governments to support it in certain areas or regions. You need utility companies, and last but not least, maybe also the end consumer to make it really happen. So I think that's a few patterns I have recognized over the last years.

8:02- Okay, yeah, it sounds like every transformation

8:04 is bigger than just an industrial shift. It's really a societal shift, a government shift, something much bigger. And today we're talking about AI, and this whole series has been about powering AI. When most people think of AI, they think of Silicon Valley right ahead, right from the start. But the growth story isn't just happening in Silicon Valley, right?

8:27 How are different regions dealing with the shift in AI, building the necessary power infrastructure for that? Can you tell me a little bit about the regional perspective?

8:36- What I observed over the last quarters is that

8:41 we have a massive energy demand increase all over the world, and we see also massive investments all over the world, but the driving factors behind are different. So, as you say, everybody is thinking of Silicon Valley when we talk about AI, and that's also somehow understandable because it's massively what we see in terms of investments, what the hyperscalers are doing these days in the US.

9:11 If you look into the statistics, you see that the demand increase of electricity in the next upcoming years in the US is mainly driven by AI data centers. So it's a large portion. Some people say half of the demand is coming from this. So that's why, not a big surprise, that we connect these two dots.

9:28 But for example, if you look at China, the overall electricity demand is much larger. And, yes, data center is also here a very important topic. They invest aggressively into this area, but it's also fact that the electrification of the industry will take a huge portion. We have the infrastructure programs from the China government running, which has a big impact, and still even they are, I would say, number one in e-mobility, there is still some way to go to electrify the whole transportation sector.

10:09 So that's about China. So then when we have a look at Europe, I would say also here data centers are important, but the driving factors are very different. So here it's more about digital serenity. It's about meeting sustainability goals, so, the targets defined some years ago. And it's also about, or still about, renewables, to get more renewables for energy generation, but even more important, how to get this energy then into the grid.

10:40 So that's the thing important for Europe. And then I just came back from India. That's also an interesting atmosphere these days because also here we see a rapid increase of energy demand, and that's coming from urbanization. It's coming from buildings. It's coming from HVAC heating, cooling, actually-

11:04- Air conditioning, yeah. Yeah, exactly, maybe more cooling than heating.

11:08 But that's also a big factor. And India is maybe special, again because here we talk about modernization of a whole grid infrastructure of a huge country. So in a nutshell, I would say we see these investments all over the world, and the common pattern is that they're all looking for a more efficient, more reliable, more intelligent energy infrastructures.

11:40 And the positive news from our side is that I strongly believe that the infrastructure or the grid infrastructure for tomorrow will be a semiconductor-enabled power infrastructure.

11:54- Okay, so it sounds like each region

11:56 is approaching this challenge differently and that the race for AI is actually a race for power infrastructure. Is everyone at the same starting line?

12:06- I mean, first of all, I would say that,

12:07 you know, we say that roughly 3% of the global energy demand is related to a data center, and this doesn't sound a lot, correct? That means 97% is something else. But the difference is that with the AI data centers, the loads are very concentrated. They are very focused on certain locations, and that makes it somehow special because the grids as we have today are simply not prepared for this.

12:34 And that's important to keep in mind. And if you look at Europe, for example, 40% of the grids are older than 40 years. Or if you have a look at US, I was told that the grid installation we see today were done between 1950 and 1980. And if I talk to my colleagues in the US, they tell me that if you have a new installation today, and you want to connect them to the grid, it takes years to get the approval.

13:00 And if you ask me what is the acute bottleneck, it's not the energy generation. So we are going to have enough energy, at least for some time, but to get it into the grid and to manage it properly from generation to transmission, to distribution, to storage, to conversion, and all these things you need along the entire energy power infrastructure, this is where I see the main bottleneck.

13:30 And that means, back to your questions about the race, it's not only about the computing race. It's also about how to deliver the energy needed for this, and this race is ongoing now, right now.

13:47- Yeah, and there's no computing

13:48 without power to compute, right? And we've talked in previous episodes about Infineon's role in compute, so, inside the data centers. But what role do we play in this power infrastructure race?

14:02- Yeah, I mean, if you say the challenge

14:03 is both the compute challenge and the power infrastructure challenge, I would say that Infineon is uniquely positioned in the intersection of these two worlds because I would say Infineon is a leading power company, a power semiconductor company, first of all. Second, there are different materials now in the room, and we are able to manage all of them.

14:30 So we can manage silicon. We can manage silicon carbide, and we can manage GaN. So I think that's important. But also worth to mention that it's not only about product or technology. It's a lot about, especially when we talk about infrastructure, it's about quality. It's about reliability. And I think it's also about decades of application expertise, what you can bring in.

14:58 And talking about the unique position of Infineon, I would even add that we have been already successful in traditional markets. So now when we talk about emerging markets, we can transfer knowledge from these markets like e-mobility, like renewables or industrial automation into these new emerging markets.

15:18 So that's a value for sure. And when we talk about these kind of applications, it's about SST, solid-state transformers. I think you had it... We had deep dives in your previous episodes. We are talking about solid-state circuit breakers to protect DC grids. It's about advanced UPS systems, energy storage systems, and a lot more what will help us to modernize the grid.

15:46 So I would say the winners of this race will be the ones who really understand not only the technology and the product, but also the ecosystem around and able to work with the system integrators.

16:06- And you mentioned a lot of emerging applications

16:08 and innovations happening in the space right now, but where do you expect the next big things to be happening? Do you think the innovations will come from inside the data center or outside the data center?

16:21- Oh, that's indeed a tough question.

16:23 I mean, from my point of view, when we talk about AI today, we talk about large AI models running at massive data centers, correct? I think going forward, it's also a lot about beyond the data centers linked to the physical AI, and when we talk about physical AI, it basically means robotics. It means autonomous systems, industrial systems, let's say intelligent machines.

16:54 And I think it will become important to have more computing power, or more intelligence, I should say, where decisions are made. And that means basically at the edge at these kind of machines. So, but, still I believe that the big cloud data centers will remain important. They will remain the backbone of computing, but for sure we will see more intelligence moving to the edge because at the end, it will be about AI systems who are able to sense, to compute or think, and then also to act.

17:36 This will be the next revolution ahead of us.

17:40- Yeah, it's quite exciting to think about intelligence

17:43 and things like robotics and things like automotive, different applications. But what excites you personally most in the next maybe 10 years or so?

17:56- I mean, AI in general is very exciting, to be honest.

18:00 I mean, I'm pretty sure we will see here more progress in the society. It is about production. Efficiency will increase. It's about healthcare. There will be a lot of impact on the societies, and that's amazing what's ahead of us. But on the other side, and also very much close to my new role when I think about the infrastructure, which is accelerated now by the AI data center, it's a catalyst or accelerator for this.

18:31 But every invest what goes today into the data center is indirectly also an invest in the power infrastructure. And this infrastructure, the modernization of the infrastructure, will accelerate then the electrification of industrial sector. It will help the electrification of the whole energy sector.

18:52 And at the end, this will help to change the world in a way that we also have a more carbon-less world in front of us, that we have an increasing productivity in all areas. So I would say these things are very close connected to each other. On the long run, for sure the infrastructure part. Short term these days, of course, it's fascinating what's happening on the AI side for sure.

19:19- Yeah, well, Andreas, thank you so much for taking the time

19:23 to discuss the topic with us and to give us your perspective on the race for power infrastructure. Thanks for being here.

19:30- Thank you very much for having me. And to our listeners,

19:35 what a great way to wrap up the "We Power AI" series. Across the season, we've seen how semiconductors help turn AI ambition into real-world infrastructure and how efficient power technologies and renewable energy can help grow data centers and grow AI smarter and more sustainably. If you enjoyed this episode, share it with a fellow engineer and join us next time on the "Podcast4Engineers."

20:00 (gentle electronic music)