235. European Sovereign Neocloud - Jun26

[Music] With Laurent Segalen from London and Gerard Reed from Berlin, this is redefining energy. Today on redefining energy, Jaloud, we're going to talk about European sovereign neocloud. Yeah, which is obviously a critical topic in this AI revolution that we're currently at the beginning of. Yes, Jal, because when you see AI investment right now, it is the biggest capital expenditure cycle in history. We are seeing statistics that the CAPEX is going to reach 9% of global GDP, never seen before. The peak of the railroad boom in the 1880s was 6% of GDP. And the current AI data center build is bigger in the US than the rest of the construction company. So we are really in an extremely odd sector, but it's very US. And when you're in Europe, you have a different point of view. Well, let's, I think we really honest with ourselves what's happened in Europe is AI regulations have blocked the development of AI companies across Europe. So we don't have any foundational model companies like the Chachibitis of this world. Yeah, I'm playing catch up. And I think there's a growing realization that actually one of the advantages that Europe has in particular businesses is they have data. And do you suddenly want that data sitting in another country? And even if it's not in another country, you want to make sure that that data is not leaking out to the rest of the world. It's also growing concerned that you don't want to be just sticking with all your data stuck on AWS or Google or any of these other big American players. We need to really make sure, as I said, that your data is your data. Absolutely. So we have a super guest, Michel Boutouil, is the co-founder and CEO of German company called Polarize. Polarize is one of the few European what we call NeoCloud and is a partner of NVIDIA and is mission is to build so-rank AI factories, high performance across Europe. And Ron, I think what we also want to do is we want to thank the BMW foundation Herbert Quant for their support. And those of you who don't know the foundation, what they're all about is uniting leaders across sectors to develop solutions that foster innovative economy and a future proof society. A key focus is the energy transition and climate change where the foundation drives international collaboration to accelerate the energy transition. Obviously, with rising energy demand from AI and data centers, new partnerships, effective collaboration, exchanges, science based solutions and strategies are not just essential, but like that critical really going forward. For our listeners, we had to learn what the NeoCloud company is because most of the time we are outside the data centers, so I'm more like energy and real estate, but here we are going inside the data center. So a real definition I just found, a NeoCloud company is a vertically integrated AI era infrastructure provider that controls land, power, data center, build out and GPU cluster and to end. Offering IPR scalar great compute without being a traditional cloud like AWS or Azure. So the type of company which are NeoClouds in the US, you have core with Nibbius Crusoe. And so they are different from what they call the power shell, the whole one, the Elix, the Vantage, the Cyrus one, aligned data centers, you know, which provide the infrastructure from the outside. And of course, you've got another layer, the original one, the traditional collocation, reads, a queen X digital reality, which are multi tenant. It's more retail wholesale focus. And then you've got the IPR scalers. And on the top of that, a lot of money, the blackstone, the KKR, the Brookfield, Stone Peak, digital bridge, they all financing this absolute crazy build. We'll see what Europe has to reply to that US push. So let's bring Michelle on the show. Michelle, welcome to the show. Thanks for having me. Well, Michelle, maybe I'd like to kick off here now and listen, I'm going out with I suppose a thesis that really were in the fifth industrial revolution. And this is the biggest industrial revolution in the history of mankind. And there's two things. One is its AI and AI needs electricity and action. And action, a funny way electricity needs AI. I'd really love to firstly ask whether you agree with that view. And then secondly, just talk about Europe's role in this revolution going forward and how you see it. Yeah, I totally agree with that. And I would even go a little bit further. It's not just an industrial revolution. It's also like a revolution of the society because we have to adapt to all these things which AI brings to us. For example, it changed the way how we work in the future. It changed the way how we are going to spend time, you know, always work and so on and so forth. So that means yeah, I'm totally agree to that. My second part to the question then, really is was in and around Europe. What's Europe's role now? Because again, stand back and look at it. I see US doing a huge amount, particularly in and around learning for national models. China is going a slightly different approach, which is a more open source. And then I sort of see Europe and I wonder what Europe is actually doing. This is the problem because Europe is not doing much at the moment. The biggest issue is if we are not starting now, we are missing the train. Because you see what is happening in America. You see what is happening in China. They're taking a major role in this AI business and it means as soon as the European economy starts to adapt to AI, it will have, it needs a choice. And if there's no choice from a European perspective, then they will go to a hyperskiller to an American one or to a Chinese one. And they get sticky and then they can't go back anymore or not so easy back anymore. And it's not just about just about sovereignty is also just it's about how we're going to give out our data for free to these models to learn to replace us in the future. And this is a real threat. And on my opinion, we're doing a big mistake in Europe because we are trying to replicate things from America like we did in the European Union. There's a tender. The tender says we want to have 100,000 chips in one giga factory. And to be honest, I think this paper was made by a lot of lawyers, but no technical guys has sitting there and explaining what is really needed in Europe. And Michelle, jump in on that, it's tough because let's go back to the fact that it is a revolution that we're going through. And I'd be critical of the European Union too because when I look at their AI directive and all this type of stuff, it just slows down. You're trying to regulate something that is changing too quickly and we're not going about it in the right way. So I'm totally with you. And I suppose the question then to ask is, well, what should we do? Well, we as polarized trying to build something which is not existing at the moment in Europe, it called the Neo Cloud. So means a cloud who specialized in AI for AI purpose like AI infrastructure. So we need to enable European companies startups and so on and so forth so that they are starting to build something that they're going to invest heavily into data center into AI without that we can't catch up anymore on my opinion. So we don't have to regulate everything beforehand which we need to have some form of sandboxes where we can start to work with. Startups needs to be supported, scale ups needs to be supported, we don't have to look just at the big carbs, we have to look like a mixture of everything. And if we're not starting now to trust in ourselves means that we're going to lose on the long end on my opinion. So why I understand it and that's coming from somebody who doesn't understand much is that you've got two type of AI work, you've got the work where basically you run those big models. I have no idea how it works but it looks like it consumes a lot of power. And then you've got what they call the inference which I finally understood what it means. Basically it's the asset management once the models are running. So do you concentrate on the first or on the second? So we are concentrating on the second because maybe I'm going to make it a little more understandable. Large language models needs to be trained first so that means we have to raise the child we have to give them the ability to speak and to think this is let's say training in the I world. And we don't have so much foundation mode so we don't have so much training in Europe. So we're done by opening eye and by all the big guys and the inferencing is like AI in operation. So that means as soon as you use AI you are implementing into your workflows and it starts to do something for you. Then it's called inferencing because like consuming constantly tokens and tokens is a form of energy. So our computers are transforming energy into tokens. The tokens are consumed by the models and this is how it operates. And we are specialized in inferencing and because what we think is that we don't need to have so much foundation mode because there are a lot of open source models out there. We can use these open source models. We can fine tune them. We can put them on an infrastructure with which is sovereign and then we can start to use it in Europe or European companies can use it because we need it because otherwise we can't compete with other companies in the world who have. have this advantage of AI. So we have two choices. One choice is we go with foundation model, which are proprietary models from the Googles of the world. And that means we're giving the data into this. This models are learning from it. And then it could be in the future, it could be possible that they're going to replace work, which was done here in Europe at European companies. Or we can use open source model. We put them on a European based infrastructure with European companies. And then it has the same effect and then the companies can use the AI as well. So and this is what we're doing in polarized. Michelle, I just want to go back to open source here again, because if I look and I see all the tests and studies that have been done now, if we went back three years ago, there was like a year's difference in terms of the quality of an open source from versus a proprietary system. But now it's like one month. And so am I wrong to look at it and say, who cares if it's proprietary not? If it's open source is a way to go going forward, which would mean if I were a European and I want to catch up, I really want to be the fast follower. I don't want to be the leader of the US who's going to do all this, spending all this money and all these foundations. I just go and take these open source models and I go directly into inference. Am I thinking about this right wrong? How do you think about it? You're absolutely thinking in the right way. So I think that the open source models has catch up so fast and they are pretty good. And you have so much varieties and you also have to think how much performance do you need? How much parameters are really necessary. If you look at the biggest foundation models in the market and you see the parameters, I would say in opus 4.0 is not necessary for 99% of the people who are using AI at the moment. So that means you don't need this super, super, super high intelligent parameters, but you can go with open source, you can save a lot of costs and you can also save your data. And it will be at the end of the day, the open source model will be your personally AI model, on my opinion. And that is super interesting. So that means yes, I'm totally following what you say. If I understand well, you say your nail cloud. So I know because that's kind of funny. Michael Intrator used to be my carbon broker 25 years ago. Well, he's done very well. It's the sea of coal we've now. So you can have a European coal with, correct? Yes. Okay. What we do know from the energy guys is basically what Rohan is doing created by Queenbrook, fantastic company, but they are there's it's basically all outside the data center. So you know, they arrived, they put a terrain, they build the whole infrastructure and basically they deliver you or they deliver somebody a building, which is ready to be powered and cooled and everything. And that's where you come in, explain a bit what you're doing. No, it's in fact, it's not where we are coming. We're coming a stage earlier because if you look at the actual data center industry in the world and it's mainly let's say cloud business like storage and so on and so forth. And it was like that the infrastructure. So the real estate building was thought from real estate guys and then the computing, which was implemented there was staying in the same shape for years for maybe 10, 15 years. So inside the data center, you had like a real estate part and then in the in the real estate part, we had like a standard of some form of it called Rex. It's like shelters. And then these shelters had a lot around 15 to 20 kilowatt each shelf today. And we are seeing it at the other way around because what we have understand and this is why we are building the data center ourselves because at the moment there are no co-location data center, which is suitable for the AI infrastructure. So to say because you have to think backwards, you have to think from the IT back to the real estate. And there's a lot of differences between the old data center economy and the old data center economy will not be ending. It will stay as it is. We still going to consume data. We still going to have internet. We still going to have storage and that will also have a growth every year. And then there is a new form of data center. These are the AI factories. So we are not calling them any more data centers. We are calling them AI factories. And these AI factories are totally different constructed. First of all, we are not going with 15 kilowatt per rack. We are going with up to 115 more. And we have to think not just about the generation of compute today. We also have to think about the generation of compute of tomorrow because the iteration of compute is going so fast. So every GPU model, it will be replaced in the next three to five years. So we're going to have inefficiency every three to five years in GPUs. And we're going to consume more energy per GPU in the same amount of time. It always doubles sometimes, sometimes even triples. And that needs to be sought as well. So if you're going in a classical data center and if you don't think about all these different elements, you're going to be outdated super fast in the data center world. So that means you need to think differently in AI factories than you did in the past with normal data centers. Normal data centers are not suitable for AI. And that means there's a totally new classification of data center coming into the market with a big hunger of energy, but with a lot less space. So for example, we are putting per square meter a lot more energy than a normal classical data center. So means we can build in 2000 square meters. We can build 2025 megawatts. Very interesting. We're saying about energy there, but I'm going to ask now just come up from the energy side. When you look at data centers going forward, are we going to build big huge ones? Are they going to be small? In other words, are we going great edge? And if it is great edge, is it 50 megawatts, 20 megawatts, 5, 1? How do you see the future in Europe of building these data centers? Are sorry AI factories, sorry AI factories. No worries. So I see it in the same way like we're producing energy today. So we have like big nuclear power plants in Europe. We have smaller coal plants. We have gas plants. We have windmills. We have a lot of different power production all over Europe, which is necessary because you need sometimes a really big power plant to produce a lot of energy because you have a region where you have big cities and so on and so forth. Whereas necessary. I see the same in AI. So of course, we're going to have 100 megawatt or 150 megawatt big AI factories, but we're going to also have like 20 megawatt factories. We're going to have 50 megawatt factories. And what we think, we see it as a, and we call it giga grid in polarized. So we don't need these big centralized single point of failure, so my opinion. So what we need is all the mixture of different form of AI factories, which fits into the region and into the market. For example, if you're living in Munich and we have one data center or one AI factory in Munich where we are running 10,000 GPUs, it has around 25 megawatts. So it's in terms of data center, standard data centers, quite huge, but in terms of AI factories is pretty small. And with this 10,000 GPUs, we have doubled the AI capacity in Germany at the moment. That means at the same time, there is a lot more hunger for GPUs and AI factories. So we need to build them fast. And on my opinion, there's a lot of option than the market. If you want to go fast, because if you build a 200 megawatt or even bigger AI factory, it takes time because the grid connection is not so easy. You need to find the right point to connect. Then you need to build it. Then you need to have all this redundancy. You need to have a lot of things to build this huge factory. But if you find an old production fabric, which already has 20 or 10 or 15 megawatts in place, and you are capable to retrofit it, and you have the creativity how to transform a former whatever factory it was into a AI factory. And using the same infrastructure, which was already there, and you just upgrade it, you just put it in a way that is suitable for an AI factory. And then you have a very fast and a very efficient AI factory. And it's also, on my opinion, so that you don't have to build everything from scratch. And you're saving a lot of CO2, for example, and so on and so forth. And we have in Europe, and especially in Germany, we have so much potential sites. And we see that every day, which could be transformed into an AI factory and being a super efficient and a super fast way forward. So there's enough options. But the option needs some creativity from companies like us or others. And if you just go the normal way like we did it in the normal data center world, for example, in Frankfurt. So we have in Frankfurt, a lot of data centers because we have there the international and the transatlantic internet not. And there is DKX, which is one of the biggest internet distribution not in the world. This is why everybody wants to be in Frankfurt. But if you want to build a data center in Frankfurt, it will take you at least six, seven, eight years until you get the grid concession. And it will cost you 40% more than other regions. And that will be too late on my opinion if you're going to go the same good. It's a very interesting approach. It's a very German approach, decentralization. But if you cross the border and you are in France where everything is stopped down, decided by the president, the only thing they are interested is to sign 75 billion with a soft bank who's going to deliver or not. Is it compatible or who's right who's wrong? Everybody's right. How does it work? It is compatible because if you look at the reality in France, the grid connection times is even longer than sometimes. For example, if you look in Paris region or Bordeaux region or other bigger regions, it's a nightmare how long it takes until you get the real data center on the ground. At the end of the day, you know, more centralized country like France, they're going to have the same problems. But and there are the same options. You also have a lot of industries which are not performing any amount, any in the same way, which can be. retrofitted into AI factories and we see that all over Europe. So it's Spain in England and we have projects in Spain We haven't we have also a project in England the grids in England It's also like a very bureaucratic and not an easy way to go So that means we have that all over Europe the same problems the same kind of grids constraints and if you look in Scandinavia Where it's a little bit better, but still also in Scandinavia is now catching up in bureaucratic processes and not being So easy anymore than it there was in the past Michelle, I like your approach and I think okay, yeah using existing infrastructure makes a lot of sense to me Talk a little bit about the client perspective in other words Why would a client go to you rather than go to Google or Microsoft or someone like who they trust and they know for many years In other words, what is it that differentiates you from them? I think there are three reasons first of all we are not more expensive than Taking a hyper skiller second. We are sovereign means all our clients are protected against the US cloud So that means that a government can decide to look into your data They can even decide to shut you down and The service we have no interest in your data We don't drive our own models. We don't use your data to learn from it We don't feed our own machines. So that means everything stays with the client and you can be sure of course They are the form of legal frameworks where you can protect yourself But at the end you don't know how the hyperskilers are using your data for their own purposes and on good and We know in the past Amazon had a market place So everybody could bring his services and goods to this marketplace and set it over Amazon and then Amazon learned from it What are the best products which products can be sold very fast and which are super interesting to sell and then they sold it themselves So why should they not use your company data to train and to understand how they can make money with this? This is something every company in Germany and Europe needs to think about that So this is why the use P of polarized is clearly that we have the same quality of service in AI we have no intention to use your data for anything and We are fully compliant to the European Union laws to GDPR and we have nothing to do with the US cloud Act and I also have to say of course we as polarized are not so big at the moment So we are trying to do our best But what we do for example, we do some form of partnerships for example with telecom the German telecom who are building the trust layer Who are giving us a lot of let potential into the market but also with others big European companies or potential cloud Champions in the future like the Schwartz group where we are delivering the layers they need to repackage it and to sell it to their customers But at the end of the day what you decide to put the architecture whether you put open a little black well or a rubin or whatever else That's your decision or that's the client to tell you that's the cheap I want There's different layers so we have one layer that is the data center the eye factory Then the next layer is the compute means what type of GPUs what type of storage on top of it We have a layer which is called core which is like a software layer where a customer can just consume virtualized Server and GPUs and on top of it we have AI as a service that means we have a platform where we are hosting Different open source models and you as a consumer can directly Log into the platform you can choose your model and then you can consume tokens and your payers per token So if we've selling to a customer the GPU layer which is called bare metal then of course the customer will tell us what kind of GPUs and what kind of Computing once but if we sell it through core or to our platform then the customer they don't need to understand the Infrastructure will be lower because they don't need the GPU nobody wants a GPU We just wants to have these AI in operation to use it for two purposes to make more money or to make my company more efficient These are just the main drivers at the end of the day So that means yes Sometimes if we are selling it to a bigger client like telecom then they choosing the IT they telling us what kind of a tea they want But if we are selling core or drive then we are deciding what kind of a TV you're gonna put into the data center Michelle one thing that I see in my work here every day really is that the people driving AI demand for data centers in Europe are all the hyperscaders in the US I don't see any European players actually doing anything Yeah, that's correct and this is the problem because it has two elements one element is that the hyperskillers are going to consume most of our infrastructure So that means they're blocking the infrastructure as well because they need a lot of energy for the training clusters And the second element is that we are not really investing into it because we as Europeans we are investing in balance sheets And we are not investing in ideas means the hyperskillers they have no problem in buying Thrillions and billions of euros for GPUs and training their models But they also don't know what's coming in the next four or five years, but they have an idea in Europe For the big tail cores the big cloud players in Europe. They are not investing at the moment so heavy and this is why everybody is looking at the Governments and is looking at the European Union to subsidize some part of the capital expansion and This is a problem because that makes us super super slow It doesn't give us the real opportunity to grow fast into it. We as a company we see that so for us It's super heavy to collect the billion for example a hundred millions So which is a lot more easier if you an American company with this corvee for example So corvee or end scale or Nebios they have collected so much money and if we polarize if goes out and tries to collect the same amount of money It is it's not so easy for us and this needs to be changed in Europe So we need to have more capital flow We need to have more risk appetite in here because if we are missing this revolution And that means we need to invest in it then we have a big big problem on my opinion One thing I was just going to comment on that machine is what it all comes down to off-take is what it comes down to if you have a client We have there's no problem then actually getting finance first off infrastructure financing So I suppose I'm trying to get my head around like if I'm let's take the example Deutsche Telekom they have two systems Right so why aren't they do more? I mean sorry they have the customers already I'm trying to get my head around that I can understand who is a startup But there's a whole pile of big existing system integrators out there. So why aren't they doing this? First of all that the European Union is different. We are mainly driven by SMDs and SMBs They can't sign with you like a five years contract on several millions They need to start now somewhere means they're gonna have short-term contracts and these short-term contracts and these let's Not exactly knowing how much consumption they're gonna have in the next let's say one two years Will lead us to not having the financing option here and the big corpse and there at the moment looking on prem solutions Or they have no fast solution because what they want is they want to have compute now So means they're also working then with hyperscale at the end because we don't have the infrastructure So of course we are trying our best to convince the market and the capital market that is a good idea to invest into Compute today even if you don't have a five years contract because consumption will come and it will come in a hockey stick and not in a slow-grows Scenery and if you don't have an offer and if you don't have the infrastructure and you don't have the compute Then you have no chance to serve this customer and the customer will then find another solution Michelle well, let me just thank you for coming on the show. It's been great really to have this discussion with you Thank you for having me. It was a pleasure Ciao, I've set a portfolio on that we're going through the biggest industrial revolution in history of mankind and Europe is struggling to catch up. We need entrepreneurs like Michelle to enable us to do that Okay, John. I'm gonna try to take a Scottish accent which is gonna be a bit difficult and to quote Sean Conorine the intouchables Europe's bringing a knife to a gun fight I was actually I don't even think it's a knife. I think it's a button because a knife is sharp I think you're right on that shit. What I learned is that the LLMs are getting really commoditized So I wonder why they are putting so much money into developing those new LLMs Considering the one who are open on the market now almost the same quality of humans later What was of course much more important was the attitude of the US government? So you have that thing called the cloud act which basically the US government can have access to every data And of course they can block as well. We've seen the recent blocking of entropy the fabled 5 AI model So having a European base is absolutely critical Yeah, but I want to talk about those financial models I think the financial models by the nation are foundational they're the basis for everything that comes on top of them I know the European approaches let's focus on the applications that can be built using those foundational models Well, there's an issue with that and I give you a really simple example There was a whole pile of businesses across the world that have come out to help lawyers across the world Use AI most of them have actually been based on cloud And then guess what? Cloth for legal comes out. Mm-hmm So the question I have to ask myself where is the power? is the power of the foundational model guys or the guys an application. That's the concern I'd have from European perspective as we just don't have those foundation about them. Now I know we are moving open source and they start the other thing, but still, yeah, we have a lot of work to do in Europe. That's what I would say. It is still a gigantic bet if you look at the cash furnaces that are all those investment those models. Big tech used to print money, now they burn it and all their cash flow is gone and they are doing equity raise and they are raising debt of balance sheet leverage SPV. They are trillions, trillions committed to AI. It's a ginormous bet. We're in a revolution. In revolution, there are going to be big winners and there are also going to be big losers, right? That means also from a capital perspective, the same thing. Go back to the start of the 20th century. If you were producing horse and carats, you were gone. That's the world we're in today. It really is, but it's bigger. I think the other thing that is clear is it's the speed of change, which is really, really difficult. I think both of us agree that there's a big bubble here, but exactly when it's going to burst and how it's going to burst and who's going to win, that's not clear to anyone. In fact, I might have an opinion today and actually I'm changing my opinion tomorrow and then I have another one the day after that because the speed of change is just so breathtaking. I think we agree on one thing, Laurent. My God, it's a good time to be in the electricity space because whatever happens, you need electricity. It's time to thank Michele for coming on the show, fascinating conversation and to thank the BMW Foundation Aberquant for inspiring us those episodes, so more AI than energy, but still everything is linked. Absolutely. Good. I look forward to seeing you next week. Okay, Ciao. Cheers.