235. European Sovereign Neocloud - Jun26

1 - >

Speaker 1: With a round segle and from London and Gerard read 2 - > from Berlin, this is redefining energy. 3 - >

Speaker 2: Today. On redefining Energy, Child, we're going to talk about 4 - > European sovereign neo cloud. 5 - >

Speaker 1: Yeah, which is obviously a critical topic in this AI 6 - > revolution that we're currently at the beginnings of. 7 - >

Speaker 2: Yes, job, because when you see our investment right now, 8 - > it is the biggest capital expenditure cycle in history. We 9 - > are seeing statistics that the kapex is going to reach 10 - > nine percent of global GDP, never seen before the peak 11 - > of the railroad boom in the eighteen eighties or six 12 - > percent of GDP. And the current AI data center built 13 - > is bigger in the US than the rest of the 14 - > construction company. So we are really in extremely out sector. 15 - > But it's very US and when you're in Europe you 16 - > have a different point of view. 17 - >

Speaker 1: Well, listen, I think we be really honest with ourselves. 18 - > What's happened in Europe is AI regulations have blocked the 19 - > development of AI companies across Europe. So we don't have 20 - > any foundational model companies like the Chatchbts of this world. Yeah, 21 - > and we're playing catch up, and I think there's a 22 - > growing realization that actually, you know, one of the advantages 23 - > that Europe has in particular businesses is they have data 24 - > and do you certainly want that data sitting in another country, 25 - > And even if it's not in another country, you want 26 - > to make sure that that data is not leaking out 27 - > to the rest of the world. There's also growing concern 28 - > that you don't want to be just sticking with all 29 - > your data stuck on AWS or Google or any of 30 - > these other big American players. We need to really make sure, 31 - > as I said, that your data is your data. 32 - >

Speaker 2: Absolutely so. We have a supper guest, Michelle Butwelle is 33 - > the co founder and CEO of a German complete called Polarize. 34 - > Polarize is one of the few European what we call 35 - > neo cloud and is a partner of Envigad and his 36 - > mission is to build sovereign AI factories high performance across 37 - > Europe and. 38 - >

Speaker 1: Ron I think what we also want to do is 39 - > we want to thank the BMW Foundation Herbert Kun for 40 - > their support and those of you who don't know the 41 - > foundation what they're all about as uniting leaders across sectors 42 - > to develop solutions that foster innovative economy and a future 43 - > proof society. A key focus is the energy transition and 44 - > climate change, where the Foundation drives international collaboration to accelerate 45 - > the energy transition. Obviously, with rising energy demand from AI 46 - > and data centers, new partnerships, effective collaboration, exchange, a science 47 - > based solutions and strategies are not just the samsual but 48 - > like they're critical really going forward. 49 - >

Speaker 2: Yeah, for our listeners, we had to learn what the 50 - > neo cloud company is because most of the time we're 51 - > outside the data centers, so I'm more like energy in 52 - > real estate, but here we are going inside the data center. 53 - > So a real definition I just found a neo cloud 54 - > company is a vertically integrated AI era infrastructure provider that 55 - > controls the land power, data center build out and GPU 56 - > cluster end to end offering hyperscal a great compute without 57 - > being a traditional cloud like a WS or Azure. So 58 - > the type of company which are now clouds in the US, 59 - > you have Core with Amnhibius, Cruso, and so they are 60 - > different from what they call the powerkshell the whole one, 61 - > the aliks the advantage the cyrus one aligned data centers, 62 - > you know, which provide the infrastator from the outside, and 63 - > of course you've got another layer the original one, the 64 - > traditional collocation reads Equinique Digital Reality, which are multi tenant. 65 - > It's more retail wholesale focus. And then you've got the 66 - > eper scalers and on the top of that a lot 67 - > of money. The Blackstone, the Kickr, the Brookfield, Stone Peak 68 - > Digital Bridge, they're all financing this absolute crazy build. We'll 69 - > see what you're up as to reply to that US push. 70 - >

Speaker 1: So let's bring Michelle on the show. 71 - >

Speaker 2: Michelle, Welcome to the show. 72 - >

Speaker 3: Thanks for having me. 73 - >

Speaker 1: Well, Michelle, maybe i'd like to kick off here now, 74 - > and I'm going out with I spose a thesis that 75 - > really were in the fifth Industrial Revolution and this is 76 - > the biggest industrial revolution the history of mankind. And there's 77 - > two things. One is it's AI and AI needs alecturously 78 - > and actually a funny way, electricity needs AI. I'd really 79 - > love to firstly ask whether you agree with that view, 80 - > and then secondly just talk about Europe's role in this 81 - > revolution going forward and how you see it. 82 - >

Speaker 3: Yeah, I totally agree with that, and I would even 83 - > go a little bit further. It's not just an industrial revolution, 84 - > it's also like a revolution of the society because we 85 - > have to adapt to all these things which AI brings 86 - > to us. For example, it changed the way how we 87 - > work in the future. It changed the way how we 88 - > it's gonna spend time, you know, always work in silence 89 - > of forth. So that means yeah, I'm totally agree to that. 90 - >

Speaker 1: My second part to the question then really is with 91 - > in and around Europe. What's Europe's role now, because again 92 - > stand back and look at it, I see us doing 93 - > a huge amount, particularly in and around learning for national models. 94 - > China going a slightly different approach, which is a more 95 - > open source, and then I sort of see Europe and 96 - > I wonder what Europe is actually doing. 97 - >

Speaker 3: This is the problem because Europe is not doing much 98 - > at the moment. The biggest issue is if we are 99 - > not starting now, we are missing the train. Because you 100 - > see what is happening in America, you see what is 101 - > happening in China. They're taking a major role in this 102 - > AI business. And it means as soon as the European 103 - > economy starts to adapt to AI, it will it needs 104 - > a choice, and if there's no choice from a European perspective, 105 - > then they will go to an hyperskiller, to an American one, 106 - > or to a Chinese one, and they get sticky and 107 - > then they can't go back anymore or not so easy 108 - > back anymore. And it's not just about sovereignty. Is also 109 - > just it's about how we're gonna give out our datas 110 - > for free to these models to learn to replace us 111 - > in the future. 112 - >

Speaker 1: And this is a real threat. 113 - >

Speaker 3: And in my opinion, we're doing a big mistake in 114 - > Europe because we are trying to replicate things from America 115 - > like we did in the European Union. There's a tender. 116 - > The tender says we want to have one hundred thousand 117 - > chips in one giga factory. And to be honest, I 118 - > think this paper was made by a lot of lawyers, 119 - > but no technical guys sitting there and explaining what is 120 - > really needed in Europe. 121 - >

Speaker 1: So Michelle jumping on that, it's tough because let's go 122 - > back to the fact that it is a revolution that 123 - > we're going through. And I be critical of the European 124 - > Union too, because when I look at their AI Directive 125 - > and all this type of stuff, it just slows down. 126 - > You're trying to regulate something that is changing too quickly 127 - > and we're not going about it in the right way. 128 - > So I'm totally witcher and I suppose the question then 129 - > to ask is well, what should we do? 130 - >

Speaker 3: First of all, we as polarized trying to build something 131 - > which is not existing at the moment. In Europe, it's 132 - > called the neo cloud, so means a cloud who's specialized 133 - > in AYE for a eyepurpose like AI infrastructure. So we 134 - > need to enable European companies, startups and solence of force 135 - > so that they are starting to build something that they're 136 - > going to invest heavily into data center, into AI. If 137 - > without that, we can't catch up anymore in my opinion, 138 - > So we don't have to regulate everything beforehand, which we 139 - > need to have some form of sandboxes where we can 140 - > start to work with. Startups needs to be supported, scale 141 - > ups needs to be supported. We don't have to look 142 - > just at the big carps. We have to look like 143 - > a mixture of everything. And if we're not starting now 144 - > to trust in ourselves means that we're going to lose. 145 - >

Speaker 1: On the long end. 146 - >

Speaker 2: In my opinion, do I understand it, and that's coming 147 - > from somebody who doesn't understand much, is that you've got 148 - > to type of AI work. You've got the work where 149 - > basically you run those big models. I have no idea 150 - > how it works, but it looks like it consumes a 151 - > lot of power. And then you've got what they call 152 - > the inference, which I finally understood what it means. Basically, 153 - > it's the asset management. Once the models are running, so 154 - > do you concentrate on the first and on the second. 155 - >

Speaker 3: So we are concentrating on the second because maybe I'm 156 - > gonna make it a little more understandable. Large language models 157 - > needs to be trained first, so that means we have 158 - > to raise the child. We have to give them the 159 - > ability to speak and to think. This is let's say 160 - > training in the world, and we don't have so much 161 - > foundation model, so we don't have so much training in Europe. 162 - > So this is done by OpenAI and by all the 163 - > big guys. And the inferencing is like AI in operation, 164 - > So that means as soon as you use AI, you 165 - > are implementing into your workflows and it starts to do 166 - > something for you. Then it's called inferencing because like consuming 167 - > constantly tokens, and tokens is they your form of energy? 168 - > So our computers are transforming energy into tokens. The tokens 169 - > are consumed by the models, and this is how it operates, 170 - > and we are specialized in inferencing, and because what we 171 - > think is that we don't need to have so much 172 - > foundation mode, because there are a lot of open source 173 - > models out there. We can use these open source models, 174 - > we can fine tune them, we can put them on 175 - > an infrastructure which is sovereign, and then we can start 176 - > to use it in Europe or European companies can use 177 - > it because we need it, because otherwise we can't compete 178 - > with other companies in the world who have this advantage 179 - > of the eye. So we have two choices. One choice 180 - > is we go with foundation model, which are proparatary models 181 - > from the Googles of the world, and then that means 182 - > we're giving the datas into this these models are learning 183 - > from it, and then it could be in the future 184 - > it could be possible that they're going to replace work 185 - > which was done here at European companies. Or we can 186 - > use open source model, we put them on a European 187 - > based infrastructure with European companies and then it has the 188 - > same effect and then the companies can use the I 189 - > as well. 190 - >

Speaker 1: And this is what we're doing in Polarized Michelle, I 191 - > just want to go back to open source here again 192 - > because if I look and I see all the tests 193 - > and studies that have been done. Now, if we went 194 - > back three years ago, there was like a year's difference 195 - > in terms of the quality of an open source one 196 - > versus a proprietary system, but now it's like one month. 197 - > And so am I wrong to look at it and say, 198 - > who cares if it's proprietary or not? If it's open source? 199 - > Is a way to go going forward? Which would mean 200 - > if I'm a European and I want to catch up, 201 - > I really want to be the fast follower. I don't 202 - > want to be the leader the US who's going to 203 - > do and all this spend, all this money and all 204 - > this foundation. Might I just go and take these open 205 - > source models and I go directly into inference. Am I 206 - > thinking about this right? Wrong? How do you think about it? 207 - >

Speaker 3: You're absolutely thinking in the right way. So I think 208 - > that the open source models has catched up so fast, 209 - > and they are pretty good, and you have so much varieties, 210 - > and you also have to think how much performance do 211 - > you need? How much parameters are really necessary? If you 212 - > look at the biggest foundation models in the market and 213 - > you see the parameters, I would say in OPO's four 214 - > point zero is not necessary for ninety nine percent of 215 - > the people who are using EYE at the moment, So 216 - > that means you don't need this super super super high 217 - > intelligent parameters. But you can go with open source. You 218 - > can save a lot of costs and you can also 219 - > save your datas and it will be at the end 220 - > of the day, the open source model will your personally 221 - > EYE model in my opinion, and that is super interesting. 222 - > So that means, yes, I'm totally following what you. 223 - >

Speaker 2: Say if I understand. Well, you say you're a Neil cloud, 224 - > so I know because that's kind of funny. Michael Intrator 225 - > used to be my carbon broker twenty five years ago. 226 - > Well it's not very well the sea of core weave now, 227 - > so you kind of a European core weave. Correct, correct, Yes, Okay. 228 - > What we do know from the energy guys is basically 229 - > what Rowan is doing created by Quinn Brooke, fantastic company, 230 - > but the others it's basically all outside the data center. 231 - > So you know, they arrived, they put a terrain, they 232 - > build the all infrastructure, and basically they deliver you or 233 - > they deliver somebody a building which is ready to be 234 - > powered and cool and everything. And that's where you come 235 - > in explain a bit what you're doing. 236 - >

Speaker 3: No, it's in fact, it's not where we are coming. 237 - > We'recoming a stage earlier. Because if you look at the 238 - > actual data center industry in the world, and it's mainly 239 - > let's say cloud business like storage and so on and 240 - > so force, it was like that the infrastructure, so the 241 - > real estate building was thought from real estate guys, and 242 - > then the computing which was implemented there was staying in 243 - > the same shape for years, for maybe ten fifteen years. 244 - > So inside the data center you had like a real 245 - > estate part, and then in the in the real estate part, 246 - > we had like a standard of some form of it 247 - > called rex. It's like shelfs and then these shelvels had 248 - > around fifteen to twenty kilowat each shelf today and we 249 - > are seeing it the other way around because what we 250 - > have understand and this is why we are building the 251 - > data center ourselves, because at the moment, there are no 252 - > co location data center which are suitable for the II infrastructure, 253 - > so to say, because you have to think backwards, you 254 - > have to think from the it back to the real estate. 255 - > And there's a lot of differences between the old data 256 - > center economy and the old data center economy. 257 - >

Speaker 1: It will not be ending. It will stay as it is. 258 - >

Speaker 3: We're still going to consume data is We're still going 259 - > to have Internet, We're still going to have storage, and 260 - > then we'll also have a gross every year. And then 261 - > there is a new form of data center. These are 262 - > the AI factories, So we are not calling them any 263 - > more data centers. We are calling them AI factories, and 264 - > these AI factory are totally different constructed. First of all, 265 - > we are not going with fifteen kilowatt uperrac. We are 266 - > going with up to one hundred and fifteen more. And 267 - > we have to think not just about the generation of 268 - > compute today, we also have to think about the generation 269 - > of compute of tomorrow because the iteration of compute is 270 - > going so fast. So every GPU model it will be 271 - > replaced in the next three to five years. So we're 272 - > going to have an efficiency every three to five years 273 - > in GPUs, and we're going to consume more energy per 274 - > GPU in the same amount of time. It always doubles 275 - > the town sometimes even triples, and that needs to be 276 - > sought as well. So if you're going in a classical 277 - > data center, and if you don't think about all these 278 - > different elements you're going to be outdated super fast in 279 - > the data center world. So that means you need to 280 - > think differently in THEI factories than you did in the 281 - > past with normal data centers. Normal data center are not 282 - > suitable for AI, and that means there's a totally new 283 - > classification of data center coming into the market with a 284 - > big hunger of energy, but with a lot less space. 285 - > So for example, we are putting per square meter a 286 - > lot more energy than a normal classical data center, So 287 - > it means we can build in two thousand square meters, 288 - > we can build twenty twenty five megawats. 289 - >

Speaker 1: Very interesting you're saying about energy there. But I'm going 290 - > to ask now, just coming from the energy side, when 291 - > you look at data centers going forward, are we going 292 - > to build big, huge ones or are they going to 293 - > be small? In other words, are we going grade edge? 294 - > And if it is grid edge, jesus fifty megawats, twenty 295 - > megawards five one? How do you see the future in 296 - > Europe of building these data centers or sorry AI factory, 297 - > sorry AI factories, no whorries. 298 - >

Speaker 3: So I see it in the same way like we're 299 - > producing energy today, so we have like big new power 300 - > plants in Europe. We have smaller coal plants, we have 301 - > gas plants, we have windmills, We have a lot of 302 - > different power production all over Europe, which is necessary because 303 - > you need sometimes a really big power plant to produce 304 - > a lot of energy because you have a region, we 305 - > have big cities and so on, and the force whereas necessary. 306 - > I see the same in AI. So of course we're 307 - > going to have one hundred megawatter one hundred and fifty 308 - > megawatt big AI factories, but we're going to also have 309 - > like twenty megawat factories. We're going to have fifty megawatt factories. 310 - >

Speaker 1: And what we think, we. 311 - >

Speaker 3: See it as a and we call it giga grit 312 - > in polarized. So we don't need these big centralized single 313 - > point of failures in my opinion. So what we need 314 - > is all the mixture of different form of AI factory 315 - > which fits into the region and into the market. For example, 316 - > if you're living in Munich and we have one data 317 - > center or one AI factory in Munich where we are 318 - > running ten thousand GPUs, it has around twenty five megawatts, 319 - > so it's it's in terms of data center as standard 320 - > data centers quite huge, but in terms of AI factories 321 - > is pretty small and with this ten thousand GPUs, we 322 - > have doubled the AI capacity in Germany at the moment. 323 - > That means at the same time there is a lot 324 - > more hunger for GPUs and AI factory, so we need 325 - > to build them fast. And in my opinion, there's a 326 - > lot of options in the market if you want to 327 - > go fast, because if you build a two hundred megawat 328 - > or even bigger AI factory, it takes time because the 329 - > grid connection is not so easy. You need to find 330 - > the right point to connect, then you need to build it. 331 - > Then you need to have all this residancy. You need 332 - > to have a lot of things to build these huge factories. 333 - > But if you find an old production fabric which already 334 - > has twenty or ten or fifteen mega what's in place, 335 - > and you are capable to retrofit it, and you have 336 - > the creativity how to transform a former whatever factory it was, 337 - > into an AI factory and using the same infrastructure which 338 - > was already there, and you just upgraded, you just put 339 - > it in a way that is suitable for an AI factory, 340 - > and then you have a very fast and a very 341 - > efficient AI factory. And it's also in my opinion, so 342 - > that you don't have to build everything from Squattionally you're 343 - > saving a lot of cou two for example, and so 344 - > and so forth. And we have in Europe and especially 345 - > in Germany, we have so much potential sites and we 346 - > see that every day which could be transformed into an 347 - > AI factory and being a super efficient and a super 348 - > fast way forward. So there is enough options, but the 349 - > option needs some creativity from companies like US or others. 350 - > And if you just go the normal way like we 351 - > did it in the normal data center world, for example 352 - > in Frankfurt. So we have in Frankfort a lot of 353 - > data centers because we have there the International and the 354 - > Transatlantic Internet KNOT and there's DCAKES which is one of 355 - > the biggest Internet distribution not in the world. This is 356 - > why everybody wants to be in Frankfort. But if you 357 - > want to build a data center at Frankfoot, it will 358 - > take you at least six seven eight years until you 359 - > get the grid concert and it will cost you forty 360 - > percent more than other regions, and that will be too 361 - > late in my opinion. If you're going to go the 362 - > same root. 363 - >

Speaker 2: It's a very interesting approach. It's a very German approach decentralization. 364 - > But if you cross the border and you are in France, 365 - > where everything is stopped down, decided by the president, the 366 - > only thing they're interested is to sign seventy five billion 367 - > with a soft bank. Who's going to deliver or not? 368 - > Is it compatible? Or who's right, who's wrong, everybody's right? 369 - > How does it work? 370 - >

Speaker 3: It is compatible because if you look at the reality 371 - > in France, the grid connection times is even longer than 372 - > sometimes for example, if you look in Paris region or 373 - > Bordeaux region or other bigger regions, it's a nightmare how 374 - > long it takes until you get a real data center 375 - > on the ground at the end of the day. In 376 - > a more centralized country like France, they're going to have 377 - > the same promise and there are the same options. You 378 - > also have a lot of industries which are not performing 379 - > and in the same way, which can be retrofitted into 380 - > two AI factories. And we see that all over Europe, 381 - > so in Spain, in England, and we have projects in 382 - > Spain we haven't we have also a project in England. 383 - > The grids in England, it's also like a very bureaucratic 384 - > and not an easy way to go. So that means 385 - > We have that all over Europe, the same problems, the 386 - > same kind of grid constraints, and if you look at 387 - > Scandinavia where it's a little bit better, but still also 388 - > Scandinavia is now catching up in bureaucratic processes and not 389 - > being so easy anymore than there was in the past. 390 - >

Speaker 1: Michelle, I like your approach and I think, okay, yeah, 391 - > using existing infrastructure makes a lot of sense to me. 392 - > Talk a little bit about the client perspective, In other words, 393 - > why would a client go to you rather than go 394 - > to Google or Microsoft or someone who they trust and 395 - > they know for many years. In other words, what is 396 - > it they differentiates you from them? 397 - >

Speaker 3: I think there are three reasons. First of all, we 398 - > are not more expensive than taking a hyperscilla. Second, we 399 - > are sovereign, means all our clients are protected against the 400 - > US out So that means that a government can decide 401 - > to look into your datas, they can even decide to 402 - > shut you down. And that so is we have no 403 - > interest in your data. We don't drive our own models, 404 - > we don't use your data to learn from it, we 405 - > don't feed our own machine. So that means everything stays 406 - > with the client and you can be sure, of course 407 - > they are at form of legal frameworks where you can 408 - > protect yourself, but at the end you don't know how 409 - > the hyperskillers are using your data for their own purposes 410 - > and own good And we know in the past Amazon 411 - > had a marketplace, so everybody could bring his services and 412 - > goods to this marketplace and sell it over Amazon, and 413 - > then Amazon learned from it what are the best products, 414 - > which products can be sold very fast, and which are 415 - > super interesting to sell, and then they sold it themselves. 416 - > So why should they not use your company data to 417 - > train and to understand how they can make money with this. 418 - > This is something every company in Germany, in Europe needs 419 - > to think about that. So this is why the usp 420 - > of Polarized is clearly that we have the same quality 421 - > of service in AI. We have no intention to use 422 - > your data for anything, and we are fully compliant to 423 - > the European Union laws to GDPR and we have nothing 424 - > to do with the US Cloud Act. And I also 425 - > have to say, of course, we as Polarized are not 426 - > so big at the moment, so we are trying to 427 - > do our best. But what we do, for example, is 428 - > we do some form of partnerships, for example with Telecom, 429 - > the German Telecom who are building the trust layer, who 430 - > are giving us a lot of potential into the market, 431 - > but also with others big European companies or potential cloud 432 - > champions in the future, like the Schwartz group, where we 433 - > are delivering the layers, they need to repackage it and 434 - > to sell it to their customers. 435 - >

Speaker 2: But at the end of the day, what you decide 436 - > to put the architecture, whether you put upper or Blackwell 437 - > or Rubine or whatever else, that's your decision. Does the 438 - > client to tell you that's the cheap I want? 439 - >

Speaker 3: There is different layers. So we have one layer that 440 - > is a data center, the I factory. Then the next 441 - > layer is the compute means what type of GPUs, what 442 - > type of storage. On top of it, we have a 443 - > layer which is called core, which is like a software 444 - > layer where a customer can just consume virtualized server and GPUs, 445 - > and on top of it we have AIS a service. 446 - > It means we have a platform where we are hosting 447 - > different open source models and you as a consumer can 448 - > directly log into the platform. You can choose your model 449 - > and then you can consume tokens and your pay us 450 - > per token. So if we're selling to a customer the 451 - > GPU layer, which is called bad metal, then of course 452 - > the customer will tell us what kind of GPUs and 453 - > what kind of compute he wants. But if we sell 454 - > it through Core or through our platform, then the customer 455 - > they don't need to understand the infrastructure be lowered because 456 - > they don't need a GPU. Nobody wants a g We 457 - > just want to have the AI in operation to use 458 - > it for two purposes, to make more money or to 459 - > make my company more efficient. These are just the main 460 - > drivers at the end of the day. So that means yes, 461 - > sometimes if we are selling it to a bigger client 462 - > like Telecom, then they choosing the it. They telling us 463 - > what kind of a t they want. But if we 464 - > are selling Core or Drive, then we are deciding what 465 - > kind of a TV you're going to put into the 466 - > data center. 467 - >

Speaker 2: Michelle. 468 - >

Speaker 1: One thing that I see in my work every day 469 - > really is that the people driving AI demand for data 470 - > centers in Europe are all the hyperscalters in the US. 471 - > I don't see any European players actually doing anything. 472 - >

Speaker 3: Yeah, that's correct, And this is the problem because it 473 - > has two elements. One element is that the hyperskillers are 474 - > going to consume most of our infrastructure, so that means 475 - > they're blocking the infrastructure as well because they need a 476 - > lot of energy for their training clusters. And the second 477 - > element is that we are not really investing into it 478 - > because we as Europeans, we are investing in balance sheets 479 - > and we are not investing in ideas. Means the hyper 480 - > skillers they have no problem in buying trillions and billions 481 - > of euros for GPUs and training them models, but they 482 - > also don't know what's coming in the next four or 483 - > five years. 484 - >

Speaker 1: But they have an idea. 485 - >

Speaker 3: In Europe with the big tail coos, the big cloud 486 - > players in Europe, they are not investing at the moments 487 - > so heavy, and this is why everybody is looking at 488 - > the governments and is looking at the European Union to 489 - > subsidize some part of the capital expansion. And this is 490 - > a problem because that makes us super super slow. It 491 - > doesn't give us the real opportunity to grow fast into it. 492 - > And we as a company, we see that. So for us, 493 - > it's super heavy to collect the billion for example or 494 - > one hundred millions, so which is a lot more easier 495 - > if you're an American company with this core weave for 496 - > examples to core weave or n Scale or anibio. They 497 - > have collected so much money and if we polarize, if 498 - > it goes out and tries to collect the same amount 499 - > of money, it's not. 500 - >

Speaker 1: So easy for us. 501 - >

Speaker 3: And this needs to be changed in Europe. So we 502 - > need to have more capital flow, We need to have 503 - > more risk appetite in here, because if we are missing 504 - > this revolution and that means we need to invest in it, 505 - > then we have a big, big problem. 506 - >

Speaker 1: In my opinion. The one thing I was just going 507 - > to comment on that Machell is well, it all comes 508 - > down off take is what it comes down to. If 509 - > you have a client, there's no problem then actually getting 510 - > finance for structure, infrastructure financing. So I suppose I'm trying 511 - > to get my head around, like if I'm let's take 512 - > the example Deutsche Telecom. They have t systems, right, so 513 - > why aren't they doing more? I mean, sorry, they have 514 - > the customers already. I'm trying to get my head around that. 515 - > I can understand you as a startup, but there's a 516 - > whole pile of big existing system integrators out there, so 517 - > why aren't they doing this? 518 - >

Speaker 3: Well, first of all, that the European Union is different. 519 - > We are mainly driven by SMBs and SMBs they can't 520 - > sign with you a five years contract on several millions. 521 - > They need to start now somewhere, means they're going to 522 - > have short term contracts. And these short term contracts in 523 - > these lets are not exactly knowing how much consumption they're 524 - > going to have in the next let's say one two years, 525 - > will lead us to not having the financing option in here. 526 - > And the big corps they're at the moment looking on 527 - > prem solutions or they have no fast solution because what 528 - > they want is they want to have compute now. So 529 - > means they're also working then with hyper skills at the 530 - > end because we don't have the infrastructure. So of course 531 - > we're trying our best to convince the market and the 532 - > capital market that it's a good idea to invest into 533 - > compute today, even if you don't have a five years contract, 534 - > because consumption will come and it will come in a 535 - > hockey stick and not in a slow growth scenery. And 536 - > if you don't have an offer, and if you don't 537 - > have the infrastructure and you don't have the compute, then 538 - > you have no chance to serve this customer, and the 539 - > customer will then find another solution. 540 - >

Speaker 1: Ye well, let me just thank you for coming on 541 - > the show. It's been great really to have this discussion 542 - > with you. 543 - >

Speaker 3: Thank you for having me. 544 - >

Speaker 1: It was a pleasure. Joe, I've shut up a foller 545 - > on that we're going to the biggest industrial revolution in 546 - > the history of mankind and Europe is struggling to catch up. 547 - > We need entrepreneurs like Michelle to enable us to do that. 548 - >

Speaker 2: Okay, John, I'm going to try to take a Scottish accent, 549 - > which is going to be a bit difficult. And to 550 - > quote Sean Connery in The Untouchables, Europe's bringing a knife 551 - > to a gunfight. 552 - >

Speaker 1: I was actually, I don't even think it's a knife. 553 - > I think it's a baton, because a knife is sharp. 554 - > I think you're right on that. 555 - >

Speaker 2: What I learned is that the lms are getting really commoditized, 556 - > So I wonder why they're putting so much money into 557 - > developing those new lms, considering the one who are open 558 - > on the market now almost the same quality a few 559 - > months later. What was, of course much more important was 560 - > the attitude of the US government. So you have that 561 - > thing called the Cloud Act, which basically the US government 562 - > can have access to every data, and of course they 563 - > can block as well. We've seen the recent blocking of 564 - > entropy the fable five AI model. So having a European 565 - > base is absolutely critical. 566 - >

Speaker 1: Yeah, but I want to talk about those fundacial models. 567 - > I think the foundational models by the nation are foundational 568 - > that the basis for everything that comes on top. And 569 - > I know the European approach is let's focus on the 570 - > applications that can be built using those foundational models. Well, 571 - > there's an issue with that, and I'll give you a 572 - > really simple example. There's a whole part of business across 573 - > the world that have come out to help lawyers across 574 - > the world us AI. Most of them have actually been 575 - > based on cloud and then guess what cloude for legal concersions. 576 - > So the question I have to ask myself where's the power? 577 - > Is the power the foundational model guys or the guys 578 - > as an application. That's the concern I'd have from European perspective, 579 - > as we just don't have those foundational moments. Now. I 580 - > know we are moving open source in this, that and 581 - > the other thing, but still, yeah, we have a lot 582 - > of work to do in Europe. That's what I would say. 583 - >

Speaker 2: It is still a gigantic bet. If you look at 584 - > the cash furnace that are all those investment in those models. 585 - > Big Deck used to print money. Now they burn it 586 - > and all their cash flow is gone. And they are 587 - > doing equity raise and they are raising debt. They have 588 - > of balance sheet leverage SPV. They are trillions trillions committed 589 - > to AI. It's a ginomous bet. 590 - >

Speaker 1: We're in a revolution, and revolution there are going to 591 - > be big winners and they're also going to be big losers. Yea. 592 - > That means also from a capital perspective, the same thing. 593 - > Go back to the start of the twentieth century. Yeah, 594 - > And I mean, if you're producing horse and carts, you 595 - > were gone. Right, that's the world we're in today, right, 596 - > it really is, but it's bigger. And I think the 597 - > other thing that's that is clear is it's the speed 598 - > of change, which is really really difficult. You know. So 599 - > I think both of us agree that there's a big 600 - > bubble here, But exactly when it's gonna burst and how 601 - > it's gonna burst, and who's gonna win who's not, that's 602 - > not clear to anyone. And in fact, I might have 603 - > an opinion today and actually I'm changing my opinion and 604 - > tomorrow and then I have another one. The day after that, 605 - > because the speed of change is just so breathtaking. I 606 - > think we agree on one thing wrong. My god, it's 607 - > good to good time to be in the electricity spos 608 - > because whatever happens, you need electricity. 609 - >

Speaker 2: Okay, it's time to thank Michelle for coming on the show. 610 - > Fascinating conversation, and to thank the BMW Foundation a berquant 611 - > for inspiring us. Those episodes so more AI than energy, 612 - > but still everything is linked. 613 - >

Speaker 1: Absolutely good and I look forward to seeing you next week. 614 - >

Speaker 2: Okay, yeah, cheers. 615 - >

Speaker 1: Thank you for listening to Redefining Energy. Don't forget to 616 - > rate the show and subscribe on Apple Podcasts, Spotify, or 617 - > the platform of your choice.