AI Factory CEO Debunks Every Myth About the AI Super Cycle | Nebius Partner Charles-Antoine Beyney

0:00 Our guest today is Charles Antoine Bin, CEO of Data 1, the company behind some of the world's most advanced AI factories, including a 300 megawatt facility in New Jersey built in partnership with Nebius. AI is the largest super cycle the world has ever witnessed. And the single constraint on how fast it unfolds is the speed at which capacity can come online. If there's anyone who truly understands the bottlenecks, challenges, and qualities required to build a world-class AI factory, it's Charles. Welcome to the show.

0:27 >> Hi, thank you very much, guys. Charles, we wanted to start with a reality check. According to Reuters, up to 50% of US data centers are delayed or cancelled. Does that match what you are actually seeing on the ground? >> I will only talk about our our own experience here because I think this is the most relevant. Right. So I can tell you that uh we have some very large support locally and at the same time some opposition. Uh I need to be blunt and and truthful. So um of course like we can see a lot of fear and a lot of questioning when it relates to AI data centers right now um because people don't have the right level of informations and I think this is one of the main thing here is that a lot of push back comes from disinformation and fake news uh related to our industry. Um so I don't know if it's 50%, I don't know what what's the percentages. I can tell you that it's tough and I we can sit in the press obviously as well. Um on our side we need to inform a lot of a lot of people and that and that's the thing because when you are used to build data centers like in the old ways you were not like specially on the hotspot you know you didn't have to prove yourself or anything everybody was welcoming you with open arms. Right now you got a lot of more questioning like what's your water usage? what's going to be the impact on the day center on my electrical bill like will you be uh polluting my water like you know all kind of environmental issues so of course we we have to make education we have to spread the word uh of what is correct you know so yeah definitely we we see a lot more questioning than before a lot of more woriness than before that's clear >> so I think it's important to understand that this is like not normal people that just go out randomly this is there are political motives behind this by people that literally sponsor protesters Because the idea here is the framing is like oh data sense it's like the evil billionaires and then it's against the normal poor little people on the ground that get polluted the water but you have to understand that the sponsors behind those people they are billionaires themselves.

2:26 >> Yes. And and this is where you got multiple uh parties at play in this case because you've got the ones which are the foreign uh governments that don't want to see you as the most smart guy in the in the in the room because obviously more GPUs you have more intelligence you have and then better service you can provide to your companies to your army to your healthcare to wherever you want. So because the applications of AI is almost infinite, right? So more power you have, more powerful you become on the on the planet earth basically. So obviously you got lot of foreign um impact or foreign people that will not be willing to see you successfully achieving your goals in time of like AI uh factory spreading right so this is number one the number two as well is that when you are touching some uh big investment like these and you are an investor in Wall Street uh it's easy to short right so having someone which is stressfully uh constraining you on your like lead times or whatever then you have an edge So for a very cheap price, you just take few environmental uh guys and you push them to say bad stuff about a data center and suddenly you've got massive opposition coming through and then what happens next? Oh, you are slowing down this project six months or 12 months or you cancel it. You your short becomes like very material for you. All right.

3:47 So and you can you can earn like big money in this case because the impact on the balance sheet of this company is real obviously. So, so yeah, of course you got like plenty of a position in this case and obviously this is not like the normal worker which is like in any case don't give a damn honestly because the guy is working 40 60 hours a day a year a week sorry and obviously he's going to be focused mainly on his on his work and on his family and protecting his family and not really looking at what an AI factory impact is honestly because especially these days our impact is close to zero because well maybe we're going to talk about it later but we are generating ing our own electricity. We are not consuming any water. Our water is pure is absolutely pure because you don't want to see any bad water into your GPUs or into your very expensive like um micro electronics components obviously. So everything needs to be 100% pure and so there is absolutely zero impact or close to zero impact and then okay noise. Yes, maybe we get some.

4:48 >> Can we just go through all those points one by one? So water basically you're saying you do not consume water. No, it's it's it's tiny amount of water because of course you need to to load up your your system once just like you know the radiator in your car you just get water inside the radiator of your car and just like cooling down your engine but it's a closed loop right so you don't add up all the time water inside your system very rarely right so except if it's extremely hot outside like literally extremely hot meaning like more than 45°C which is I don't know like 140 or something like this Fahrenheit so in this case yes we can consume during like maximum 100 hours a year but what are we talking like we're talking one two three four five swimming pool maximum at the scale of what we're doing so it's not even a drop in the ocean honestly so there is no question to ask anymore so yes of course you had uh first generation data centers with open water cooling tower which was consuming millions of gallons per day that was that could have been an issue long term right now in AI factories because of the technologies we are employing we are using what we call dry coolers and closed loop system at a very high temperature. So we are consuming close to zero water. So it's not even a brainer. We are consuming way less than even a small tower of office basically.

6:02 So >> but but then is it fair to say that some data centers are still kind of like old school or like not using the newest technology and for those so basically the people should differentiate like it sounds kind of stupid to say but to speak it out simply people should look at what technologies the actually deploy and then decide based on that right but obviously the nuance often gets missed in political debates >> yeah that's exactly the same like if you were saying I've got a car within the 90s or I've got an electrical car of these dates you Am I polluting the same when I'm entering your city? Not really.

6:36 Right. So that that that's exactly the same with the AI factories. This is a new breed of data center with a new breed of infrastructure which is completely uh different in term of technologies at every level. Before this, I was trying to do some research and see what videos of you there are on YouTube. There's two and it couldn't be more contrasted. One of them is you basically showing to investors like how big of an opportunity it is. And the other is you like basically defending this data center that'll use a couple swimming pools worth of water and like having to like promise to the public that like that's all you're going to that's all you're going to need. Do you see like this scientifically unsubstantiated push back? Does that actually lead to real delays?

7:12 >> No. Honestly, absolutely not. And that's the thing and that's where I was saying like about fake news because these people are pushed with a biased which is absolutely not true and and again as we just say it's like water is not an issue in our case and and and I I cannot say more than that like I mean guys you just need to do your homework and just check what we are using in term of like you know supply and return temperatures into our cooling system. This is nothing is rocket science in this case, right? When you got GPUs which are certified like W45, it means that they can have supply of water at 45 degrees CC which is huge.

7:48 That means in this case that as long as you got the right dry coolers then you can go up very high in the temperature without using even any electricity or close to non electricity. So this is super efficient as well and that's all physics you know. So this is like like physically and I cannot re that right. So it's heat dissipation. It's quite basic math. So if people are serious and they can just ask GPT or orac or code or whoever they want and they can explain that to them as well. So especially these days if you really want to have the real source of information with AI, it's kind of easy to have both sides of the of the spectrum, right? So you just take five minutes and you've got a full education on this system. So yeah, one of my favorite moments from when you were defending the the water thing was uh you collect some water through condensation and like this guy was just asking you to promise that like condensation will work like what is your guarantee that condensation promise the pipe >> what can I say you know like it's it's okay. So yeah that's that's the other thing as well. So the the we got multiple ways uh in data one because sometimes we're connected to the grid especially in France because in France we have like gawatt of power coming from nuclear and coming from renewables. So we don't even need to get our own power behind the matter. It takes more time though, but it's it's doable. And and in this case, we are cons like as I was saying, we're going to be storing a little bit water on the site just in case it's a very hot summer, which is very rare, but still we we need to cover this kind of risk. So we got like a small tank on site just in case uh because we're going to be collecting the water when it's like, I don't know, fall or spring or whenever when you got like lots of water in the system. Anyway, it's not a problem for anyone. um or in the US when we have like two ways. So either the grids but it's not it's not the case right now especially not in violent we're off grid. So in this case we generate the the electric the electrons ourselves and as I was saying in my previous videos is like we have engines with very high temperature exhaust. So in this case we we can slow down the temperature exhaust because we are using like what we call absorption shooters. So absorption shields works with heat basically. It doesn't work with electricity. It it works with heat because you got like a chemical reaction which is giving you a chilled water on the return. And then because you slow down this temperature at one point you you just uh catch what we call the due point. So this is where the condensation happens right and when you condensate that much uh heat then you can have like a lot of water in the system. So you're just extracting a little bit like I don't know not even half a percent of of uh of condensation in the air around your stuff and it's already millions of gallons that you are generating every year and especially when you are living at 60 miles not even from the coast you can imagine that water is not like is everywhere in the in the in the air anyway so this kind of difference doesn't make any difference right so this is how we are generating water in this case but I will tell you I have invested and I'm the main sponsor of a company called CLA which is a French company which is doing extensive research on um seaweeds, microorganisms etc. which is capturing carbons or CO2 all kind of pollants and transforming them into something very valuable. It can be biomass, it can be biouels for sustainable avation fuel, it can be acidalic, it can be all kind of different stuff that you can use for cosmetics or for food or whatever. This is really next level stuff.

11:10 >> You don't even need natural gas. Oh no, I I do need natural gas. Yes, at one point I need something that I can burn or mix that like Yeah, because physics works like this. So yeah, no, we we need this at least. Uh so but we are using the natural gas in a way that we are like not generating knocks or vox or um like all kind of bad pollutants you don't want to breathe because our mission is to be at net zero and even longer term mission is to be above net zero meaning we want to uh clean the air around us more than what we are degrading is it's very funny the disconnect between narrative and reality. Uh it's like if the people have a problem with air pollution and you're going to be above net zero, they should they should actually be welcome you.

11:53 They should be paying you to clean their air. >> That that's exactly the point. And if we're not using everything by the data center, we're going to resell that to the local grid at the local rates and even lower than that. So we're going to be able to even maybe decrease the price of the elections for the local folks. Right. So in this case, this is the opposite the full opposite of what you can learn learn or read in the press.

12:14 And I guess that's you just made the perfect transition to uh behind the meter energy and because the narrative is still I know this is silly to you but I want to like for the viewers to help maybe spread the message about the reality. So the narrative has been the data center come they they steal our like they steal our electricity we cannot afford to heat our apartments. So the reality is I think that current administration made it quite clear they are pro AI they're pro innovation they're pro data centers but you have to bring your own energy can you explain that quickly like can you maybe explain like the direction that the current administration is taking the industry >> at one point when you got an industry which is sucking like half of the country power even more if we believe the trends the long-term trends that everybody is trying to push which I'm not convinced about honestly but that's for another day. So when you see this kind of impact on a grid which has not been upgraded for many years and on a system which is very old honestly in the US that's the case uh obviously it's demand and offer right so you got very powerful industrial people with a lot of money that will be able to pay 15 cents 20 cents per kilowatt hour even so but the impact on on the local guys is massive in this case and of course if you see your bill getting from I don't know $200 to $500 because of uh freaking data center you are super pissed right so that makes sense I mean there is nothing to say about that so obviously when you've got this kind of industry at that scale you have to ask the guys hey take responsibility and be human at the same time and then just do your stuff and you know back in the ' 90s uh on the 20th century 19th century all the big factories they had their own poor plants because every single big factories had to come with its own like uh gen sets basically or generators to generate enough steams or enough energy for their own factory because there was no grid.

14:11 There was no such thing. So it's like something we had already many centuries ago that we forget in between but that is just coming back to us because of the scale of what we are doing right now and that's not even a surprise for me. So that's why when we saw that the grid was not here, we say, "Oh, we got pipeline. So let's do our own electrons." And that's even better because in this case, you can ensure a level of SLA to your customer which is way higher than what a usual utility company will be able to offer you anyway because you are you are mastering your level of redundancy.

14:41 You're engineering from A to Z, etc. You know, for for instance, for us, if we got like a pipeline issue, we got LG tank on site where we're going to be able with four vaporizers to continue to feed the sides. whatsoever unlimitedly until the pipeline is repaired, you know, and that's the same if the the the local folks they need the gas for their own usage, they are prior to us. So the the the South New Jersey gas can say, "Hey guys, I'm sorry. I need to interrupt you for five or six or 10 or 10 days because it's too cold out there and we need more gas for the local farts." And in this case, we got truck on standby coming from another state that's going to refill our our LG tank in order to sustain the load of the data center 247. And we got redundancy embedded into this LNG tank as well. So that's that's how you're building something which is very responsible with close to none no impact for local folks whatsoever. But that needs to be thoroughly thoroughly uh sought out since the beginning because if it's not properly designed then it's never going to work. But for us it's been embedded into the vision since day one basically and and that's how we are doing like like sustainable engineering as well.

15:50 >> You also built data centers in Europe right recently France actually quite pro data centers like allowing data centers to like tap into the grid of nuclear. Does the power strategy change that much from US to Europe? >> This is mainly different. Yeah. Because we are a lot more centralized and the grid in France uh is like miles away or million miles away from the quality of the grid that we have in the US. >> As in much better. Yeah, it's much better. Like uh 100 times much better.

16:14 You don't see any >> 100 times. >> Yeah. Yeah. It's massive. We have like zero down time basically. You don't see blackout. You don't see downtime in France. It's very rare. I mean like it can be very localized because you got a transformer which has blown up. Okay. That can exist obviously, but it's very very localized on the on the national grid. It's redundant at every level. It's insane the level of redundancy and over supply we have because we got like more than 63 gawatt of of um uh nuclear plants that are running in the system plus all kind of gawatts of like solar and and all all kind of things and at the small scale of of a French country with only 70 million people you imagine this is a lot of power basically so we are even reselling power to Germany to Spain to all the neighbor country so yeah that that's that's a different stuff so in this case you can rely on the on grid it's fine because you know that you will never have like more than 3 minutes of interruption per year and even so you know so if you got like 10 or 15 minutes of batteries on your site you can sleep yeah you can sleep on your two ears especially if you got like two substation feeding you and that's that's how we are setting this and in France what we do is that we are paying slightly more in order to have the certainty from the national provider that we have 100% of green energy so all the energy we have is coming from uh the hydro dams uh from the halves or from solar panels or from wind turbines. So it's a mix of like renewables in this case and we got 100% certificate of green energy but that's because we are sustainable at core. There is no obligation for that. It's just like because this is our philosophy in data one. It's very particular to our company basically.

17:48 >> Charles I think you freaking crushed and nailed that part. I do believe the viewers the people that invest in data centers will be excited to hear this to be honest. >> Yeah. and and even and just if I can go back to to the gen sets as well because we got amazing gen sets. We signed most probably the best gen sets for for this industry uh with like a super nice uh uh technology behind them to stabilize the the the micro grid against AI load and everything. So the engineering is absolutely flawless. The full engineering we had with the retrieval of all the you know the exhaust system that we've developed to be close to net zero and everything because this is honestly perfect and it's all publicized by the way. It's all public. If you go on to the NJPP website, all our engineering is open source. So we have opened up everything we do uh for people to see that we are really intending and we are very serious on on the fact that we are like killing 99.5% of all the poly basically or close to it. I guess it's coming down now to speed to power, right? Like how quickly can you how like how much can you deploy and especially like how quickly because recently there were like big headlines about entropic and their official languages. The reason why they're so short compute is because people said 10x would be crazy, but now actually turns out the demand was 80xing. And so they did crazy stuff like they they lowered the limits. They increased the pricing. They even did stuff that I thought was unimaginable, which is they didn't allow new users to sign up. Like which company grows so quickly that they literally are like, "Guys, we don't like don't join like because it's going to degrade the quality of our other users because we have to split the compute across them, right?" and open everything open says the same and obviously now with video with video AI with real time image generation coming that's going to get much much worse and then if you have physical AI and like basically everything in the world it's moving that's going to be like a a nightmare or like a dream for you right because of the demand coming in so the question that I would have is if the demand is so massive what stops us from just throwing as much money as we can at the problem to speed up the time like what are the kind of bottlenecks that pure money because we see like literally hundreds of billions going into this. What are the bottlenecks that pure money cannot solve? So many so many I I've got like a list which is almost not of course it's limited by nature but it's close to unlimited when you look at this because it's you know when you got a data center it's so many different industries intricated in each others which is like absolutely mind-blowing honestly that goes from of course the gas itself or the way of powering up at the very basic level your your infrastructure up to obviously the generator or micro turbines or turbines or whatever you want to use. So all these is already industries by themselves of course and they have bottlenecks at every level because they have a certain throughput for all of them. But then you go up and up in the chain you talk about transformers, switch gears, busways, UPS, batteries, all this kind of stuff use rare earth elements or like all kind of different things that you need to to power them up. And then you go even above in the chain you talk about GPUs, microchips and all those kind of things.

20:58 And then it's other struggles, you know. So on every single layer of this industry, you've got multiple interconnections with hundreds, thousands of companies interconnected to each others to solve these issues together. And that's a very high complexity. And if only one of these bricks falls apart, nothing works. Nothing. Right? If if I miss one gen sets, you can have all the GPUs inside your data center. Doesn't work. If you if you have like just helium which is missing into your GPU process, you are not able to print your GPUs anymore and Helium is a big topic right now especially with the the stuff that has been happening into Qatar. Uh you know so that you have so many different disruption aspects that can arise. Are you confused by the fact that kind of like as of today the straight I mean it was open for like I think one day and then it was closed again and people talked about helium shortages also like to the big HBM like like highex and and other but that's a real topic >> but like it's it's it's still closed and those companies just like you know what I don't care and they're just ripping they're like they have doubled like they have I think Esa Hanks has doubled from the >> but but the reality has not changed >> demand and offer because the the demand will be there the offer will be very low. What does it means? Price increase, right? We are just at the very early stage of internet right now for the AI.

22:18 So we are the early adopters. Most of the people they are still not using it. The majority of the people are still not using it. So imagine once you're going to have like 80% of the people connected to AI like what's going to be the impact on demand? It's going to be even more. And and as you were saying I I don't believe that video will be such a thing right now because they will limitate a lot video because the ARI on video is so slow so small that like they will be putting some heavy limitation but anything which is related to um personal uh um models like healthcare model all these kind of things that's going to blow up everywhere of course and and yes that so that means we're going to be in a scarcity environment or a scarce environment for the next 10 to 15 years even. So, so what's going to really change the the the the game I believe is when we're going to see the first fusion reactors coming and hitting the the space. That's when we're going to we're going to see that. I don't believe that data center will be in the space anytime soon. I don't believe this is going to be a short fix.

23:21 uh maybe in 10 years of course but anything in between that we we have to be on earth and the sole thing that we can solve right now is clearly redefining how we are distributing power >> we've seen Elon's track record is very clear he always delivers but never on time >> exactly 100% the vision can maybe be right but but and it poses a lot of challenges as well uh this data center in space but anyway that's that's for another topic another day most probably >> it's funny you mentioned the video model so maybe like not everyone will use it immediately. But the amount of like GPU like the amount of comput actually uses per user is very high. You're actually speaking to like a major power user.

23:59 You'll maybe see in an episode later when we publish, we use it for as much as we can like to animate a time lapse of the yeah New Jersey data center being built because that doesn't exist anywhere. We can kind of do that with AI. So speak to the power user. >> I'm completely convinced about the usage. Uh, trust me, I know what I'm saying is the ROI for the operators is so small right now even is like they they lose money basically by doing so.

24:24 So, I don't think they will be pushing it big time or you're going to have to pay a lot more per minutes or per seconds of video that you are really generating or re-engineering basically because right now they are clearly dumping the stuff. So, that that's that's not good. That's not sustainable at scale basically. That's what I'm saying. But yes, definitely video is a big thing in AI. It's it's massive 100%. >> Wait, who's dumping the stuff? The lab.

24:46 >> Everybody. Everybody. Yeah, everybody. When you are like getting tokens and and and you are making GPUs work at the price they're asking for. This is not sustainable long run. So this attract the first customers to showcase the technology, but they are not even on balance right now. So they need to increase and bump up a lot the price to to get profitable basically. So, and then at one point you're going to see that uh maybe you you will not be using that that much because then your video which was costing you 10 bucks will cost you 500. It's not the same anymore.

25:14 >> Well, maybe for our intro maybe we use $100 of credits. Uh if I were to hire like a graphics artist like maybe that's like $1,000 of >> that would be a lot more. Yes. >> And the thing is like maybe that would have been worth it if the video gets enough views, but I wouldn't have considered or even tried it if that initial cost wasn't as low. But now that the cost is low, maybe in the future, even if it goes up double or triple, I know the value and I'll still use it >> 100%. And that would have served their purpose anyway because for now they are educating the users to use that and so they pay for it. This is a customer acquisition and then they're going to increase the price accordingly. But that's that's the that's the usual aspect of any dealers, right? So uh get you hooked first.

25:54 >> So just now you mentioned what like increasing prices, but two also uh you're trying to get this take this long-term view. Something I'm really curious about is like how do you actually plan for this? Because even even if you just want to build a simple like a house, usually you go over budget on a construction anyways, how do you budget in like a for construction in an industry where component prices are going up like every month like going up very significantly?

26:18 >> So you know I built like 26 data centers in my life already. So I got a kind of extensive track record of data centers uh everywhere in the world in Africa, in South America, in uh all over Europe close to the North Pole etc. So small data centers, larger data centers, nothing closed from what we are building right now in New Jersey. Um so we redefined how we are building it because we have like plenty of USPS uh in this case we are cheating basically because we are next to one of the biggest or largest uh uh pre-cast factory in the northeast uh of the US. So already when you got these next to you, literally next to us like you take the GPS like satellite maps and you will see uh we got the trucks getting out of the factory and plugging the panels just on the other side of the fence. So when you get these you got a shell which is erected in no time because you got amazing teams as well on the other side and we got a full vertical integration of the entire chain as well. So from the metal to uh the electrical components everything is integrated by ourselves like almost all. So that gets us into a spot where we can be that fast basically and right now because we can see delays on busways on on switch gear we are opening our own factories to do so made in USA you know so we are going very far in term of electrical and uh and vertical integration into all the value chain of our field of the data center itself. uh transformers we have them in a month, two months, three months.

27:50 Everybody else has them in three years, two years, four years. So we have we have innovated into the sense of like finding the right supply chain at the right locations with the right partners that nobody was trusting them above but we did and we knew that we were right because we knew the guys inside out. So it's a mix of engineering, super smart engineering, uh because our engineers are very freaking smart, honestly. Um super design at the same time a super innovative uh supply chain and super high uh vertical integration. So right now we're going to announce as well data and construction because we are our own GC basically we have joined forces with some others and we acquired all their teams. So we got more than 500 600 people that are on our peril just for New Jersey to build the stuff. So the electricians, the pipe fitters, everybody is working within data one directly. That's the sole uh answer to be able to be uh as profitable as possible and at the same time being able to deliver the quality that we want and to be able to get access to enough people uh that we need as well because this is a very tense uh obviously field in the US the real estate market. So that's that's a a very short question.

29:06 That's a very long answer because this is like multi-layer answer. Uh and it's not like just one USB. Uh this is a blended uh um uh reasons or like multiple reasons of why we are so fast in comparison with the rest of the folks. It doesn't mean that we're going to always be matching our lead times because obviously we have our own delays uh because our supply chain can be delayed as well. uh we are trying to get multiple sources all the time. Uh but at one point our financial resources are finite as well. So we can't uh order like 10 times more skill that what we need. We can't order 10 times more stuff you know that what we need. So yeah that delays can always happen and uh and or the weather as well. This winter was the worst winter in New Jersey for the past 30 years. And of course this is the winter where we are like uh delivering the the data center you know. So at one points when you got people that cannot work outdoor because it's too dangerous for their health but what can you do you just stop everything right so you lose weight you you lose you lose uh time uh then can you cope it can you cope this time which is wasted maybe maybe not but in any case I can tell you we're very close we are very close so yeah I think we are the fastest in the industry way faster than any any competition honestly because the first 15 megawatt we turned up in 6 months nobody's even close to that like even you alone he needed 14 months because they talk about the commissioning and installation but they didn't talk about the 12 months of engineering they did before we had zero plan when we started uh this construction. So clearly uh when you were talking about budgetization uh budgetization was high level budgetization uh because we took all the major capex in consideration we knew them before but then the big question mark was what about the workforce because nobody has done that before. I mean all our design is first made nobody has even an AI factory close to it. I can tell you because this is coming from our gaffam factor gaffam customers they all say the same we never seen this kind of AI factories before or data center before that's that's the first design that we see like these and so when you get there obviously you cannot have like a playbook and saying oh it's going to cost me these or these etc you just take estimates and this is what we did basically so we take estimates we take buffers everywhere we can and this is the the way that you are budgetization making your budget basically >> when you create that budget are you already modeling or projecting some of these cost increasing.

31:32 >> Um so our cost they are kind of steady. Uh the the only impact that we have uh is the freaking tariffs because when we started the project there was no tariff and and when you take these into your into your teeth uh then it's it's bad but you need to live with it right so you find money and we find money that's that's absolutely fine. So um so yeah that's that's mostly the budget increase that we had from original budget is there. Then you take a bunch of contingencies on the on the top and and then this is how you're building it, right? So when you take suddenly 15% or 25% on a multi hundreds of millions of of order, then that that cost me like tens of millions more and you know that's that's the that's the drill. But we live with that because we take contingencies. We've got very good uh partners as well that are like providing everything we need. Uh and so yeah, we will not complain about that. So that's it. And now we're going to be able to claim back even more money by the way because the states it's been uh as you know it's been ruled uh illegal uh these tariffs of most of them. So we are claiming back a decent amount of money back. So yeah that's good.

32:40 >> When it comes to data centers when you say that you have tariffs but that's a increase on the costs on the input costs that everybody in the industry has and you have incre you have delays and and and like problems throughout the supply chain. So basically the input costs for everybody are increased. The delays for everybody are increased. But this obviously means you have less supply from everybody for insane demand. So if you are very certain that you're faster in industry, could the net effect for you be positive because then you and your customers can charge more dollars per megawatt. I mean let's say your cost increase by 20%.

33:21 And you're delayed by 30%. But you can charge 70% more like how can you can you explain like how those supply demand dynamics interact? I I can tell you that on our side, we are not taking this kind of opportunity to uh bump up our prices because we have we are establishing very long-term partnerships with our customers and uh we don't want to benefit of a short-term situation to uh uh break the trust they have in us basically and um and so we are trying to swallow as much of the increases that we have on our side and this is part of the etos of the company too. Um so I understand the global aspect of everybody is to make as much profit as possible. We are profitable enough I can tell you. So uh we are not here to take advantage of this kind of situation. We want to be here to support them especially in tough times like this because otherwise that's going to beat their uh profit margin as well on their side if we do this kind of things and uh that's not the intent because they are locked in with the prices they have got with their own customers for many years too. So if I if I start to do these kind of things then I'm just taking profits out of the hands of my customer and this is exactly what I'm trying to not do. Um and that's that's how you build long-term relationship and um I mean for us at least. So yeah for for the industry I can see some bump already in the cost and in the price which is offered to to the to to the other operators. Um, but we have such a way of making data centers that we have a nice profit margin. So, we can absorb a lot of cost increase. Honestly, I will not complain about this, right?

34:54 >> When we publish this episode and some of our American viewers watch this and they hear like we don't need to absolutely maximize profits all the way, they're going to think like how European. Uh, but I get it's a partnership, right? It's not just a customer, it's a partnership. >> No, no. I think I think I I think what Charles is saying he would rather have a decent profit margin for the next 15 20 years than to have like a great profit margin for two years.

35:16 >> Exactly. That that's long-term investment. >> I think he's very capitalistic actually. I think that's smart, right? I'm trying I'm trying to Yeah. I'm trying to support my customers on the long run instead of like thinking short term. There there is two two two two way of thinking it like either you you you you secure your next eight quarters or you secure the next like 100 quarters whatever >> and data one is a is a private company right >> yes I own the majority of the shares >> so you don't optimize for next quarterly reportings you optimize for the long term >> yeah 100% and we are not publicly listed either not yet at least >> yeah that would be cool that would be interesting uh company to invest in >> I hope so you're is smiling a lot this you're smiling a lot. It's like did I hear an IPO?

36:01 >> Maybe maybe maybe next year. >> But but it's interesting what you're doing, Charles, because um we a lot of people in the industry are confused about the exact same dynamic with TSMC. They're always saying like why don't ASML and TSM TSMC double their prices? Why are they playing so nice with Nvidia? And it's probably because they have like a 20 30 year relationship and they trust each other. And all these people who are like, "Oh, you should double your prices." To me, they sound like idiots because you you're going to you're going to destroy decades of trust built for like your next couple of quarters really and and not and not only at one point it gets so much expensive for the end user that nobody will be able to use this technology anymore. So the reverse effect will be massive too because what happens if tomorrow Chad GPT or Open PII is not asking you $200 per month but $10,000 a month like nobody will be paying for that right so so and this is what happens because this is all a nil sum effect at one point you know everything which is charged before needs to be charged downstream so at the end it gets to the end user to pay the difference and so yeah of course you can always charge more until a certain point and and I think we are very close from the equilibrium these is so uh it's already a ton of money that you need to up front for only five seven years of lifetime into your GPUs you know so and and when you got six months of delay you already already taking a chunk of the profits off the table so it's all very like yeah that's a tough game and uh and GPUs are so expensive already that like bumping up even more the price right now will render the stuff most probably impossible for many operators anyway so that's why day one with us we are not entering the the neo cloud game we don't want to uh be the next nebus. We don't want to be uh the next ncal or whatever.

37:45 We are very happy on being like the infrastructure provider because I think we have like a unique niche here which is already a massive market and where we are welcoming the new clouds the hyperscalers directly into our walls uh because we developed the perfect product for them to fit in. So they just need to to concentrate and focus on their GPUs and the added value solutions they can build upon. And so this is where you need to be very costefficient as well because otherwise you just think okay I will do it myself right and if you start to behave like a predator as well to them they're gonna say okay Charles you're nice but next one I would do it myself and this exactly the reflexes that you don't want to see happening either you know so that's it that's the full that's the full story of it >> I think like we wanted to do like a quick fire around and just like uh power sources what you think about them >> yeah when do you think each power source when do they make sense and when do they not make sense especially interested now with lead times, US all that stuff. So when do they make sense? When do they not make sense? Starting with gas turbines.

38:44 >> So gas turbines they make sense. They make a lot of sense for any large scale data centers. The only flip side of it is that you need to have like very strong emission system in place to be able to not just kill mother earth. But that's the European which is talking into me when I say that because of course the regulation is lots less demanding right now. So if you take the right oxycat and SCR you you're good already and that's it. But for me that's a massive additional pollution that you can put in the atmosphere. So you you have to do your homework properly if you want to do it at scale and not having some big push backs uh down the line.

39:26 But this is definitely what you need because they can deliver 200 250 megawatt a piece uh with double combined cycle system etc. So that that's definitely where you need to go if you really want to to be serious on the long run with like sustainable uh not sustainable but yeah sustainable power in the sense like sustainable in the uh in the power level uh not maybe in environmental level but you can you can deal with that if you got the kind of technologies we have vertically integrated again in the group. Um but and the price per kilowatt hour would be the lowest definitely especially for the US uh because you get gas cheap gas and then that's it. But again longterm because that's three years, four years, 5 years now lead time to get these or you overpay them if you want to have them in 18 months, 24 months at city prices. So the other stuff that you can do is refurbishing them. So you can you can take like all the uh like old turbines that you can completely refurbish because they are like uh you know the same kind of turbines you have in the airplanes right so you can overall them completely you can unmmon them piece by piece and remond them piece by piece so you can refurbish these kind of things in six eight months and replug them into your facility and they are zero hour so that's that's still doable uh that's the kind of things we can contemplate for some other sites too um then micro turbines I don't believe into them not at that scale honestly this is way too small.

40:46 >> What about what about for smaller data centers? Because there's actually one company looking into uh is it called Capstone green energy? >> Yeah, Capstone is a great technology. The problem is yeah, if you do 5 megawatt, 10 megawatt maybe, but honestly these days I don't believe into this size of factories because uh right now the the the efficiency of the concentration of power is so high uh that having anything below 100 megawatts becomes irrelevant. And this is where there is a for me personally it's not a financial advice but there is a massive short on all those is legacy players because they got thousands of data centers with like 1 megawatt 2 megawatt 3 megawatt what you going to do with that honestly in the long run as the market is switching right now anything below 50 to 100 megawatt doesn't interest anyone >> so in their in their in their latest earnings call they said that they they made a model about how transformation would be for their business to get an order for a 50 to 100 megawatt data center and you think so but you think basically the solution is cool but not for that scale.

41:45 >> Yeah. No, no. Exactly. So because you got at the end you got too many engines, you got too many stuff, it takes too much land and it's it's a yeah it's complicated. Then you got linear generator as well. You get some very nice technology coming from a company called Mainspring. Uh like that's very interested. Uh but again for me it's lower scale because uh there are only 3 like 350 250 I think kilowatts uh each.

42:10 So uh and you you got like a lot of maintenance to be made on on this kind of technology. So but very nice because almost no knocks, almost no VC's like uh it's it's really good uh actually and that provides like yeah a good level of electrons but good for again up to 15 megawatt most probably above that difficult. And then you've got all the uh the uh like green energy like the wind turbines and the mix with the solar panels and the battery energy storage system. That's absolutely the perfect mix for me. That's the real answer. If you're in Texas or if you're in a very sunny state um with a lot of winds because the mix of it gets you in a very nice spot if you invest into a decent amount of batteries. Um so at scale that can work clearly. Very quick one. What do you think about zinc based batteries?

43:02 Do you know about the company EOS? >> Yeah. So, uh that's good. You got plenty of different technology of batteries. Uh you got nickel and zinc, you got zinc, you got lithium, you got you got plenty of different stuff. So, uh they just need to under the heat cycle and to to be still there within the next five to six, seven years. Um and having the right cost as well because most of these new emerging technologies they just cost a lot because they don't have yet the scale uh to get you uh in the gigawatts that you need basically and so you ends up paying a tremendous amount more uh but of course they can they can take the AI heat cycle like the huge drops and the ecaps and everything much better than the lithium but then the per that you see is like just the cost because for having 15 minutes of lithium ion batteries you've got five minutes of uh nickel zinc for instance or even less.

43:53 So that that's the kind of things we see. Um what else? Uh what did I miss? Like nuclear plants. Nuclear plants is definitely the best for me like by far. Yeah. Yeah. Especially if you but you need to shield the the the the grid in this case because when you got gen sets you can put flywheels in front of them. Uh which is the stabilizers of the grids. >> Carl and you said in your like thick French accent nuclear is the best. I guess that's very like French. Um I'm glad we went through went through all of this because like Daniel and I were actually having a sort of a debate um about like especially aerody turbines. I started my career in Seammens working on these aerody turbines. Daniel got really interested in it recently because when you look at the lead times for the large gas turbines, those are like many multiple years, but then maybe now you have some smaller like microurbine suppliers that can have like shorter lead times. So just now you went through all of these different sources in like a idealistic world. Uh but then in reality you have all of these all of these lead times. Um so can you now take the real world into account?

44:54 >> Yeah and think about the again gigawatt scale because this is what we're talking about right? So we are talking about gawatt plus. So it's not only one turbine or it's not only two turbines or three turbines you are ordering. In this case it's it's a ton of turbines that you are ordering. So you imagine the complexity of it. So of course you will not be able to have only one vendor. you're going to have to mix all kind of turbines all together and it creates even more complexities and etc etc. So uh there is not a single answer in this case. It's a multiple layer of answer again because you cannot work with only one technology. If you work with only one technology you already screwed or you're going to beat five five years down the line. So there is no easy answer. All the people that are in this industry was comfortably sitted on their seats and having five years down the line plans right now. If you think like this, you are already dead. Uh and you need to think out of the box all day long. So you need to have like again a strategy which is multi-layered. So it can be a mix of grid with a mix of gen sets, a mix of of like of gas turbine and at the same time you got enough land to put some turbine like wind turbines and solar. You know it's it's this kind of strategy for me which is the winning one. Uh and that's the kind of things that we are working on actually. So we have like a full uh cycle of suppliers all around the place so we can tap into all these different pockets when we need a new site because at the end as we said it's uh speed speed of execution. So >> yeah, that was sort of a con at least a conclusion from my side when I was discussing with Daniel like I would probably not use a bunch like a hund of these turbines in a data center but if I cannot get my power source I'm just going to take whatever is available. Um do you think maybe not for you guys cuz you planned in advance but do you think that is potentially how other builders think like they just will take whatever power source they can get?

46:45 >> Yeah. So, so the the bull case the bull case the basically the bare case for companies like capstone has been well it's so efficient at the scale but the bull case has been but what do you do if nothing else is available >> yeah that that's right but at one point you have as well the reality of engineering and the reality of cost too so if you end up paying like 20 cents a kilowatt hour you don't have market anymore you know it's like it doesn't work at one point so and when you got inefficiencies at that level that cost a tremendous amount of money and you need to think about it's it's just like a investment for one or two it's investments that you're going to be having on your balance sheet for the next 10 to 20 years. So, one mistake is replicating over the years and cost you like a lot more. So, yeah, I I don't know about I think that you can still find some plenty of other solutions around the around the market because you got and this is where you need to be very careful as well. You got a lot of announcement in these markets but very few players which are of like the legitimacy uh the background, the understanding, the and the financial capacities as well to convince others.

47:44 there there's a lot of companies pivoting into AI because it's hot but that doesn't mean that they have an actual offering >> that they don't have a clue of what it is they don't have a clue of how like a cooling system in AI factory really works honestly I can tell you firsthand the discussing with with all the big customers we have in front of us we have the echo of the rest of the market obviously but I can tell you like very few players knows how to cool down a a 250 kilowatt rack for instance like efficiently uh and and the rest of the chain is I'm not even even talking about the rest of the chain like integration of of you imagine like you are a poor provider at the same time you are a coding provider at scale which is insane and insane and you are like a a critical um uh provider for uh for uh um compute power solutions. So this is three different trades that you are merging into one kind of a new breed of companies and you need to understand absolutely every single stuff every single beat of of efficiency and tech.

48:48 >> It's clear why you have such a busy day Charles because this is I'm I'm going to be honest. I would rather invest in people like you than your job sounds >> your your your job coming from a form entrepreneur. I I respect you a lot because this is very difficult. This is very this sounds very stressful. It sounds like it's aging you quickly. It sounds like a lot of fun >> and pressure and responsibility, but a lot of pressure as well.

49:12 >> No, it's it is it is only pressure. I will tell you honestly. >> And and and you are in between your customer, the customer of your customer and all your supply chain and all your guys at the same time and you are taking the best of everybody and that's a lot of pressure. Yeah, it's it's an insane level of pressure. But the but this is good. I mean, you know, this is like Elen when he was running his uh his first factories. You know, the guy was sleeping in his factory to make it happen. That's exactly the kind of living moment we have. You know, it's the the hell of production, right? So, and this is exactly what we are doing right now. And when people are asking us like, do you have the playbook? Do you have the the processes? Hey, come on man, we are drafting them as we speak, you know, because nobody did what we did before. So of course we can inspire ourselves to like best practices on other trades but still we need to do all the even the SOPs and all these kind of things. So it's it's a very interesting very interesting moment but yes very stressful as usual. Maybe I would like to hear your your your final words on cooling like where you think that industry is growing because from what I understand is like the the power like power is becoming more and more important and efficiency is getting more dense, hotter and the industry is moving for example to glass substrates and basically that obviously has implications for cooling I assume but also in general like where do you see what kind of big changes you see happening not just today but also next to the five years?

50:34 >> This is a very wide question. Yeah, I would be even more philosophical if you if I may. Uh that's the French side of me talking. Um because I'm here for the long run. Um and and so basically I believe in the trade that we are occupying right now in the sense that we are really dealing with intelligence, right? So when you are doing artificial intelligence, you have this this manifent word which is intelligence. So what we're like dealing with is intelligence. So more AI factories you have, smarter you become as a nation. as humankind, as entrepreneurs etc. because this gets us into a spot where we can have discoveries in matter of days or hours instead of years or centuries. So we are able to redefine what we see, what we think, what we how we want to evolve in a scale at a speed that has never been done before. And and for me this is super inspiring honestly. So the rest of whatever we do is just to serve this vision honestly. The rest is just irrelevant. It's noise I I would say almost uh it's just solutions for engineers to figure out and I'm an engineer at Kora. So we got plenty of solutions. It's fine and and with amazing people all around. So but what people needs to understand is the bigger message or the bigger picture of this industry and this industry is about intelligence and at the end the sole question that you need to resolve upon is do I want to be the dumbest guy in the room or do I want to be the smartest guy in the room? So, do I really want to fight this industry which is bringing me light and brightness or do I really want just to embrace the way that it is and that is figuring out how to deal with that, right?

52:13 >> Oh, wait, wait. Now I get it. It sounds like you're you're talking about like the modern version of the enlightenment. That is the French side of you. >> Yeah, that's the French side of me, right? So, yeah, I'm strongly believe that this kind of technology will solve death for instance. That's my strong belief. >> Biological aging or death? >> Death. Death. death as as a as a disease. This is a natural disease that we have inated basically. So because of cellular regeneration technologies and things like this that are making possible right now through AI because we have like magnificent models that are dealing with DNA and with uh proteins and everything we are able to uh literally emulate as you know already like uh the degeneration of our cells and everything. So we can reverse engineer ourselves due to this technology basically and this is how you kill death basically definitely that's that's the nice stuff but and everything which goes with it right so and by doing so space become a reality new physics become reality like everything's change when you see that because then time is just a variable uh and it's not essence anymore so that changes everything and so that's why we are focusing on the on the I would say the foundational break to enable this market to get there because I believe this is the the best and the quickest way to solve this and because this is such a complex realm or a complex uh issues to solve this is where I'm focusing the most of my time right now and when it's complex that brings value as well obviously because more complex it is more value you create usual stuff so cooling is one aspect but the rest is as essential as this, right? The the rest of the food chain is as essential as this. And I think we are living a a really a tremendous amount like a tremendous time like where the civilization is really switching and is shifting. I it's just like common sense, right? So when you think about energy, this is indeed a game of energy like all the guys like everybody is saying the same like now AI factories becomes power like electrical power. It's an electrical power game. So the next big thing will be fusion reaction for me.

54:28 When it's going to happen, we'll see, but maybe sooner than later. >> There's a there's actually a couple of very interesting European fusion startups. I I've heard some very I I've seen I have friends that are investors that have some deep pockets that invest in that stuff and they are saying they're make like they're overpromising maybe, but the claims they're making are mind-blowing. Some of them actually claim we solved energy. Like they some of them are it sounds way too good to be true. If 1% of it is true, I will be happy.

55:00 >> Okay. Okay. Daniel, your your friends are investors, but I study engineering. So I have friends that are engineers in these fusion companies and agencies. And I wouldn't I wouldn't say they're that they're that uh optimistic about they're excited by it. But um it will take some time. It's okay if it happens in 20 years. That's fine. >> Yeah, we'll see. >> Yeah, there's a different perspective from the engineers on the grand investors hoping for >> if stock goes up, it's soft for me.

55:34 >> No, I'm joking. Okay. >> No, no, no. I I I I think I think we are most probably only 5 years away from real fusion. >> And Charles is an engineer >> and from France where like all this is happening with experts in nuclear. Uh Charles, you're literally building the future. Thank you so much for your time. This was so information, so good. >> Thank you very much, guys. Thanks for the invitation again and for your time.

55:58 It was a real pleasure. >> Hey guys, thanks for watching. If you like that episode, the best thing you can do is just share it with someone else who you think might like it as well. Otherwise, a rating or comment on whatever platform you're watching on goes a long way. And if you're looking to sponsor the podcast, then I'll drop an email right here. Otherwise, I'll see you guys in the next