 Hello everyone and welcome to JSA TV and JSA Europe the newsroom for telecom and data center professionals across the continent I'm John Max Lim and on behalf of the team here at JSA. Thank you for tuning in to the first of three parts of our end of the year special broadcasting series Today, we'll be looking at the future powerhouses of Europe aka data centers We will delve into the geographic locations on the rise in the digital infrastructure connectivity and data center space across the region As well as the benefits of these new destinations the role of the upcoming regulations affecting many of the EU territories And what to look out for in the new year Having said that it is now my pleasure to introduce our exceptional executive lineup Which includes Tom Glover head of data center transactions for EMEA JLL David Hull senior vice president of collocation for Atlas Edge and Matthew Bains vice president for the design and construction partners for the cloud and service Provider segment at Schneider Electric Everyone thanks so much for joining me. I thought before we jump on questions I thought if you could just give us a very quick elevator pitch on what you do and what the company focuses on It just gives our viewers a bit of a background on on the things that the company does Maybe start with you Tom, and then we go around And then yeah, so what's your elevator pitch? The 30 seconds so JLL is well known as a real estate Organization but within the data center arena specifically we act on behalf of operators on behalf of cloud providers on behalf of investments and developers Investment funds across a whole plethora of different areas on the transactional front from finding sites to buying sites to buying companies to Advisory work to research and then flip across to the facilities management part of our business as well Pretty much everything except actually owning any data centers is perhaps the the elevator pitch for JLL And that's what I'm responsible for. It's the transactional side Almost like the Airbnb model in a very different way, but you know what I mean David Hi JLL. Yeah, so I described that edge as the world's maybe controversy the world's only real date to edge data center business And what I mean by that we're real because we're not a software or network business We're really building and buying data centers to expand our reach Real because we're really on the edge of the network and that's kind of what we mean by edge I mean that's partly thanks to some of our investors and we'll talk about that bit later Real because we've got real customers who actually pay us and real books. We're really generating returns for our investors So it's kind of that snapshot about this edges business, I guess Yeah, and I think actually now it is important to say that you're real because you do have real estate assets People out there they say they exist but there's nothing to show But we won't we won't get through that one today and then Matthew Yeah, hi Joe. Hello everybody and and David you are real because I've seen you in real life. So we Should I don't let you we're a global, um, let's say a Manufacturer technology company We are French headquartered, but we are a global company. You could say we're manufacturer of data center solutions offers in terms of hardware, but not just that we are also a big software company in terms of the management of your data centers as well, but actually What we really try and do is empower our customers to make the most of their energy and their resources and really it's it's to position ourselves to be your your digital partner for sustainability and efficiency and That's on a global scale. We are a global company. I'm based out of London. I'm on a goal I have a global role based Here in Europe working with the global Cloud and service provider segmenting And yeah, and what you do is a very important part of the instrument because it actually keeps the lights on Thousands of data centers across the globe and then the region of course But I think I think this panel is very exciting because we do have the operational side We have the energy side and we have the finance real estate real estate not real estate tree Side of the the conversation as well. I mean so I mean Tom You're joining us from Frankfurt's all of the other ones or all of us are based in London at the moment It's To the flap markets so the tier one cities Frankfurt London Paris Amsterdam and Dublin as well We've know a lot about them, but what's really in people's minds at the moment It's the tier two locations the new uponcoming Hubs like Milan Madrid Warsaw So my first question really is what are the new locations across the continent? What's really driving investment? In terms of geographies and I'll open to the floor I don't know if Tommy from to start off someone else wants to start. Yeah, I'll kick it off I think the right flat D. Everybody's very familiar with them and they're going nowhere. Let's not kid ourselves They're going to continue to see growth over the for foreseeable future And not least I would throw into this thought process because of AI But putting that to one side, I would say locations like Lisbon Portugal Madrid the land Italy Barcelona you're going to see more more developments there But let's not forget noting the Nordic region where all these locations as there's been investment There's been data centers operational for for many years, but the Nordics especially with the advent of the sort of the learning data centers that are going to be driving the growth of AI in The future I think the Nordics offers a huge potential interest area Just because of the the climate and the renewable sources of power so Without wanting to grab all of the answer to this one. It's everywhere is interesting I do not see one country. It's not interesting. I including that in the Baltics right now And we're not even touching on Middle East of Africa today. We're just talking about Europe frankly, so it's all exciting Looking closely at Vienna looking closely at Madrid looking closely at Lisbon looking closely at Barcelona and Milan They're the key what I would say right there. Yeah, I think that's this very interesting I mean, I can only imagine the amount of the straining that's going on in boardrooms When it comes to putting pins on the map because whatever you put a pin you can actually build something And surely that you will be able to find somewhere so someone to anchor the facility and the Nordics is interesting It's I mean the Nordic start developing around the meat 2010 And we've seen a massive boom especially with the hyperscaler with the hyperscaler client And then it kind of went a bit softer because we've seen the other regions grow But it's kind of coming into a 2.0 euro when it comes to the Nordic's a lot more people Investing in the region and I really want to get David's and Matthew's view on this as well Especially on the energy side because that's I guess it's the biggest UPS of the Nordic's is the energy availability And the pricing of set of said energy It's funny actually you talked about pins on the back to actually at the stages name Come from we've been thinking about it in terms of the pudding pins in an atmosphere So you hear all exactly how we name ourselves that you we have kind of an opposite view I so we talk about two kind of motions in the kind of where data sensors being developed We talk about this cloud out motion So I guess we all kind of know that the cloud essentially lives in the flat D markets But if you think about some of those flat markets, they're kind of not really interesting from a GDP perspective There's not that many Dutch or Irish people and actually from a GDP perspective the Netherlands and I don't actually create much GDP Compared with something like a Germany or a France, right? So actually we kind of think those markets are certainly Sort of fairly saturated and I think Tom's right probably possibly some new applications might make them more interesting This to be determined whether or not those applications are better hosted close to the end users And I think there's certainly a debate to be had about that But I think that our expectation certainly is that those markets won't grow as quickly And it's always hard to grow a market when it's very big, right? But certainly I don't think we see that the kind of double-digit growth in those markets that we might expect to see somewhere like a Berlin So we talk about this kind of cloud out motion, right? So, you know, if you're a cloud provider, you've been able to deliver all your services from Dublin, Amsterdam, right? But there's now applications whether those are machine learning or whether those are server hungers or whether they're Latency sensitive or bandwidth sensitive that require you to move that kind of closer to your end users, right? So I think that that motion is essentially what we're seeing kind of primarily driving demand at the moment Certainly the places that Tom might mention like Lisbon and Berlin and Hamburg and Stuttgart And Germany's an interesting market because you've got such a distributed GDP So only where we're looking is where there's big GDP and not many data centers, right? That's just kind of a logical thing that you might want to do if there's lots of people if there's lots of companies And there's not really many data centers. It seems kind of logical You might want to build some and then on the other side It's the kind of age in motion, right? And I think that that's probably somewhat less developed, but that's kind of saying well, okay I don't want data centers in my you know my office anymore Or I don't want data centers in my factory anymore I benefit from consolidating some of that IT into a data center Which is still quite proximate to my locations and my people but a bit further away than my office So I think for us that that's certainly what we see driving demand But but you know as Tom kind of alluded to That cloud out motion is is massively more significant at the moment than the the kind of age in motion Which I think is why you're seeing certainly asked building kind of 20 30 megawatt sites Not the kind of 500 kilowatt a megawatt sites that you might expect in the kind of far-edge Movie in was but certainly that was the kind of market. Everyone was talking about five years ago Which doesn't seem to develop yet anyway Interesting and I think just out today that I think maybe to a lesser extent at this stage still and Matthew I think you're probably gonna be able to to add on what I'm gonna say But it's it's also about disperse in the power consumption from the bigger hubs into into other parts of the the the countries To not supercharged the grids in those more more heavy locations In a way, I don't know. I think what I'm saying is it's Correct what I'm saying or can you give us a view on this inability front and the energy front? I mean just just firstly I echo Tom and David, of course, I mean David is interesting you say in terms of growth Are the core markets gonna grow as much as the other markets in terms of caga? Probably not but then it's like it's like the US market really they what's the caga may be low Not flat, but low. Let's say they're still building 3,000 megawatts That's it's still a there's still a huge demand for new capex investment for data centers in these core markets I I think it still feels a little bit like and Tom you can either that if you build it They will come in Frankfurt if you can get it They will come at least anyway, so still still great markets, of course, but then we're seeing the spillover I think into Berlin for example and places like that spillover markets where capacity is constrained, but just how you said Around power. I mean, I I don't need to talk about it too much Everybody knows that some of the biggest challenges are land and power in our sector for sure People are looking at how they can generate power in core markets whether through on-site generation of micro grids Hydrogen generators, whatever that might but whatever that might be, but of course that is difficult not always easy to get the permits not always easy to negotiate and work with the energy providers the DNO's or The regulators whoever that may be and sometimes the ROI is a little bit tricky for that as well in terms of your your business case And your investors might not see that as a best Investment to make some of the investors we we see have a Short-term view than others of course as well. So you've got to balance that that out But other markets tier two Tom's mentioned them David has for sure Milan We're seeing at the moment is a very ripe market Madrid Barcelona Lisbon. You said and Zurich is Zurich still tier two. I don't know I don't like the term tier two because they're tier one in their own country, of course Major capital markets, let's say, but there are some challenges. I think I was talking to one Major operator where they're saying they're actually gonna focus. They're not gonna focus should I say so much on tier two They really want to focus on the major campus sites and that's where they want to put their energy There's enough demand there They can't spread themselves too thin to go into all of the the Secondary markets if we just use that term And so there is there is opportunity David for people like at this edge and other place to really hop on and be early Early adopters in those in those cities. I think it's I think it's fantastic and it will help Breed newcomers and new entrants perhaps as well the new investors in that in that space Maybe some areas that it could be a little trickier to do business with I do think again some investors It depends what the fund is Ease of doing business and return and the cost and things but to finish on what your point was our power of course is Tricky and then we need Western grade data centers built to Western grade with sustainability in mind Of course as well top of mind. I think there's a question perhaps later on on Regulation and sustainability as well. So we'll touch on it later On the power front now I mean actually power power is hard, right? But you can dig a hole and you eventually can get power generally speaking some will upgrade the substation It will come it may not come exactly where you want it I think Madrid you're looking at like eight years at the moment and no one wants to eight years But actually the bigger challenge that I foresee certainly in the flat T markets is the requirement to recover and reuse heat Right because I can build a gigawatt campus in Frankfurt But if I have to recover and reuse 300 megawatts of heat Must hope that people want a lot of saunas nearby Because that's a huge amount of heat and actually that's that's going to be certainly in Germany And as you say, we'll cover it perhaps when we talk a bit more about sustainability But that's certainly going to be a big challenge where you've got like In the U.S. But you know Ashburn, Virginia, what are you going to do with eight gigawatts of heat? All right, so it's an awful lot of heat if you want to recover and reuse it And I think you know 30% is the kind of threshold that Germany is setting But we're clearly in the future We're going to look back and wonder why we could console ourselves that PUE was the measure of how efficiently We were boiling the oceans or heating the sky in a world where energy is scarce, right? And I think clearly we're going to want to recover and reuse all our waste heat And that's certainly going to drive a more distributed data center architecture I think just purely because you know, I can get I can get power relatively easily Somewhere in a concentrated half Frankfurt or Amsterdam And it's much harder to see how you could recover and reuse the heat And I think if that German legislation that comes a template in Europe, which I hope it will And that's really going to make it very difficult to substantiate building more data centers in Dublin and Amsterdam Yeah, I mean you you you mentioned PUE I don't want to open the can of worms and now and I know Tom You're very passionate about PUE, but it almost brings the question of do we need PUE Of where we are with things, especially when we look on a global scale. I mean a PUE number in Europe Compared to an Asian PUE number, for example, a Singapore number It doesn't it doesn't match. It doesn't fit. It's completely different realities So it's almost the case like like you said, we want to evaluate these things on a PUE basis On the actual benefit of the data center to the local community And to the point of the the heat reusable and the grids a lot of infrastructure is not built On the grid front. I mean, if you go to southern Europe, there's no infrastructure built to send Heats to to even to a local pool Let it let alone to the housing to the housing pipeline I know yeah, Barcelona, you've got a nice LNG recastification plan that's desperate for your heat So they can recastify their LNG And I you know, I think there are other trends things like the recording and stuff that allow you to recover Waste heat at much higher float return temps than you could and if you're able to recover heat at like 60 70 c Then you can start to think about running things like absorption chillers So I don't I don't think that it's he recovering reuse is something that only applies in northern Europe Which I think we often kind of have historically thought about I think it's just as valid in southern Europe as well But I'm sure the other guys have have thoughts only Matthew. I'm sure has thoughts around that Um, yes, I have thoughts It's uh, it's an interesting one. I think it's been in discussion for a while. I've seen some use cases I've seen some examples where people have have um Made use of this waste heat and I think it's um, but it's pretty unique at the moment Regulation will probably dictate that there's got to be something around The use of it or the reuse of it somewhere somehow and that may certainly dictate the location and uh geographical nature where we put data sensors and what we can do with that waste heat Or as you said David, we actually the technology inside the data center Utilizing that waste heat differently and then actually we design data centers differently and It's almost like a circular Technology economy inside the data center One or the other I think it's tricky. It's certainly There's a bit of nimbyism. So you don't want data centers by your Residential areas to heat homes. I mean it's that would be ideal in a way or public offices and schools and things but as we know It's just not really viable that 500 megawatt data center campus is on the doorstep of a primary school or something like that. It's just We often focus too much on this idea of using waste heat for residential when there's actually a thousand Other uses for that heat. I I love the things that The guys at pure dc have been doing the Nordic you're using it for salmon farming of all things and then you've got the the use of it in um Vertical farming scenarios, you know, there's lots of other things that the waste heat can be useful beyond the residential aspect Maybe there's an argument to be had that says A data center should be in the middle of a big farm field Because if you then export the to support that farm to help it, you know produce better crops more crops For the population and that's actually a good use of the lab But no, but you're right. I think people are going to have to be imaginative and innovative here because It's likely it will come through in regulation Right with the with the european commission. I suspect for sure. So we're going to have to be creative I don't there's no I don't think there's anything written about what you do with it necessarily, you know, there are different grades of of how you utilize it and Do you get special credits for utilizing it in one way versus another? I don't know but that's the calm I guess tom and People who are thinking about it now will obviously have a bit of a head start Everybody on this call agrees So the industry itself our industry we're all saying the same thing We should be reusing our heat somehow somewhere some sometime as soon as possible. So we're all saying this I think we're all saying the same thing here Yeah Question around pue, right? I'd much prefer to have the data center with a pue of 1.5 When I'm recovering and we're using 100% of my waste heat and heating a hospital because essentially I'm displacing The load that that hospital would be running their boiler Well, so if actually it's fine heating the ocean and having some salmon farming Although I think salmon farming in that scale is bad for the environment generally But but it is better than probably having electric heaters in the ocean to farm salmon And but actually if I'm displacing a hotel or a hospital that would have been running a boiler There's actually a negative carbon impact from my data center, right? so That's a massively powerful That's a massively powerful thing to be able to do and actually I think that it really is incumbent on us as an industry to Make choices about where we build data centers based on us having neighbors who want the heat And if I look at projects that we have under construction every single one of those projects has a neighbor who will Not all of our wasted, right? So I'd love to be telling you truthfully that we found people to export 20 megawatts of waste here And that there are some examples where we can But all of our projects were able to export some of our wasted And that for us is just because we make that a criteria of site selection So that's totally up to the industry. You can do it or you cannot do it Right, but it's totally but you totally within your grasp to be able to do it You just have to make different choices when it comes to to selecting a site And I think with respect to building a 500 megawatt data center next to a primary school Probably not right, but if I went to a local council and said, hey, look, you've got a sports center Or you've got a swimming pool. You've got a bit of land. You're not having as many people drive there now Why don't I build a data center in your car park and I'll use my waste heat To provide you essentially with free heating for your swimming pool library, whatever I think there are some UK mini councils who'd be interested in the opportunity to not pay a heating bill And to lease a few car parking spaces to a data center operator And of course there's people like deep heat or deep green, I think who are doing exactly that So, you know, and then that's a small scale, but certainly I think that The trend is moving that direction and as operators certainly is incumbent on us to make sure that we are choosing actively to participate in that not just hoping that someone will come along And want to use our heat, which of course is not realistic in some locations Yeah, it's like guys. I think we will have to get creative On how to do this and have to be willing to do it as well I think some operators still need to make their shifts from we can do it to actually doing it In that sense, and there's plenty of use case. I mean, we have for example, Q scale even though it's in Canada Where they use the entirety entirety heat for use to To farm tomatoes and other vegetables, which are then sent out to schools and community centers and everything We have equinix next year is going to warm up the olympic pool in paris With the data centers heat. So we were starting to have some some nice use cases And like I said, we will have to have creative use creativity to come up with different ones So we cannot even think about that the moment But you've all mentioned the keyword of regulation and I think that's kind of There is actually the hot words for for 2024 I would say with everything that's going to come into into effect in europe within this is the first First half of the year. So my question will be I mean to all of you What sort of regulations should everyone be aware in europe right now? That's coming in 2024 and how is that going to shift? The the could shift the landscape. I mean, we already mentioned that maybe in the future These things will dictate where data centers get built Which will I don't know if it's a good thing if it's a bad thing we can also ask the question But let's go through the regulation scenario now. So What should we be looking out for and I'll let someone start CSRD, I'll kick off with that one. So corporate sustainability reporting directive measurements started this year The actual effect of those measurements will start taking impact in 2024 What I like about this is it's It's new. It's the first time. I think that regulation in europe has actually come into force And not all but most data center operators will have to abide by it But not all because there are certain sort of the Benchmarks you've got to achieve before you you have to To report on it I'd like to think that those data center operators that don't fall within the criteria that mean you have to report on this Reporting directive. I'd like to think that they would see it as a best practice anyway But for me what I like most about it is it's starting to measure multiple facets of Sustainability not just PUE is looking at the whole plethora of attributes within industries not just data centers, but across commercial industries And if you yourself think You're not doing particularly well against some of these metrics. That's a good thing You can start now putting things in place that you can measure against a consistent number You started off at x we want to make improvements of x plus 5 or x plus 10 or whatever you decide is a business So as opposed to it being a stick to beat yourself up with as an organization I like to think of csrb as as a initiative as a regulation that should Over time enable people to see and companies to see just how well they are performing and behaving Um, and then making improvements against that if you're not measuring it you can't do anything about it So that that'd be my starter for 10 Yeah, I think the measuring is very important and of course the reporting as well Um, and then I a lot of third dimension, which is actually understanding the data And who's reading it needs to understand it especially policy makers, uh, but uh, davey mathew Yeah, I mean, uh, I'll go I'll go with tom. I mean you've got the energy efficiency directive from the european union as well I mean these are He's a similar and coupled with what tom's saying, you know, there's got to be transparency in Reporting from from data center operators now. I think it's very important that that that takes place. There's still some um, the industry needs to Needs to eventually agree on what the reporting looks like and what are we reporting? um, and who too as well, of course, I mean, that's uh Uh, we're getting there and I know the eu dca is working on this quite um actively with the european commission um self regulation is something we would like to do as an industry certainly If not if not then a form of it, which means we actually come together as one voice um, and we have that and the voice in front of the european commission is a um coordinated Voice if we look messy as an industry, it's more likely that they'll Do quite heavy top-down regulation? Uh, when they don't really understand the industry that much. So I think I do I do see some active, um, um coordination activity with the a lot of operators in europe with the eu dca um in this in this space and the eed um Was influenced let's say by the european data center association to a certain extent for sure Um, it certainly had its say So we need to continue on this path. I think we need to work together We've got to um eventually at some point soon Agree on what those measurements are and what they look like But it's going to be key. I think and um, it's going to be interesting how people Take this on board and when they take it on board they, um Believe climate data centers need to be climate neutral by 2030 That's the and and the eedca is working on a climate neutral data center pack to really gather the thoughts gather the um Intellectual property of the operators on how we can all collectively achieve that so we've started in 2023 as tom said It's going to ramp up in 2024 all the way to 2030 and probably beyond but I think um In general this industry can give itself a pat on the back anyway because it's um strive Towards sustainability energy efficiency. Let's say Is helps the bottom line anyway and it helps differentiate them and it helps you win business and investors like it So all of that put together. I think we're on a on a good path But regulation is going to come in europe and we'll follow in other places, but it may be different You talked about the us What are you going to do with all that wastey? Well, they probably won't regulate on it then because it's there's not much As you said, perhaps you can do with it, but we have the climate accord from the eye masons and all of these Organizations will come together and are working together to make sure that we have something that's sensible Um, and that we have one voice in front of the regulators. I think is important That's interesting. I was actually chatting to the climate neutral. They sent a pack guys earlier this week and um The amount of ground work that has been done this year alone It's this quite extraordinary. So I'm expecting big things next year in terms of actual visible palpable um reports and announcements and news and and actions Um, not lobbying. They were very very they they made sure I didn't say it was lobbying But uh, actually getting things off of the paper onto to the streets Of brussels to actually get action on on the way and david give us your view as an operator Yeah, so I mean, I guess I come in this war of kind of engineering perspectives I worry about the practical impacts of this legislation actually I don't really worry about it because wonderful thing about average is that we have essentially a clean sheet So rather than having to defend Tom years of capital investment that I worry that legislation is going to Cause me to have to invest pots of new capex in fixing Actually for us, it's a fantastic opportunity to just be really good from a sustainability perspective and then And have that as competitive advantage, but I mean, it's most practically the things I worry about The the dynamic seems to be that people want lower pues zero w ues and to get rid of all the nice refrigerants That used to make those things possible Now that's quite a hard circle to square um because typically The kind of you kind of think of those as things that pop up and down as you push one down The other two kind of pop up and solving all three of them in one step is a really interesting challenge And so I think that that's quite quite a big one And certainly on the refrigerant side. That's really massive because we've benefited from these quite sophisticated refrigerants that enable us to do Lots of really interesting things and have low pues and zero water usage But getting rid of those and replacing them with butane or propane makes zero w e And low pue quite hard So that's an interesting one. And I'm not quite sure anyone's quite Squared that that circle and the other big one that I think no one's really talking about but it's massive Um, and I'm sure Matthew can talk more about it knowledge be the me is the banning of sf6 So you know sf6 essentially is what we have in all of our breakers I know Shido electric have these lovely vacuum breakers that they've solved the problem At least for Shido electric customers But there is a whole bunch of a whole bunch of electrical breakers that Like air plus max shoes as well But they all rely on this gas that's being banned that no one really talking about but that's that's massive, right? So, I mean, you know, think about two things that we care about and data sensors Vector distribution is pretty important and cooling is quite important and both of those really key systems Um are already being penned in in terms of how we design and manage them So, I think that you know, there's a big shot coming frankly for perhaps some Unsophisticated investors who invested in properties that have a you know, kind of headline PUE of 1.3 It meets the objective of their green fund. Um, but oh by the way, it's got really high wwe Um, it's maybe in an area with climate stress Um, and it's got all these breakers for the sf6 that can need to be replaced So there's this this is quite there's quite a lot of capital investment coming I think in the sector Potentially and actually sometimes building replacing the some of this infrastructure in the existing site is significantly more expensive than building a new one So I think the the dynamic that I think is really interesting is how some of the incumbents are going to deal with You know really going to their investors and say actually we've got really significant capital requirements And that's not going to have more capacity. It's not going to make the site operate more efficiently Actually it might increase our PUE But it's going to keep us on the right side of the law. So um, certainly I think that that's a really interesting dynamic And I'm looking forward to see how that one plays out. No, I'm not at one of those large incumbents anymore No, those are interesting points And I mean there's a lot of conversation as well around the the investments into sustainability Um, and how much money was porting to the industry over the last two three years And how some operators are having to to renegotiate Those terms because things have not been achieved Um On what they agreed So that's that there's something that apparently it's happening in the backstage as well And it'll be interesting to see what happens. And if that's going to impact the 2030 goals Of some of for some of the larger players I was going to ask as well. So because I mean we've over of course gone through some of the the upper incoming markets We've gone through sustainability a little bit regulations But then of course ai is the other big words They in 2023 sorry the the world of 2023 really is ai Well two words the fish intelligence How I mean one how do you see ai really affecting the development in a good way the development of data centers Towards secondary locations And what are the trends are you seeing beyond of course in ability power ai is there's something else happening lurking underneath that's driving An operate into a next location Well It's a fascinating topic the ai one I had a breakfast seminar on this Before it hit the press, but let's not kid ourselves ai has been a topic of conversation for 15 course years When I was in the software industry ai was a conversation machine learning was real So this isn't modern day stuff. It's just come up to the surface for everybody The impact for the data center space from a from my perspective what I get fortunate enough to see Is that it's it's writing a new chapter in terms of what it is We're expected to be doing to enable ai To evolve the way that the world seems to want it to go if you follow the trend of the digital economy growth Plug ai into that it's going to be exponential growth as a result of it But it has created a new chapter that we have to all learn to write read and understand that chapter being one where most likely and The the things we're seeing sort of go off faster the milk goes off in the fridge as one of my colleagues reminds me every single day ai is shifting and changing in real time But one of the things I think we're going to see from a location perspective is it used to be location location location with power You know in brackets It's going to become where's the power? Where's the power? Where's the power? And the location actually can be far more flexible for for these big Let's call them mother ships mega gigawatts sites The learning centers where the brains evolve Can pretty much be in the middle of nowhere the inference sites can be back close to your typical Locations, which is going to be a driver for increased demand in your typical locations The third site location. I think this is where it maybe plays nicely into Atlas Edgeshounds Is it we've been looking for a case study for the likes of the edge for a long time? I think that artificial intelligence is only going to give us that that edge case study in A deep way in a meaningful way that the inference sites based out of frankfort Will be very helpful will be necessary to execute whatever the mothership has created But you're going to want that execution getting even closer to the people whether it's through a lamppost in the car Or through an edge data center that might only be 500 kilowatts I only be a megawatt in size, but is servicing an artificial intelligence set of widgets to Porter's head or somewhere else. So it's a fascinating Chapter I think in the data center world and it's one that continues to evolve I'm glad you said it be around for a while, Tom. My master's degree. I finished in Edinburgh in 2003 in machine learning And no one really believed me that I studied it, but now everyone I only did it on my own. There was no practical use of machine learning in 2003 So I agree with everything Tom said. I think there's there's an interesting challenge Which is somewhat subtle in the training is that the workload is very asymmetric So they're kind of used to data centers having fairly consistent workloads, right? And that makes sense if you're investing a ton of capital However, whether you build it yourself or whether you're selling it to someone else You essentially want that capital to be being used, right? You build a ups you want the ups to be under 70% low Now when I train like chat gpt 5 right and a large language models are a bit of a toy Right, there'll be a new machine learning. That's more interesting the large language models But let's call it chat gpt 5. There's a very very high demand for power until there's not right because I finish training and now I don't need it anymore So if I've got a two gigawatt site so I can train my model very quickly There's a very kind of odd asymmetry in that work code So why I disagree with Tom actually is because the hardware that you use for training and inferencing is the same So whilst one could buy a lot of very expensive hardware Have it highly centralized and because we will think about data sensors as being really expensive things But of course actually the expensive thing is what you put inside the data center, right? So if I want my data sensor to be working all the time I sure as hell want my invidia gpu's to be working all the time because they're the real expensive thing So I definitely don't want is to go and build two gigawatts of gpu's in the middle of nowhere Train chat gpt 5 and then have nothing to do So I'm gonna jump in quickly David because it's Take chat gpt. It's a great example because that's just what everybody talks about There's far more artificial intelligence being built behind the scenes that we're not talking about So but take chat gpt to build chat gpt 3 it took 10 000 gpu's nvidia gpu's The projection to build chat gpt 6 and these are for nvidia now the projection in terms of how many gpu's that's going to need Is 10 million 10 000 to 10 million The amount of power that's going to be consumed is going to be in the gigawatts. It's not just going to be for chat gpt From a hardware utilization perspective the same gpu for do inferencing and training right So it what's clear is that inferencing is going to have a much steadier state in terms of workload, right? Because if I have 100 cttv cameras and they're all providing a feed to a gpu and I want to recognize You know someone getting you know mugged in the street, right? Then that that's a constant workload So I know that that workload exists all the time So what might make more sense is rather than doing the training centrally, right? Which also has the challenge of having to move the data to the training which could be really expensive And there's a lot of data that's required to actually use really trainer What might make more sense is to actually distribute training as well, right? Because then the same hardware that's doing the inferencing is also doing the training So rather than having a pool of hardware that sits there doing nothing and I agree with you you could build another model, right? So as long as I've got enough models being built I could fill up my brains and brain creations Honestly, the problem with the answer is a mixture of both of these things are going to happen But there will be some highly centralized hubs and then some smart engineer will find a way I will put a wager with anybody on this call right now that within the next five years Next five years. We are going to see some gigawatts sites for AI You're going to also see plenty of inference expansions around your existing regions around the existing zones And I genuinely think you're going to see The edge at last having a reason and a way to take off bigger than it has to date and that will be The ultra inference at the ultra edge locations. It's necessary Cascade, sorry, Matthew. Go on. I'll let you guys have it. Just to say I don't disagree with any of that, of course I mean, certainly I didn't study machine learning David. So I'd take your expertise on that. That's for sure and really state Tom You're the man, of course Yeah, we're seeing similar and we're hearing similar. We're talking to people having similar conversations You have the large learning models and you have inference, of course I think Some numbers I've seen is the the the training model will be 15 percent of the total demand Actually, the inference will be 85 percent of the total. Whatever the total demand is and by when we, you know It's a bit fuzzy, of course the timing is a little bit fuzzy But I do believe we We will see locations like the Nordics be be leveraged for the Learning models for sure You said David. Yeah, what happens when the the training model is finished? I mean If we've got chat GBT two three four five six, is there a chat GBT 24? I don't I don't know, you know Does it stop or what else comes down the line? We don't know but I think we have seen Or we are hearing that locations like Nordics will be used for learning Models and then the inference David, I'm sure you'll come into play there is a huge opportunity for us It's just when that really starts to hit the ground and when that really starts to take off What it does pose though, you've talked about some of these is technology and design challenges I guess this is um On the mind of schneider electric for example in what technology we need to develop to support These large loads you talk about Nvidia. I was chatting to them a few weeks ago. They've developed a 10 u GPU that is 100 kilowatts of That requires 100 kilowatts of power to mine You put that in a 42. You rack four of them 400 kilowatts now And they're not stopping by the way, but we're not seeing 400 kilowatts at any stretch at the moment, but it's not beyond Uh, the width of man to see maybe up to some people are saying 60 to 80 kilowatts per rack of of demand for these Learning models we need to change the way we design we need to look at the way we design data centers Cooling is going to be a big challenge. And maybe we see the emergence of a really quick cooling, but um, you know Proximities one Technology and design is the other of these data centers for sure. We're going to see we're going to see a little bit of a shift A bit fuzzy We're still all trying to figure out as you can tell by this call where and when I think is the 64 million dollar question I think the design piece is a really valid point you make there Matthew in the sense You know, I look back at the facebook story that we were all horrified to see earlier this year Facebook cancels all these build construction projects And then people suddenly dawn them that you know, they're not cancelling them. They're pausing them and the reason they're pausing Look at redesign and the design piece of this whole conversation is so critical because Yes, you could put rear door cooling You know up to 60 But no no exactly As soon as you get beyond that and I'm stretching it here 40 60 Excuse me. You've got to be looking at, you know, how can you bring water shock horror? How can you bring water into a data center something you'd never talk about five years ago? Today people have got to talk about proper water coming direct to cooling chip But also I mean the the the tier rating of these data centers. I talked to one Hyperscale and no names nothing of course And they're alluding to the fact they are considering what their design looks like in terms of the tier rating for the the learning model data center training data centers Do they need Generators do they need ups? Look, I don't want to say that sort of stuff because we sell ups But you know, it's just a matter of the fact that they're looking at the way they design data centers Do are they mission critical or not? Or are they like Mining data centers exactly You know, there's a lot of considerations that people are looking at in terms of design And the location may support them in the cooling as well. So yeah, Tom I agree with you. It's interesting. Yeah, I think I agree with typical We we say three three plus whatever tier 80 that is that's the mission critical side of the application Yeah, and I mean I agree with everything you said and I was actually at a conference in Toronto a few weeks ago And this was exactly one of the main topics of discussion and there was so much discussion about several operators Not even just hyperscale even smaller than hyperscalers They have had to change the designs within six months of them being designed And they went back into the engineering board the architecture board Just to to really redevelop them to to be able to to fit in all the new equipment and what's coming and I think Tom what you were saying about the five years I think it's going to be probably the lesson that to be honest for the gigawatt scale Facilities, there's one already being built which will be announced soon Which I will have the exclusive on that one. So they'll be I'll give more of that But At the present it is I think things will probably shift a little bit Over the next five ten years, but it is and I think the Nordics is quite interesting because it is really coming through a second coming Um, but but guys, thank you so much for talking to me David Matthew and and tom That's what we have time for today. Unfortunately It is this good last it could be a days on end conversation But on behalf of the speakers, I would like to thank everyone for tuning in either here here on linkedin youtube x or jsa.net Our next session will take place in the week's time on november the 8th at 3 p.m UK time and we'll delve into the sustainability energy hubs of europe and green data center development Make sure you visit our linkedin page to sign up for november the 8th and with that We wrap up today's conversation look out for the playback of the session coming soon to jsa tv youtube linkedin X and more from me and the team at jsa europe happy networking