Host Chris Adams is joined by special guest Karl Rabe, founder of WoodenDataCenter and co-lead of the Open Compute Project’s Data Center Facilities group, to discuss sustainable data center design and operation. They explore how colocating data centers with renewable energy sources like wind farms can reduce carbon emissions, and how using novel materials like cross-laminated timber can significantly cut the embodied carbon of data center infrastructure. Karl discusses replacing traditional diesel backup generators with cleaner alternatives like HVO, as well as designing modular, open-source hardware for increased sustainability and transparency. The conversation also covers the growing need for energy-integrated, community-friendly data centers to support the evolving demands of AI and the energy transition in a sustainable fashion.

Learn more about our people:

Find out more about the GSF:

The Green Software Foundation Website
Sign up to the Green Software Foundation Newsletter

Resources:

Windcloud [02:31]
Open Compute Project [03:36]
Software Carbon Intensity (SCI) Specification [35:47]
Sustainability » Open Compute Project [38:48]
Swiss Data Center Association [39:07]
Solar Microgrids for Data Centers [47:24]
How to green the world's deserts and reverse climate change | Allan Savory [53:39]
Wooden DataCenter - YouTube [55:33]

If you enjoyed this episode then please either:

Follow, rate, and review on Apple Podcasts
Follow and rate on Spotify
Watch our videos on The Green Software Foundation YouTube Channel!

Connect with us on Twitter, Github and LinkedIn!

TRANSCRIPT BELOW:

Karl Rabe: That's a perfect analogy, having like a good neighbor approach saying, "look, we are here now, we look ugly, we always box, you know, but we help, you know, powering your homes, we reduce the cost of the energy transition, and we also heat your homes."

Chris Adams: Hello, and welcome to Environment Variables, brought to you by the Green Software Foundation. In each episode, we discuss the latest news and events surrounding green software. On our show, you can expect candid conversations with top experts in their field who have a passion for how to reduce the greenhouse gas emissions of software.

I'm your host, Chris Adams.

Hello, and welcome to another edition of Environment Variables, where we bring you the latest news and updates from the world of sustainable software development. I'm your host, Chris Adams. How do you green the bits of a computing system that you can't normally control with software? We've discussed before that one option that you can do might be to shift where you run computing jobs from one part of the world to another part of the world where the energy is greener.

And we've spoken about how this is essentially a way to run the same code, doing the same thing, but with a lower carbon footprint. But even if you have two data centers with the same efficiency on the same grid, one can still be greener than the other simply because of the energy gone into making the data center in the first place and the materials used. So does this make a meaningful difference though, and can it make a meaningful difference? I didn't know this.

So I asked Karl Rabe the founder of Wooden Data Center and Windcloud, and now increasingly involved in the Open Compute Project, to come on and help me navigate these questions as he is the first person who turned me onto the idea that there are all these options available to green the shell, the stuff around the servers that we have that also has an impact on the software we run.

Karl, thank you so much for joining me. Can I just give you the floor to introduce yourself before we start?

Karl Rabe: Thanks, Chris. This is an absolute honor and I'll have to admit, you know, you're a big part on my carbon aware journey, and so I'm very glad that we finally get to speak. I'm Karl, based out of North Germany. We initially, I always say I had a one proper job. I'm a technical engineer by training,

and then I moved into the data. Then I fell into the data center business, we can touch on it a little later, which was Windcloud, which remains, which was data center thought from the energy perspective, which is a very important idea in 2025. But we pivoted about four years ago to Wooden Data Center, probably can touch upon those a little later, in also realizing there is this supply chain component to the data center.

And there are also tools to action against those. And I'm learning and supporting and providing, you know, as a co-lead in the data center facilities group of the OCP where we work, you know, with the biggest organizations directly in order to shape and define the latest trends in the data center

and especially navigating the AI buildout in somewhat of a, yeah, sustainable way.

Chris Adams: Okay, cool. And when you say OCP, you're referring to the Open Compute Project, the kind of project with Microsoft, Meta, various other companies, designing essentially open source server designs, right?

Karl Rabe: Correct. That is the, initially started by then Facebook now Meta, in order yeah, to create or to cut out waste on the server design. It meanwhile involves and grew into cooling environments, data center design, chiplet design. It's a whole range of initiatives.

Very interesting to look into. And, happy to talk about some of those projects. Yeah.

Chris Adams: All right, thanks Karl. So if you are new to this podcast, my name is Chris Adams. I am the director of technology and policy at the Green Web Foundation, a small Dutch non-profit focused on a fossil free internet by 2030. And I also work with the Green Software Foundation, the larger industry body, in their policy working group.

And we are gonna talk about various projects and we'll add as many all the show notes to all the links we can think of as we discuss. So if there's any particular things that caught your eye, like the OCP or Wooden Data Centers, if you follow the link to this website, to this podcast's website, you'll see all the links there.

Alright then Karl, are you sitting comfortably?

Karl Rabe: I am sitting very well. Yeah.

Chris Adams: Good stuff. Alright, then I guess we can start. So maybe I should ask you, where are you calling me from today, actually?

Karl Rabe: I'm calling you today from the west coast of the North Sea Shore in northern Germany. We are not a typical data center region for Germany, per se. We, which is Frankfurt, you know, 'cause of the big internet hub there. But we are actually located right within a wind farm.

You know, in my home, which is, initially was, you know, home growing up and turned to my home office and eventually to what was somewhat considered the international headquarter of Wooden Data Center. Yeah, and we're very close to the North Sea and we have a lot of renewable power around.

Chris Adams: Oh, I see. Okay. So near the north of Germany, near Denmark, where Denmark has loads of wind, you've got the similar thing where, okay. So

Karl Rabe: Yeah, absolutely. Yeah.

Chris Adams: Oh, I see. I get you. So, ah, alright. For people who are not familiar with the geography of like Europe, or Northern Europe in particular, the north part of Germany has loads of wind turbines and loads of wind energy, but lots of the power gets used in other parts of it.

So, Karl is in the windiest part of Germany, basically.

Karl Rabe: That's correct, yeah. We basically have offshore conditions on shore. And it's a community owned wind farm, which is also a special setup, which is very easy to get, you know, the people's acceptance. We have about a megawatt per inhabitant of this small community.

And so this is becoming, you know, the biggest, yeah, economic factor of the small community.

Chris Adams: Wow. A megawatt per, okay, so just for context, for people who are not familiar with megawatts and kilowatts, the typical house might use what may be about half a kilowatt of constant draw on average over the year. So that's a lot of power per person for that space. So that's a, you're in a place of power abundance compared to the scenario people are wondering where's the power gonna be coming from? Wow, I did not know that.

Karl Rabe: No, that, is, yeah, that is the, so it's a bit of that background, so to speak. We are now trying to go from 300 megawatts to 400 megawatts. There has been, you know, Germany's pushing for more renewable energy, and we have still some spots that we can, under new regulations now, build out.

And the goal or the big dream of our organization, the company running this wind farm for us is trying to produce a billion kilowatt hours per year. And so we're now slightly below that and we're trying to, Yeah, add another, yeah. For, we need to reach probably another 25 percent more production. And, it is, so to speak, you are absolutely right, we are in an energy abundance and that was one of the prerequisites for Windcloud. 'Cause you know, the easiest innovations, is one and one is two. And so we have in, we had energy, I was aware that we also had fiber infrastructure in the north to run those set wind, parts.

So we said, why don't we bring a load to those? That was the initial start of Windcloud.

Chris Adams: Okay, so maybe we should talk a little bit about that. I hadn't realized the connection between the geography and the fact that you're literally in the middle of a wind farm, which is why this came together. Okay. So, the, so as I understand it, and now this makes sense why you are so involved in Windcloud.

So for context, my understanding of Windcloud is it's essentially a company where rather than like connecting data centers via big power lines to like somewhere else where the actual generation is miles away from where the data centers are, the ideals instead was to actually put the data centers literally inside the towers of the wind turbines themselves.

So you don't need to have any cables and, well you've obviously got green energy because it's right there, you're literally using the wind turbine. So, apart from this sounding kind of cool, can you tell me like why you do this from a sustainability perspective in the first place?

Karl Rabe: Yeah, so the way we discovered that I wanted to, and this is the, probably the biggest reference that I can give on the software developer front, and I came out of a study in the UK. We had a really nice cohort.

We were constantly bouncing ideas off of each other. I wanted to actually build small aircraft, because we have a wind farm and we have wealth with that. We actually have people building small planes in our location. They told me I needed about 5 million euros to do it, which I didn't have.

So I started pivoting to a software idea. And why the software did to host that, I just quickly discovered, you know, the amount of energy going into data centers, the amount of, you know, associated issues, and back then, 2015, 16, we were literally just discovering the energy aspect of it. We need didn't discuss, you know, water and land use and all of that.

We really focused on the energy and then we say, "look, well wait a second. You know, we have all this excess of energy. We literally cannot deliver that at this point. So we have a very high share of shutting down our wind turbines when there's just too much energy to move it around. Why not bring the data center as a flexible load close to the production, and enable, you know, sustainable compute

to then send package rather than energy, which is way easier, you know, over the global fiber grids." And that's how I got started and fell into the data industry. Big benefit and big learnings from that stand that I didn't know nothing about data centers. And as an engineer, a lot of things were not adding up.

We looked at the servers back then, and even then it said, okay, this is good, you know, to run from 15 to 32 degrees. I said "32 degrees? Why? What is data center cooling and why is data center cooling? We don't have 32 degrees in the north." Most likely now, we probably ought to do within eight years.

Karl Rabe: But the important thing was really challenging this, and we started with very little money and we couldn't afford like the proper fancy stuff that all of this data center make, you know, like a chillers, you know, spending electric energy to cool something which really does not need cooling, in my opinion, up to now.

That was the start of this, you know, and so this is, the company of Windcloud is still ongoing. We had what we were, what we had as a huge problem. And I was always, my gut feeling for this was always we need to find a way to be able to compete with the Nordics.

So we have renewable energy, but we need to have it cost effective. And that was something that we tried two or two and a half times, I would say with the, with always having a legal way to access the energy in a proper setting. It was always extremely difficult and extremely frustrating also because the German energy system is very complicated.

It is, you know, geared or developed from a centralized view of this, and is benefiting, you know, large scale industry and large scale energy companies, to putting other terms, as you know, in, you're probably familiar with the, Asterix comics. You know, that far off and north in Germany that probably people, you know, there was a bit suspect, you know, what we're doing there or now we start producing energy and now we also want to use the energy so that is not adding up.

It's very hard and close to impossible to access your own produced energy at scale, you know, which is in an abundance. And that was, yeah, that was something what we always faced, which led to other innovations. So we build the first data center or one of the few data centers to reuse the heat in Germany, putting into an algae farm.

And we trying to create really efficient, PUEs already back then, you know, whereas the industry stranded is quite still quite high in ours. Claim I never had enough money to build a data center with a PUE above a 1.2, or even 1.1. The first servers were cooled with a, you know, a temperature regulated fan, you know, out of the, we built with the same guy who built, you know, a pig stale for my father, you know.

that was, you know, we nearly didn't call it Windcloud. We nearly called it Swines and Servers,

Chris Adams: Okay. Pigcloud.

Yeah.

Karl Rabe: Yeah, Pigcloud, but it could have been, you know, could have been misleading. And the, so the good thing turning out of that, you know, and going back to that, to those struggles in getting started is that we were forced to uncover a lot of the cooling change and the energy distribution change, which are were not, you know, not really adding up for us.

And that is, you know, still one of the biggest support for us to build efficient data centers and to create, you know, sustainable solutions.

Chris Adams: Okay. Cool. Alright then. So. Okay. There's, I didn't realize anything about the Schweins und Servers aspect at all actually. Would you even, I'm not sure what German is for server would actually be in this context. Was it literally gonna be Schweins und Servers, or?

Karl Rabe: Yes. Some. So, yeah, something like that.

Chris Adams: Okay. Wow. That's, I was not expecting that.

I think Windcloud sounds a bit better, to be honest.

Karl Rabe: Yeah, thanks. No, the brand, the name is great. I think it's still, yeah, I'm very simple like that. You know, we had Windcloud, so we take wind, we make cloud. Now we are have, we are Wooden Data Center. We build data centers outta wood. So we, but there's this, but it's, to be fully honest, is right now, is so to speak,

we call Wooden Data Center, but what we do is try to decarbonize the data center. So wood is obviously, is a massive component of that, but we do see real good effort in the supply chains. Happy to go into that a little later, but there are some examples from fluids. We just found, you know, bio-based polycarbonate for hot and cold containments.

So the amount of components throughout the data center that have, that has a bio-based, ergo, a low carbon alternative is ever so increasing.

Chris Adams: Can I come back to that a little bit later? 'Cause I just wanna, touch on the

Karl Rabe: Yeah. no.

Chris Adams: So the wind thing, so basically Windcloud, the big idea was putting data centers in the actual wind turbines themselves. So that gives you access to green energy straight away, because you're literally using power that otherwise either couldn't be transmitted because there were, because the pipes weren't big enough essentially in some cases.

And, I guess plus point to that in that if you are already using a building that's already there, you don't have to build a whole new building to put the data centers inside. So there's presumably some kind of embodied energy advantage there because there's a load of energy going into, kind of, that goes into making concrete and stuff that you don't have to do because you are already using an existing, like, building, right?

Karl Rabe: Yeah. So to clarify on that, it is good that you touch on that because there is, this is literally is a bit of a crossover because the company you're referring to is Wind Cause, which is a good friend of ours and they are using the turbine towers.

Chris Adams: Ah,

Karl Rabe: They can do so because they use a little bit different type of turbine.

And they're also based in the south of Germany, we had the same idea because it's also very difficult to build next to a wind farm. The big difference is that the towers used at Wind Cause, they are concrete and they have quite a lot of space. They're about 27 meters wide. because of the initial, discussion that we have onshore, or offshore conditions onshore, we have steel towers, which are shorter and hence don't have this big diameters.

You know, we build tall. And so we always had the challenge of still needing a data center. And so that's where our learnings inspirations came from For Wooden Data Center. But we still tried to reuse existing infrastructure. So we were at one point within the Windcloud journey,

I was the co-owner of a former military bunker area. And so we wanted to place within those long concrete tubes, we want to place data centers in order to yeah. Have a security aspect and don't need, you know, a lot of additional housing or even bunkering. And there's obviously the dodging bullets where has spent a lot of concrete and steel concrete in order for those facilities.

Chris Adams: I see. So you're reusing some of the existing infrastructure, so rather than building totally new things, you're essentially reusing same, you're reusing stuff that's already had a bunch of like energy and emissions spent to create it in the first place. I see. Okay. All right. So,

Karl Rabe: And, back then, you know, also to, because it's such a short time back then, really need to emphasize that we were, we really, you know, only had a hunch and a feeling, oh yeah, sort of has CO2 associated to it and probably also the building of a data center.

You know, we have, we really, it was so hard to quantify, and I think we still, carbon accounting is still, is somewhat of, not wizardry, but it's really hard to pull the right numbers. You know, only two years ago at the OCP Summit, so in a Google presentation, the range that they mentioned, you know, for steel and concrete carbon was, you know, 7 to 11 for equally both. So the range of the total uncertainty, I feel, is quite high. You know, and this is the biggest, one of the biggest and most funded, best funded organizations in the world. You know, we're still not being able to get it more concrete, you know, and that's something we really need to work with the industry and supply chains in order to be even aware to specify the problem.

Chris Adams: So, can I unpack that for a second before we talk a little bit about this? And so you're basically saying even the largest companies in the world, they don't necessarily have that good access to know how, what the carbon intensity of the concrete they've used in one data center compared to another one,

it can quite, it can vary quite a lot. Is that what you're saying there?

Karl Rabe: So this was basically specifying the global numbers for steel and concrete. So, I do believe that we have now relatively good visibility for our own builds and projects and also what we do now moving forward. But to really try to grasp the global problem of it, that was still, you know, two years ago was still had this high uncertainty, you know, 'cause we were working with numbers,

maybe they're now five years older, we don't know the complete, you know, build out of every city, every building globally. You know, it's just a lot of guesswork in that, globally. And so I especially believe that although we, Wooden Data Center, the amount of innovation that is put into concrete, you know, has the potential to drastically reduce that for buildings.

You know, the, it was a, it's definitely still a huge problem in, for the data quality and the data, yeah, the emissions, you know, guesswork that's in there, you know, and a lot of those things are based on scenarios, you know, and those are getting ever so more real. But the best example for Wooden Data Center is, there's a comparison comparing a steel concrete building to a CLT one,

Chris Adams: Yeah.

Karl Rabe: and it is assuming that after 20 years, it's only living for 20 years, which, you know, can easily be 200 years, but that afterwards it is being reduced into, you know, building chairs or tools or toys. But if you take then the CLT and burn it, then obviously you have a zero sum gain. Every, all the carbon that was stored.

It's Cross-Laminated Timber, you know.

Chris Adams: Yeah. So this is the kind of like the special, the, this, it's a special kind of, essentially like machined timber that is, that provides some of the kind of strength properties of maybe steel or stuff like that, but is made from wood, basically, right?

Karl Rabe: Correct. So we need to stretch the importance that this is actually a material innovation. It's a relatively young material based on a, I think a thesis, PhD thesis from Austria. And so we only have CLT or cross-laminated timber for about 25 years.

Chris Adams: Oh, I see.

Karl Rabe: Or maybe now 26, years.

So the, you probably are familiar, or you have seen there are those huge wooden beams in, you know, storage buildings.

Chris Adams: Yeah.

Karl Rabe: Those are called GLT, like glue laminated timbers. And the difference is those boards are basically glued in one direction and they're really good for those beams or for posts.

But to have like ceilings, walls, and roofs, those massive panels, you now have the material of cross laminated.

Chris Adams: Oh, okay. In both directions, right? Yeah.

Karl Rabe: Correct. And those now enable like full massive wooden buildouts. And that's something, and so the biggest challenge is that we, if we say wood, then the association we probably will touch on now or later is fire.

Chris Adams: Yeah.

Karl Rabe: But in reality, in nature, we don't have those massive panels which don't, you know, just flame up. They have, they're fully tested and certified to glim down, which is, you know, they turn black and then they slowly, you know, in a thousand degrees, they slowly, you know,

shrink

Chris Adams: Like smolder, right? Yeah.

Karl Rabe: Yeah. And so, but, the, how we design data centers is basically factor in this component, and we are able to create really fire secure data centers built out of those new wood materials basically.

Chris Adams: Okay. All right. So a lot of us are typically thinking of data centers as things made entirely with wood and with steel, concrete and plastic all over the place. And essentially you can introduce wood into this place and it's not gonna burn down because you have this material, which is treated in such a way that it is actually very fire resistant.

And that means you could probably replace, I mean, maybe you could talk to him a little bit about which bits you can replace. Like, can you rep, would you replace like a rack or a wall or like a roof? Maybe we can talk about that so we can like, make it a bit easier to kind of picture what this stuff looks like.

Karl Rabe: No, absolutely. I'm afraid I'm still, always very liberal in sending out samples to my clients, you know, but I don't have it here in my hand, but, so that is a very good the question, is basically like, if we would talk like slide deck or something that I'll try to show in terms of scope one, two, and three, what we can do and what we have now, and that it's like the, biggest component is in obviously the housing. You know, what is your building or your room of a data center? When you are touching on existing buildings we CLT is also ideal for building and building concept of existing large storage or logistic buildings to put in data centers.

We can build that up quite quickly out or create rooms very quickly in those, and there is other huge advantage of CLT is that we get those pre-manufactured and they just fit,

Chris Adams: Oh, like stick them together like Lego rather than have to pour

Karl Rabe: concrete?

Yeah. Little, Yeah, a little bit. You need like, a little bit of leveling foundation.

If you have an existing floor, still, some datas, you know, preferred to in the greenfield also have a new floor. But that is is something that it helps to, with those, we can create the IT room relatively quickly and then have the build out of those averaging up to 40% quicker times than traditional steel sandwich concrete, you know, data center.

So it is enormously easy to work too. It's very precise to pre-design and pre-manufacture and then very easy to work with. If there's something, if there's a problem on site, you know, you just crank out the chainsaw and adapt and adjust.

Chris Adams: Okay. Just to carve it down a bit.

Karl Rabe: And yeah, so to speak. But once you have then those assembled and secured, it has like a lot of mass to it and a lot of volume to that which creates very good fire protective

physical resistance and availability properties. And that is something that we now, it's really being seen as one of the core benefits. You know, the speed what we can build this out.

Chris Adams: Oh, ok.

Karl Rabe: We have introduced wooden racks, and we also see more and more attention for those.

Chris Adams: Wait, sorry. Can I just, you said wooden rack, like as in the big steel rack that holds the servers themselves, you could, that could be made of wood as well now, so you'd have like a rack thing holding a bunch of servers, right?

Karl Rabe: Correct, So we built this also. We have, one of our clients has send us like a server casing and ask to also think about to do the casing, but we probably, we're not a hundred percent there yet. In order to do that, we would have, we would've an idea, in terms of the spirit of OCP, which is, you know, like,

reduce and cut out stuff. You know, one vision of that would be just a wooden, you know, board where you have dedicated spaces. You slide in your main board, connect power, connect liquid cooling, have fans on the back and then cycle through only the boards. Remove, you know, not even fancy, but just base frames of a server.

But right now the, it's a combination, for the 19 inch standard and also the OCP standard, to use, you know, reduce up to 98% of the steel in those constructions and then only have functional parts in order to stick in the servers made from steel railings and have then wooden frames.

And we do that for the OCP format, which is very popular. We get a lot of the special requirements because we are the only ones who producing like a small version of the rack, which, the OCP has a lot of advantage, but the base rack format is a two meter 30 high, which is like a really hyperscale, you know, mass density approach.

Which doesn't fit even through the doors of most data centers I know, you know, they still have relatively, you know, standard two meter high doors or able to fit in like a 42 inch rack. But you need like a very special facilities because those racks come also pre-integrated and then you roll them in place.

So you need a facility that has high doors, ramps with small inclines, you know, or no ramps at all, in order to be able to place a fully integrated rack. We started building OCP racks because back at the time only hyperscalers were really getting those, and we wanted to do more of this open compute format and were able to offer that.

And the version three rack, you know, was a good candidate to convert to a wooden based structure.

Chris Adams: All right, so we'll come to that a little bit later because I actually came across some of your work when you were building, designing some of these on YouTube so people can see what all this stuff looks like. But if I just come back to the original question, essentially, so it sounds like you can replace quite a lot in a data center.

So you can replace the shell of the building, like literally green the shell by replacing the concrete, which is one of the largest sources of, you know, creating concrete and cement is one of the largest sources of emissions globally. So you can switch, you can move from a source of emissions to, is it a sink?

Cause CO2 and carbon gets sucked out of the sky to be turned into trees. So you've gone from something which is a source to a sink and that, and you can replace not just the walls, the outer building, but also quite a lot of the actual structure itself. Just not the servers yet.

So that's probably like a, I mean, maybe I could ask you then like, If I'm switching from maybe regular concrete and regular steel, I mean, is there any, like, do you folks have any idea about like what the kind of changing quantitative terms might actually be if I was to have like an entirely concrete, entirely steel data center and then replaced all of that with, say, wooden alternatives, for example? Like is it like a 5% reduction or is there any, like, what kind of changes are we looking at for the embodied figures, for example?

Karl Rabe: So the conservative industry figures are somewhat off between minimum 20%, only having the production change up to 40%. So Microsoft, we, the good thing also we have to mention is that we are an industry now. Microsoft announced those productions I know the other hyperscalers are looking at that.

We only, in Germany we had two other companies started getting into construction. That's why it's for us really important to be on the decarbonization path.

Chris Adams: Ah, I

Karl Rabe: see.

So we do come with our own data center, even concepts and philosophies, which I can talk about a little later.

But coming back to the point it is still very hard to quantify. And the, but really positive things about carbon accounting or calculations, as I mentioned, we now have as a data center, we have this negative component, which I have to laugh 'cause an engineer immediately and said, can we then just use more wood?

You know, can we make the wall thicker? You know, obviously yes, you could do that, but there's a cost to that and there's also, you know, it be betrays the idea, you know. But, the really exciting thing is that I now go to show, from show to show, and I was two weeks ago in London

and just on the flight somebody showed me, a picture of an air handling unit inside of a wooden enclosure. And I was chasing an hour through the London show, 'cause I assumed it was there, but it was on a, it was on a different show. but that is the kind of things that we can really think about is enclosures.

So also we have started, we have one, well, for the OCP reg or for this AI build out, we have also created a rear door, which is, so to speak, a wooden rear door. So the fans are traditional, the heat exchanges obviously needs to be traditional, but it is also like a micro aluminum micro channel heat exchanger, which is derived from other industries, which is, you know, helping mass production, reducing cost, reducing emissions.

And that is the other thing that is happening in the industry that

we're trying to find, not data center specific solutions, but rather find mass produced industry solutions and adapt them to the data center in enhanced reducing cost and time.

Chris Adams: Alright. So in the same way that basically cross laminated timber and the use of wood is something that has been in use in not just in the data center industry, like people make, what are they called? Are they called plyscrapers? You know, skyscrapers with wood. Plyscrapers.

It's, so the idea was that, okay, things which are made, being made in volume here can be made more efficiently and like this is one way that you are adapting 'em to a new domain.

And it may be that if people are getting much, much better at making say very efficient heat pumps, 'cause they can cool things down as well as heat them up, that might be another thing you're looking at. Say, "well actually that might be able to be used in this context as well." Okay. Alright. And if I am, so if I go back to the original thing about saying, okay, we're looking at possible savings maybe 20% up to like possibly 40%, like that's the kind of

Karl Rabe: yeah. That's the range that we have, you know, I think, so the problem is do we, if, did Microsoft evaluate with IT or without IT? So for the facility, I think we can potentially come to net zero approach, which we, you know, but by first principle, I think we can at least achieve realistic reductions to let's say 80, 70-80, 85% with those tools that are set, you know, basically the easy steel replacements, the, like, the rack, the enclosures, the housing, fluids is something we have. There's a very interesting, you know, no-brainer replacement for fossil diesel on backend generators.

It's a liquid called HVO

Chris Adams: Yeah. Let's come to that in a second actually. 'Cause I did wanna ask a little bit about some of the things you can do for the fuel here. So maybe if we just, so basically the, so there are some savings available there and these should be something that you could, some, this is something that should show up in some kind of numerical description.

So if you had like, maybe two data centers and one was using wood in strategic places, then the embodied carbon should be lower. So if, I mean, if I was looking for this there like a label to look for or a standard I can look for? Because in the Green Software Foundation we have this idea called Software Carbon Intensity, which includes like the carbon intensity of the energy you use and stuff like that.

But they also look at the building itself. So theoretically, if you had a wooden data center and a bog standard concrete data center, you know, if you run your code in the greener data center, you would probably have a better score if you had access to these, the data or stuff like that. Do you know, like, do, any places share this data or have like a label for anything like this?

Karl Rabe: They definitely share the data. I, for example, so we definitely also need to Eco Data Centers in Sweden's and we, which were, you know, basically we approach to them. Our whole world was shook. You know? It's like, oh, so we come from this energy perspective, but they didn't have idea and they build it, you know, sustainably. They build it sustainably.

So we need to change, you know. That was, you know, it was a huge eyeopener. And they also are the few first ones to, I'm not sure if they used like the LCA method, but they were quantifying the embed carbon and are certifying to you annually too, as a client, which I think is the way to go.

And we need to figure out how to standardize that. I assume there's potentially a standard that we can use. I know that other data center providers are building sustainably and putting this effort forward. But we don't have a unified label yet, I'm afraid.

Chris Adams: Okay. Well this

Karl Rabe: I know that some, also challenge, like, there's like a data center climate neutral act and some of them specifically exclude scope three, which, you know, I know where they're coming from.

Also in Germany and Germans, you know, they're all about energy efficiency. They love to talk about, you know, just the, energy and the scope two, basically. But then, you know,

Chris Adams: Most of the

Karl Rabe: missing out, this dimension, you know,. Missing out the dimension is being faithful to your girlfriend or wife, you know, like three days out of a week.

You know? It is, it's not

Chris Adams: You are not showing the full picture, right?

Karl Rabe: Yeah. You're not, doing it at all basically. Right. I would probably, you know, just need to Google it and there are, you know, building labels that you could be used in construction. Quantifications, I'm sure, but there's not yet like a data center specific label.

There is good work also in OCP to do metrics and key performance indicators, and they're all looking at that and there is, I think they're trying to build towards something like real, like true net zero.

Chris Adams: Oh yeah. Okay.

Karl Rabe: But...

Chris Adams: So there are some, so there are some initiatives going on to kind of make this something that you could plausibly see, and, but it's quite early at the moment right now. So like, let's say that I, you know, we spoke before about, okay, I can run my computing jobs in one data center or I can choose to run it somewhere else.

These numbers don't show up just yet, but there is work going on. Actually, I've just realized there is actually a embedded carbon working group inside the OCP who have been looking at some of this stuff. So maybe we'll share a link to that, because that's actually one of the logical places you'd look for that.

Okay,

Karl Rabe: And they do real good work. They do a lot of good initiatives, happening there. There's also, it's Swiss from the Swiss Data Center Association. They also have a label, that is looking at some of this, and they want also to include scope three.

So this is coming up, but it's, not as easy as, you know, having an API, you know, pushing it to the software developer and saying, look, we have this offset because this was constructed, you know, with concrete or steel, and this is, you know,

Chris Adams: Okay. So we're not there yet, but that's a, that's the direction we might be heading towards. Okay. Alright. We'll add some links to that. And now I'd like to pick up the other thing you mentioned about HVO and stuff like that because you spoke before about, you know, Windcloud or wind node and like data centers running in,

or like, you know, relying on wind right now, we know it's a really common refrain that the wind doesn't blow all the time. And like it's news to some people that sun, that's, you know, it is not always sunny, for example. So there'll be cases where there'll be times where you need to get the power from somewhere and, you know, in the form of backup power. And like loads of data centers, you said before they rely on like fossil diesel generators, right.

And that can be, it's bad from a climate point of view, it's also quite expensive, but it's also terribly really bad from an air quality point of view as well, because, you know, people are up, you know, you can see elevated cases of like asthma and all kinds of respiratory problems around data centers and things like that.

But you mentioned there's options there to kind of reduce the impact or have like more responsible options there. Maybe we could talk a little bit about like what's available to me there if I wanted to reduce that part, for example.

Karl Rabe: No, happy to go into that. That is something that we are now thinking about quite heavily this year. And we're already presenting on two occasions, a sense. So the easy options in order to reduce your carbon on the scope one part for data center, which is basically, you know, that's just the direct burning of fossil resources and that is the testing of your backup generators.

The easy option for that is this second gen diesel, HVO 100. And the, when I realized the key feature of this fuel, which about 15% more expensive, is that it doesn't age. Fossil diesel and especially, you know, biodiesel, the first generation and fossil diesel with biogen, there's always, in Europe there you have a certain degree of mixed in of this, it ages through bacteria biologically.

So it's degrading. So, the, which is, you know, really bad because this diesel sitting there in a tank, you run it half an hour every two weeks, and you maybe change the fuel filter twice, once, twice a year. But if you really have an issue, you know, all of a sudden you use this diesel

for four hours and then your system, your filter clocks, and you still have a problem, right? If

Chris Adams: So your backup isn't a very good backup.

So backup needs to be a good backup. Yeah.

Yeah.

Karl Rabe: Yeah. so your backup can run

Chris Adams: you had one job, right? Yeah.

Karl Rabe: Yeah. Yeah. And so, how it's mitigated is people try to use 'pure' diesel or, you know, heating oil, you know, which is not so prone to it, but still ages. They are recycling or, you know, really pumping out the fuel and pumping it in again every three years or they continuously filter it.

All of this is either adding energy or cost. And so, the, this new form of biodiesel, which is, you know, your old frying fat, cracked with hydrogen to, is it looks very clear and it's very chemically treated that it's not really aging. People don't know really yet how long it stays.

They certified 10 years, potentially it's stays, good longer and is also burning cleaner. So

Chris Adams: Ah,

so it'sn't going to be bad

like bad air and stuff as well then?

Karl Rabe: Yeah. So for the majority of your enterprise IT, your standard data center that's around you, you know, cutting out the whole AI discussion, probably that's the easiest way to do something about that.

This is like a drop in replacement. You just, you know, you empty your tank, you put it in, or you burn your old fuel and put a new, that is something that is, you know, easily increasing the availability of your facility and you can change with that.

Chris Adams: Can I just try to like summarize that? So, because I don't work with data centers on a daily, so there's like basically fossil diesel, the kind of stuff that, you know, you might associate with dieselgate and like all kinds of bad air, air quality issues. And then the, kind of the other option, which is maybe a little bit more expensive, you said around 15%,

there's something called HVO, which is essentially like biodiesel that's been treated in a particular way to get rid of lots of the gunk so it burns more cleanly and works better as a reliable form of backup. So the backup is actually a decent backup rather than a thing which might not be a very good backup.

Oh okay. So that's like one of the things, and that's like the direction that we might be moving towards and that's kind of what we would like to see more of for the case where you need to rely on some kind of liquid fuel power. Right.

Karl Rabe: Yeah.

Chris Adams: Okay.

Karl Rabe: That is, I think is for most people, you know, just a very easy low hang fruit to just replace, you know, it does not, you know, most engines are certified for, nowadays, all engines run on it, you know, it's, it has the same, Yeah, criteria, properties like traditional diesel, the only thing that's different is it's 4% lighter, you know?

Chris Adams: Oh, I see.

Karl Rabe: So that's the only real on the spec sheet

Chris Adams: Oh, okay. Alright. So if I may, so that's one of the options. These, so you can replace fossil diesel with essentially non-fossil cleaner, slightly less, you know, less toxic diesel. So that's one thing that you might have in for your backup. Now, I've heard various people talking about, say hydrogen, for example.

Now hydrogen can come from fossil sources. So people, most of the time, actually, most hydrogen does come from basically cracking natural gas or methane gas, but it can come from green places. And that's why is, that's another option that you might have to generate power locally.

Karl Rabe: Is that something that people tend to use?

So I think the best, the best reference for hydrogen is like the champagne of our energy transition. You know, we need, we need to put in a lot of energy to put, to produce it. It's not easy to store, that we need a lot of facilities to actually create green hydrogen.

Karl Rabe: The majority of hydrogen is not green hydrogen, but it's gray or blue, which is basically

Chris Adams: like a carbon capture hydrogen, which is still a bit questionable. Yeah.

Karl Rabe: all based from fossil tracking, you know, so it's, it potentially, you know, you, you also have the same goal.

everything that we do for our clients is under this extremely short impact of time.

You know, we have solve everything within now, within five years time, not even five years. Right. And so that's also something that I'm always, you know, spark a good discussion. When we talk about SMRs, you know, have the big pushback for nuclear over in the US, and also in Europe we have voices for that.

And the short answer is, the three reasons I don't believe in it. They're not quick, you know, they're not cheap. Two projects were just, a year ago, there were two potential very, you know, hopeful projects for SMRs were canceled in the US, and half a year later it was a big thing.

The big solution. like, what changed, you know? And then the third point is that is the, very German, perspective, you know, all the fears or the, challenges around the fuel, like getting it mostly 70% from Russia or, then the waste, you know, dumping it somewhere is not solved, still.

And so, this is not a 2030 technology basically.

That's my, the point what we can do and what I'm happy also to link, there's a good article from some of the, hyperscalers looking into solar combined with batteries, combined with gas based backup. The gas based has the one flexibility that it can start fossil, can move to bio, and potentially also can run on hydrogen. So this is, in terms of the speed with which we are now deploying hundreds of megawatts, you know, every data center for AI is now, you know, 100, 200, 300 gigawatts.

You know, things that we did not,

yeah,

yeah. So it's things that we, you know, like yeah, we're discussing, you know, five or one to five gigawatts for the large people. And every other data center is all of a sudden is now a hundred megawatts, which used to be like a mega facility, just two years back.

that build out can only really be achieved not with grids or interconnects, those are too slow. This can only be basically with micro grids.

Chris Adams: I see. Okay.

Karl Rabe: Helping, you know, that are battery backed and gas based backed. And the big advantage of this is that if we think about the data center,

traditionally a data center is a data fortress, right?

You don't get in, data doesn't get out. It is, you know, is like a bank, you know, in terms of the security measures to do that. And also all of the infrastructure was handled that way. But if we are thinking about the UPS, and the genset not being sitting straight at the data center or only sitting straight at the data center but technically belonging to the utility and being able to provide flexible power, you know, because we have this, as mentioned, underlying flexible build out of renewable energy, and we need, you know, reliable switch-on power, which data centers all have. And so if we can put those together, there's a little bit of this working together, finding the right location where it would make most sense, and then allowing for SLAs and with clients to bidirectionally use batteries, gas turbines,

Chris Adams: Oh, I see.

Karl Rabe: engine power.

This would, you know, help our, yeah, help us to transition, especially if we go into, you know, renewable shares, 60% and above and at latest from 80% we need those backup technologies. And then, and that is coming back to the question of hydrogen. Hydrogen is a technology that would, is so expensive that it would need to run all the time, basically.

With renewable energy, we have high loads of

abundance

of energy and only need short times of flexible energy generation for which gas and batteries is virtually ideal. And so we promote this idea of an energy-integrated data center, which has the electrical part supporting into the grid and is also, you know, taking advantage of the heat reuse, especially for liquid-cooled facilities in order to give heat out.

And the benefit of that is not only from an economical perspective, but also we see more and more discussions about not in my backyard. If a data center is energy integrated, it's not a question, you know, it's a must have. And there's also a reason why it needs to be there, you know, in order to be able to stabilize the, your town grid or your local area.

And so that's what we are trying to promote. We got a lot of good feedback and we see the first, hopefully we'll have the first data center realized with a medium voltage UPS this year, which is like a first step in moving the availability components of a data center, the batteries and the gen sets to a higher area, which, a lot of the cost in a data center is from the low voltage distribution.

The power that you put in the batteries is also first transferred down, and then it's moved, you know, through the data center until it sits in the battery and then needs to go out. And all of those are rectification steps. And all of this makes, yeah, all of

Chris Adams: You lose, so do you lose power every single time you switch between them? Oh, okay. So it sounds like you, there's a shift from, like, data center as a fortress where, you know, you could do that before to like something where you have to be like a bit more symbiotic with your local environment because for a start, if you don't, you're not gonna allow it,

you won't be allowed to build it. But also it's gonna change the economics in your favor if you are prepared to like play nicely and integrate with your, essentially be a good neighbor. All right. That seems to be what you seem to be suggesting.

Karl Rabe: That's a perfect analogy. Having like a good neighbor approach. Saying, "look, we are here now, we look ugly, we always box, you know, but we help, you know, powering your homes, we reduce the cost of the energy transition, and we also heat your homes."

You know, and that is then, is then a relatively easy sale.

Chris Adams: Okay. So that points to possibly that, honestly, that points to quite a different strategy required for people who are, whose job it is to get data centers built. They need to figure out how to honestly relate to communities and say, well, which bits are we supposed to be useful? Rather than the approach that you come to sometimes see where people basically say, "well, we're not even gonna tell you who, it is or who the company is, but we're gonna use all your power and use all your water."

That approach's days are probably numbered, and that's not a very good strategy to use. It does make more sense to actually have a much more like neighborly approach and these are maybe new skills that need to be developed inside the industry then.

Karl Rabe: Absolutely correct. And so, you need the, you need an open collaboration approach to that, and that is, you know, mirrored, so we trying to be a bit of an example there. And if you go, if you talk about, you had a good point in there, which we usually don't have a lot of time to expand on,

but I think podcast a good format for that. You ask like, where do you get the ideas or what's the guiding star on that? And so,

I was fortunate to be an exchange worker, you know, on a farm in Canada. And they introduced to me the idea of holistic management, which is like a, basically, decision making framework, that is based on financially viable, socially viable, and economically viable.

And so those three bases are necessary in order to create sustainable decisions or holistic decisions. Those need to be short and long term viable.

And that has been, you know, my guiding star as an entrepreneur and really being able to cut out those things. You know, there's a lot of startups, especially in Germany. We had those Berlin startups who all came from a business school and all of their ideas worked on an Excel sheet,

always cutting out like a social perspective, you know? And so that was, you know, that's the opposite basically of what we are trying to do. And this framework was found by a farmer who first applied it to grass management and cattle farming, technically. But it is, and it is wildly interesting what he's able to do. He's basically retaking, stopping desertification and reversing effects in subtropical, semi arid areas.

Yeah. So we'll definitely put that in the notes.

It's a tED Talk from Alan Kettle, which I think he's still alive. He's in his, he must be 90 now. And it's fascinating. But that was a guiding star. And in order to promote our ideas, you know, a lot of our designs, you know, we put on YouTube, but we also put the files up.

The, racks, you know, you can download the CD files. There's, we believe they're created with open source tools. So especially in engineering, we only recently really have powerful open source tools for CAD, for single line diagram. So we can give the source files with that.

And that is is something how we believe that open collaboration and openness helps to build, you know, the trust

Chris Adams: Ah,

Karl Rabe: to build with speed and to really work together, you know? And that's what we get mirrored in the Open Compute Foundation. Yeah, that is something that we believe is, for challenges that we face as humanity,

you know, I believe that only this open approach, and especially an open source, open hardware, open data, framework can help us.

Chris Adams: All right. Okay, so we're coming up to time and I just wanted, and you did alluded to it a few times and I just wanted to provide a bit of space to let you talk a little bit about that before we kind of finish up. You spoke a few times about the fact that these models, when you work, bunch of designs for the racks and things are like online and available, and did you say that they're on YouTube, like people can see the videos of this or can like download like something in blender to mess around with themselves or work with it? Maybe you could just expand on that a little bit because I haven't come across that before and

Karl Rabe: Okay, sure, sure. So, yeah, we initially, when we started, you know, we designed everything and we put it, we still need to, shamefully, we still need to put, do the push for, to GitLab and GitHub. We use right now, we put those model on a construction setup, of course, called, GrabCAD.

Chris Adams: Mm-hmm.

Karl Rabe: And for our, it, you know, it's not only our own thoughts to open source this and to build the trust, but it's also our biggest, easiest marketing tool. You know, create a model, publishing it, put a video tape. We are a bit behind. We have a lot of new and great ideas and things to share.

But that's how we approach it, you know, we'll come up with idea, put it out there and, also, you know, make ourselves criticizable, you know, we'll, we are the only ones comfortably saying, look, we have the best data centers in the world, 'cause you can go, you can download, you can fact check our ideas, and if you have something against it, you know, just give us a feedback.

And we are open to change that. And this way forward, you know, helps us also to approach the biggest companies in the world. They really like this open approach, you know, and they're happy to take the files in the models and to work on that.

Chris Adams: So you basically have like models of like, this is a model of

Karl Rabe: Our rack, you know, this is our module data center. These are ideas behind that. And so that's how we are moving this forward. So people can approach this, they can download, they can see if it fits. They can make suggestions.

Chris Adams: And like see if it's tall enough for the door and all of the basic or the practical things.

Karl Rabe: Yeah. All those things, you know, and see, okay, we have smaller data center, oh, the base design doesn't fit in this setup, or we need to change something where we place, you know, the dry coolers or something like that.

And so that is really, you know, really good feedback and sparks discussions.

Chris Adams: Yeah, I haven't heard about that before. All right. Well, Karl, thank you so. This has been a lot of fun. Now, we've come up to time and I really enjoyed this tour through all the stuff hap that happens below the software stack for engineers like us, for example. If someone does wanna look at this or learn about this or maybe kind of check out any of the models themselves, if they wanted to build any of this stuff themselves, where should they look?

Like, how do we, where do people find you online or any other projects that you're looking at, you're So, working on?

Karl Rabe: So the best thing technically to, is LinkedIn. This is, you know, our strong platform, to be honest, we are very active there. We publish most there. The webpage is still under construction. You know, people already understand what we do from going to that.

LinkedIn is great. Look, go and, you know, trying to reach us and see what we do at the Open Compute Foundations is also often very great. But yeah, just technically why Google is very easy to find us on LinkedIn and to reach

Chris Adams: So Karl Rabe on LinkedIn, Wooden Data Center, there aren't that many other companies who are called Wooden Data Center. And then for any of the Open Compute Project stuff, that's the other place to look at where you're working. 'Cause you're doing the open compute modular data center stuff.

Those are the ones, yeah?

Karl Rabe: Yeah. Correct.

Chris Adams: Brilliant. Karl, thank you so much for this. This has been loads of fun and I hope that we've had listeners follow us along as well to see all the options and things available to them. Alright,

Karl Rabe: It was a pleasure. Thanks so much. And,

Chris Adams: Likewise, Karl. And, hope the wind turbines treat you well

where you're staying. All right, take care mate.

Karl Rabe: Yeah. Thank you. Bye bye. Cheers.

Chris Adams: Hey everyone, thanks for listening. Just a reminder to follow Environment Variables on Apple Podcasts, Spotify, or wherever you get your podcasts. And please do leave a rating and review if you like what we're doing. It helps other people discover the show, and of course, we'd love to have more listeners.

To find out more about the Green Software Foundation, please visit greensoftware.foundation. That's greensoftware.foundation in any browser. Thanks again, and see you in the next episode.

Jailbreakout of History

Mentioned by

Mentioned in 0 episodes

The AI-powered Podcast Player