When data centers move in II

Send a voice message to LivingPlanet

Subscribe to the podcast here via your favorite podcast app: Living Planet Podlink

TRANSCRIPT

Charli: So, I’ve arrived in Hattersheim for the second time in two weeks. It's a very sunny day and I’m going to go inside one of these large black buildings today. I’m just trying to find the Haupteingang...  

After weeks of false starts, here I was, finally, on the other side of the high security fence.  

Charli: Ok, here we are. 

Ready to go inside the cloud. 

Günter: Hello, I'm Günter Eggers... Charli: Hi, lovely to meet you, how you going... Günter: Great, thank you... So, you could leave your stuff, in this room here...  

And so, the tour begins. 

Is this how it works?

I’m taken through metal turnstiles. Down concrete corridors, past warning signs. Through steel door after steel door. Right into the bowels of this giant building. 

This room is very loud!  

And it makes me think... if this were a cloud, it would have to be a cumulonimbus. The densest, most impermeable type of cloud. The dark, towering kind that rumbles and vibrates with electrical activity. 

But really, it’s not like a cloud at all. It’s cables, concrete, machinery, fans, pipes, and fluorescent lights.  

This is, oddly enough, where the internet lives. 

Every email you send, show you stream, TikTok you watch, and, increasingly, every prompt you send AI – somewhere along the way, they all pass through a place like this.  A data center. An engineering marvel. That til recently, we haven’t really noticed. 

I’m here because... we’re building more of these places than ever before. And that makes understanding what is inside them and what they ask of the world outside more important than ever. 

Series jingle 

Jensen Huang: This is the beginning of a new industrial revolution. And this industrial revolution is about the production – not of energy, not of food – but the production of intelligence. 

Alex Karp: We're in an arms race here... We're running. We're running like hell. 

Sam Altman: Um, it does feel like it's going very fast. 

Ursula von der Leyen: We will spare no effort to make Europe an AI continent... This is the great mission of our time. 

This is the future, we're told. Of economic growth. Of innovation. And we'd better not get left behind. 

But this intelligent digital universe we’re building doesn’t live in the clouds. It’s planted firmly on the ground. In data centers. 

I'm Charli Shield. And in this two-part series from Living Planet, we're digging into the data center boom on the ground from Europe's biggest data center hub – Germany. This is part two. If you haven't heard part one yet, I do recommend going back and giving it a listen – you can find it in our podcast feed. 

Günter: So here you can see these different buildings. We have started construction here in 2017... 

We’re inside a data center campus, in Hattersheim, a small town near Frankfurt in central Germany. The company behind this center is NTT Data. It’s a Japanese multinational telecommunications firm, with about 160 data centers worldwide. Around 25 of those are in Germany. 

Günter: So, my introduction. My name is Günter Eggers, more than 13 years with NTT global data centers.  

Günter Eggers is NTT Data Germany’s spokesperson.  

Günter: And I'm currently responsible for the political and public communication. So I'm the spokesperson, but even the main political contact. 

Günter is showing me around today. In Part One, we skirted the perimeter of this place. Speculating from the outside in. From the other side of the fence, I’d seen five buildings. In fact, there were seven.  

Günter: It depends on the point of view... what we have here with our seven data centers is a data center campus. 

Seven buildings, one campus. 

This particular mass of cloud is about 28,500 meters squared. Roughly the size of 5 American football fields. 

Günter: And the overall IT capacity is 70 megawatts... you could say it’s substantial. 

10 years ago, around the time he started out in this business,  

70MW was considered a very large data center, Günter tells me. But not anymore. 

Günter: The new data centers we are talking about have a capacity of 30, 50, in some cases, 100-megawatt IT capacity.  

That’s per building. 

Günter: So that's a lot more. So from that perspective, from today's perspective, it's a smaller, these are smaller data centers, or let's say mid-size.  

We set off. Günter and the technical team leader guide me on what feels like a well-trodden data center tourist trail... 

Günter: I would like to follow the path of the energy...  

There’s actually only one building we can enter, which houses the backup power supply, some cooling systems, and one set of servers. 

The other buildings – where the bulk of the servers live – are off limits. Even, apparently, to company staff. Beyond the threshold of the doors to those buildings, Günter explains, is strictly the domain of their customers, who bring their own IT equipment to run there. 

That’s because this is a ‘co-location’ style data center. Meaning, basically, it’s rented out to a whole bunch of different customers. Customers who shall not be named. That would be a security breach. Mostly, he says, they’re cloud providers.  

Günter: We know of course who our customers are. We are not allowed to talk about the customers but we know of course necessarily. So, we can anticipate what kind of digital services they are producing here in our data center. But we don't know exactly... 

'Co-location' is one style of data center. It’s different to what are called ‘hyperscale’ data centers. Those are the biggest kind. They’re typically owned and operated by a single company, like Google, Microsoft, or xAI.  

These are the type being built to train generative AI models – like ChatGPT, Gemini and Claude. And, as Big Tech races to build ever more ‘intelligent’ models that demand ever more compute, these hyperscale facilities are getting bigger and bigger. 

News clip 

One of the world’s largest examples belongs to xAI, Elon Musk’s company. It built two campuses in Memphis and Mississippi called Colossus I and Colossus II to train the AI model ‘Grok’. And, according to Musk, they have a combined capacity of 1.5 GW – with plans to expand that to 2 GW. 

‘Co-location’ centers tend to be smaller and often built in areas closer to people, so their users have faster data transmission.  

They don’t just provide empty warehouses for their customers though. In fact, what they’re really providing is power, cooling, electrical infrastructure, and security.  

Günter: We have an SLA promise to our customers that we are able to deliver the energy, the power... 

SLA btw is service level agreement. In other words, a binding contractual promise. 

Günter: ... for every single second a year. 

Every single second of the year. And not a single second less. No hiccups, no dips allowed. Like the internet, data centers never sleep.   

Günter: So, meaning, we need to prevent any major outage...  

So, to do that... to unfailingly provide power 24/7, 365 days a year, data centers need to have a constant, highly secure energy source. 

Most data centers in Germany, including this one, are plugged into the grid.  

In fact, there are so many of them in the Frankfurt region now, they’re estimated to consume about 41% of the city’s electricity. In Dublin, Ireland – that figure is 80%.  

Frankfurt’s grid is so chokkas with data centers now, there are no new grid connections available until at least the 2030s. We'll come back to that point later, because it’s important.  

In the event of grid failure, data centers need to have back-ups to make sure power keeps flowing. For most centers, like this one, that is diesel generators, so a polluting fossil fuel, sometimes coupled with batteries. 

Günter: These batteries are only capable for keeping the complete power distribution for the IT system for five to ten minutes, maybe 15... 

Testing these diesel generators, which happens here every two months, is often  what contributes to the noise pollution created by data centers. Because diesel generators are very noisy, especially the big ones. But on the day I visited the center, apart from the various whirring, screeching and droning I heard in each of the rooms we entered, I didn’t think the center was particularly loud from the outside.  

But to come back to the grid. Most data centers in Germany, like any energy consumers, are as dirty or as clean as the grid itself. In 2025, renewable energy covered about 55% of Germany’s gross electricity consumption. The real problem for the grid now is the scale and speed at which data centers are being built.  

News clip 

According to the International Energy Agency, global data center electricity consumption is set to more than double by 2030 - to around 945 terawatt hours. So, consuming more power than the country of Japan by the end of the decade. 

We'll be back after this short break. But, before that, just a reminder - if you want to reach us directly, you can email us at livingplanet@dw.com or leave us a voice-note by going to dw.com/livingplanet and hitting record on that big red button. You can only talk for a minute... so make it count. Ok, back soon. 

Trailer 

And we’re back. 

So, NTT's facility isn’t one of the giant AI data centers training the world’s biggest generative AI models and grabbing headlines around the world. But Germany wants a lot more infrastructure like this. And for companies like Günter’s, business is booming. 

Earlier this year, the government unveiled its latest “data center strategy”. There are currently more than 2000 data centers in Germany. By 2030, the country wants to double data center capacity and quadruple its computing capacity for high-performance computing and AI. 

Clip Katharina Reiche 

The industry argues that Germany has little choice here. It still lags far behind the US and China when it comes to data center capacity, especially relative to the size of its economy. Building more, they believe, is also important for Europe’s ‘digital sovereignty’ - which is a term thrown around a lot that basically means making sure the continent isn’t wholly dependent on US American tech companies and infrastructure. 

But others are concerned. About where all the extra electricity will come from, and how the country will reasonably provide this power while also meeting its climate targets and rapidly cutting emissions from the power sector. 

Jens Gröger: What actually happens right now is that due to the extreme electricity consumption of data centers, this electricity... it can't be produced by renewables only because it's extra consumption. And so, what happens is that existing fossil fueled power plants are running more intensively. 

This is Jens.  

Jens: My name is Jens Gröger. I'm working at the German Öko-Institut, which is located in Berlin. And my function there is to be a research coordinator for sustainable digital infrastructures. 

Jens spends a lot of time digging into the data on data centers in Germany, and trying to square that with the country’s sustainability ambitions – as in, those goals to cut the emissions heating the planet and making it increasingly unlivable. 

Jens: Electricity has a carbon footprint for sure, especially if you produce it with fossil fuels. So there is also a rise of carbon emissions due to data centers. 

The rise in emissions due to data centers that Jens is talking about is basically happening because there are suddenly a lot more of them, demanding a lot more from the electricity grid. And even though renewable power generation is increasing, if it doesn’t expand at least as quickly as this new demand, fossil-fuel powered electricity has to fill the gap. And that means more planet-heating emissions. Just as we’re trying to desperately dial them down. 

An analysis from Bloomberg New Energy Finance estimates that the additional fossil fuel power generation required by the data center boom will increase total global power sector emissions by 10% over the coming decade. 

What if we were able to power all this new infrastructure with clean energy? 

Even if enough new wind and solar farms are built to power all these new data centers, that is renewable electricity that can’t be used elsewhere. Not to electrify more transport, for example, or to help other critical industries move away from fossil fuels.  

Basically, at a time of transition, when almost every sector of the economy is trying to electrify, they’re now competing with data centers for the same clean electricity. 

This then begs the question: if our goal is to lower emissions and tackle climate change, can extra data centers actually help achieve this in any way? 

According to Günter, data centers are not just critical infrastructure for our daily lives, but also for a more sustainable future. 

Günter: We need data centers to support our economy, even to support the growth of the sustainability industry. So, there are a lot of digital services necessary to optimize the logistics to optimize energy systems to optimize kind of resource usage. Yes, you need even these big data centers especially if you want to use AI capacities. That's a trade off. 

There is an argument that we need AI to optimize things like transport, energy systems, and industry to help these sectors cut emissions. 

Jens: But the funny thing is that most of these AI appliances that really help the industry... they don't really need these chat GPT like really big AI models.  

Research shows that the kind of AI needed for this optimization is of the older, more traditional kind. For example, machine learning models that forecast wind patterns to improve wind energy generation. 

Jens: And maybe this AI model can even run on a laptop in the factory itself... 

And older AI tools like this have a much lower environmental footprint than generative AI, which is what‘s behind the rapid growth in AI infrastructure. 

Jens: If I want to, let's say, improve my heating system, I don't have to ask a general purpose AI model that knows from Shakespeare until whatever how to improve the heating system because it's only five or 10 parameters that are necessary to really improve that. 

Also, as a recent analysis from analyst and writer Ketan Joshi, that was commissioned by environmental groups, found, even if these climate benefits are genuine, they are unrelated to and outweighed by the increase in energy use from the generative AI industry. 

Another concern is the potential for this boom to be used as a delay tactic for the fossil fuel industry. 

Some environmental groups have found a link between the expansion of fossil fuel use and the growth of data centers in Germany. 

A recent report from Urgewald, a German environmental and human rights organization, found that about 12% of the gas-fired power plant capacity currently planned in Germany is linked to supplying data centers with power. 

Earlier this year, German gas power plant manufacturer, Siemens, in cooperation with the Irish power management company Eaton, announced they are now offering a “ready-to-use" energy package for data centers, which comes with a standard 500-MW gas turbine. 

This echoes a broader trend. Across the Atlantic in the US, where data center growth has been far faster; some companies have begun building their own gas-fired power generation to fuel their data centers.  

News clip  

Jens: Yeah, in situations where data centers can't be powered or can't get their electricity from the grid, the idea comes up to really build your own electricity generation. This is what already happens in the US. For example, the big AI data center, GroK for XAI, they have their own gas fossil-fueled power plants there, feeding this data center with electricity. And so they don't need any electricity line there, only a gas line. And they are really emitting huge amounts of carbon emissions due to burning this gas there. 

Jens is talking about Elon Musk’s xAI’s Colossus data center campuses in Memphis and Mississippi, where the company is currently operating dozens of self-installed natural gas turbines to power its AI infrastructure. Local environmental groups have filed a lawsuit, alleging that the turbines are operating without the permits required under the Clean Air Act. 

News clip 

Whether that could happen in Germany, though, where planning and environmental laws are very different, seems, to Jens, unlikely. 

Jens: In Germany it's quite difficult to get the allowance for having these heavy emissions there. So, this probably does not happen so easily in Germany as it does in the US. 

There is also the question of how much water data centers use. For many regions around the world, especially those that struggle with drought and water scarcity, this is becoming a serious problem.  

Jens: We see data centers in Spain, for example, or in Portugal or in South America, but also in USA, where there is really water scarcity, but still the data centers drain the groundwater to have their evaporative cooling. 

How much water a data center uses and where they get it from greatly depends on the type of center and where it’s located.  

The biggest water consumers are the hyperscale AI training facilities. These places need to cool tens of thousands, even hundreds of thousands, of computer chips, which get very hot and sweaty with all those silly questions you ask chatbots. Just kidding, but, for example, an NVIDIA H100 chip runs at an average temperature of 87 degrees Celsius, or 188 degrees Fahrenheit. So, pretty hot. 

Centers can cool servers in a number of ways – direct-chip liquid cooling or air cooling, for example. 

News clip 

A recent study from the UN University found that, by 2030, AI-related water consumption could add up to the basic annual water needs of 1.3 billion people. 

Reporting has also shown that a majority of AI data centers going up in the United States are being built in drought-prone areas. 

Günter: So, the first phase of cooling, so to say, is if the cold air cools the CPUs, the GPUs, the AI processors, all that stuff...  

We exit the building on the top floor and climb some steel steps to the roof. Where the air conditioners that cool the building live. NTT data say they use “state of the art” closed-loop technology to try and save as much water as possible. 

This particular data center campus, however, like most data center companies, does not publicly publish its electricity or water consumption. Though Günter told me they do report it to the EU, as required by law. The company says that would be breaching the privacy of their customers, and bad for business. 

Günter: Because to be honest, that impacts investors, impacts customers. And from that perspective... the data center industry see it as a threat for doing your business.  

As we exit one of the cooling rooms, I ask Günter a less technical question. 

Charli: When people complain that they are enormous, ugly eyesores, what's your response to that? 

Günter: Well, being ugly or not, or seeing something as ugly or not is, of course, in the eye of a viewer. Definitely, they are right if they say they're huge buildings. Yeah, you cannot deny it.  

Sometimes they fit well in a certain environment, other times not, he tells me. 

I think about all the other enormous infrastructure we build to run our societies: the factories, manufacturing plants, freight terminals, power stations. Beauty, it's true, is in the eye of the beholder, but I think it’s safe to say many people would consider these... not that beautiful. They are also, often, relegated to the outskirts of cities and towns. As the scale and number of data centers grows, their proximity to homes and communities is becoming an increasing challenge. 

Günter: In some cases, I wish the responsible people in the community, so the so-called city planner, would have thought more about the potential reaction of having this huge... data centers close to the living areas.  

It's kind of hard to hear me here, but I'm asking: So, that's not your responsibility? 

Günter: So, we are asking...if you find a location close to Frankfurt with enough power, with enough space at all, then we ask the community, local community if we are allowed to build a data center. And if the community says yes, we originally intended to have a kind of market there or logistic center, but data center would be even nice, then we start to build because, if it's allowed to build data center in a specific spot is the responsibility of the community itself. 

This brings us back to the company’s other data center project, located a few miles away in Hattersheim. The one that’s currently being challenged in court by a citizens’ group opposed to its construction. We heard from members of that group in part one. I ask Günter about it.  

Günter: Unfortunately, we have some issues there and honestly, I wish we would have communicated better in advance. 

I inquire if this is one instance where he wishes ‘city planners’ would have more carefully considered what people living nearby might think. Günter doesn’t want to blame anyone in particular.  

Günter: Well, let's say... In some cases, it's easier. In some cases, it's harder to build new data centers and to find acceptance in the neighborhood, and that's one of the harder cases. 

Günter: This data center, as far as I'm aware of, is widely accepted in the community. There are some people which are really fighting against this data center, and I have to accept the effort they are doing. But on the other hand... we have all the necessary approvals, and we already have invested a lot of money. Of course, we are continuing still to find a way for having a peaceful relation even to these people which are not accepting these data centers. 

The group, who don’t represent the views of the entire town, say they aren’t anti-data centers. But they believe it’s wrong to construct one in this particular, formerly natural area, close to where people live, in a small town that already has nine data centers. 

Günter: Hopefully the time will show that we still respect our neighbors and that we find a solution to harmonizing all the stress... And we are trying to engage in the community. We are trying to give back something to the community, engaging in schools, sports clubs, stuff like that. Being visible here, paying our tax here, by the way. 

Charli: You know what goes on here and what your purpose is; do you think you'd like to live next to a data center?  

Günter: That's a hard question. I cannot imagine to move out of my house. I'm living in for more than 20 years now. And as I started living there, there were just a few other houses in the neighborhood... Would I move in the direct neighborhood of a data center? Probably not. But, well, that's a theoretical question. 

I wanted to ask some more people for whom it's not a hypothetical question. 

Charli: Okay, so I'm just walking through the residential area... 

As I exit the center, I walk over to the residential blocks that sidle right up against the campus. They’re modern town house-type buildings, stark white walls, with neat little lawns and garages out front. They were built after the arrival of the data center.  

Charli: Hi! Entschuldigung, wohnen Sie hier?

I spot some people walking down the street, and I ask them if they live here.  

Charli: Hi!  

Charli: So, whereabouts do you live? 

Lise: In this building over here.  

Lise, she tells me her name, points to an apartment block about 100 meters down the road.  

Charli: How long for? 

Lise: For about a year now, a little less.  

Charli: And how do you find it? 

Lise: We love it actually because we're profiting from the center. We get the... heat.  

Charli:The waste heat? 

Lise: Yeah, the waste heat, right? So we don't know yet how much the bill will be, but I guess it it's still sustainable that it's used. 

Lise is talking about the heat that the data centers produce that can be captured using heat pumps and piped into local district heating grids. In conversations with NTT Data, they told me that new, modern heating grids are required to connect houses to this heat, which is warm, by the way, not hot, around 25-30 degrees Celsius or 77-86 Fahrenheit. The company pays for the infrastructure on their turf, while the town takes on the rest.  

Lise and Kai haven’t been connected yet, and it’s not the sole reason they chose to live there, but it definitely sweetened the deal. 

Charli: So, you will expect to pay less for energy? 

Lise: Yes, that’s what they promised us.  

Charli: And you... obviously you moved in after the data center was here already?  

Lise: Yes.  

Okay. And are you ever bothered by it? 

Lise: No. We were told that there's some sound from the fan on top of it, but actually we have never really noticed. 

Charli: What do you think about its appearance? 

Lise: Not really nice (laughs) but I mean we picked this building because there will be a lot of buildings around so we don't really have a view on it.  

If Germany follows through on its plans to rapidly expand data center and AI infrastructure, a lot more people are likely to find themselves faced with the prospect of either moving in next to a data center campus, or having one move in next to them. 

These centers are arriving at the same time Germany is trying to transform its energy system, urgently cut emissions and adapt to a hotter world. And that means, the questions raised by this story, about where data centers belong, what they demand and whether the tradeoffs are worth it, are not going away. 

Questions that Indra Jungblut, the tech journalist we heard from in part one, believes are not being asked in a public setting nearly enough. 

Indra: Especially if you connect this idea or this thought with the space we need for energy transition, and the discussions about where to build wind turbines and where to build solar power. Well, this, there it's like discussed a lot where to put all these new renewable energy sources. Regarding data centers, it's not discussed that much, but it's like a big problem. We're still building, building, building data centers. And our problem is just growing bigger. And it's already big. 

This two-part series was created by me, Charli Shield, with Neil King. This episode was edited and fact checked by Jennifer Collins. It was written and sound designed by me. and engineered by Jürgen Kuhn.  

I'd also like to thank Julian Bothe, Moritz Leiner, and Konstantin Hartmann for being generous with their time and sharing their insights with me for this story. 

If you rate this Living Planet episode, you’d make us so happy by leaving a rating and review on Apple Podcasts or Spotify or YouTube. If you want to reach us directly, email us at livingplanet@dw.com or leave us a voice-note by going to dw.com/livingplanet and hitting record on that big red button. You can only talk for a minute, so make it count. As always, thank you so much for listening.