The Future of Data Centers and Digital Sovereignty with Friederike von Franque
- ed5759
- 5 days ago
- 15 min read
Updated: 2 days ago
Can cloud infrastructure be owned and governed by the people, rather than just Big Tech? In this episode, Eleanor Drage speaks with Friederike von Franqué, policy advisor at Wikimedia Germany, about how feminist principles and decentralized models are reshaping the internet as a public commons rather than a purely corporate service.
Their conversation explores Friederike’s work on building more equitable digital infrastructure, from the environmental impact of energy-intensive data centers to alternative models like municipally owned fiber networks. She reflects on why relying solely on hyperscalers can reinforce existing power imbalances, and makes the case for greater environmental accountability and community-led design in technology systems.
By unpacking the often-invisible infrastructure behind our digital lives, the episode highlights the challenges and possibilities of creating a more sustainable, democratic, and inclusive internet.
Friederike von Franqué is a policy advisor at Wikimedia Germany, where she works on issues related to digital infrastructure, governance, and the public interest. Her work focuses on promoting decentralized, community-driven approaches to technology, drawing on feminist principles to advocate for more equitable, transparent, and sustainable digital systems.
Reading List:
Reading List:
“Wikipedia Is Running On Its Own Metal: The Power and Limits of Self-Hosted Infrastructure”by Wikimedia Europe
Data centres are reshaping the global investment landscape by UN Trade and Development (UNCTAD)
Transcript
Eleanor Drage (00:53)
Well, thank you so much for joining us. And it's a real pleasure to have you on. Can you tell us about yourself and what brings you to the topic of feminism, gender, and technology?
Friederike von Franque (01:04)
Yes, thank you so much. I am really honored. My name is Friederike von Franqué. I work as a policy advisor at Wikimedia Germany. And as you know, Wikimedia is an NGO that supports the German-speaking community dedicated to the common good. We support the people, tech, and policies that enable reliable information to be shared with the world. And therefore, I think this is deeply feminist because we are looking at the content that is not already there and that should be there.
Eleanor Drage (01:39)
So, from your perspective, then, as someone who works very closely with civil society and policy, what is good technology? Is it even possible? And how can feminist ideas help us get there? I like that you said, you know, feminism helps us fill in the blanks, put stuff out there that's not already there, and identify what's missing. Is that the secret?
Friederike von Franque (01:57)
I believe so. Plus, we have a longer time perspective. If you're looking at fast gains, then you would probably have other decisions. And if you're looking at the longer run and if you look at what is a public good, then the goal is that a society will live in a good environment, have a good life for, like, ever. So every decision should be dedicated to creating a good life, and that is feminist as well, I guess. For me, at least.
Eleanor Drage (02:28)
Hmm. It definitely is; I totally agree. You do a lot of work on Wikipedia as part of the Wikimedia Foundation. Of course, Wikipedia is a great source of data, but it's not perfect. And as I said before, when we were speaking, I'm working with this comedian and TV presenter, Sandi Toksvig, on a women's Wikipedia that is built by and for women.
Friederike von Franque (03:00)
Yeah.
Eleanor Drage (03:03)
Saying that, I use Wikipedia all the time, and I think it's an incredible bit of crowdsourced information. So, apparently, you said that recently it's had a huge uptake in data. Can you tell us about that?
Friederike von Franque (03:17)
Yes, the uptake is based on Wikidata. I won't say recently, but it is gaining traction. So you can see, we have right now over 150 million data entries in Wikidata. This is because you can share your databases, like a good library, for example, or like the European Union, which had combined two databases for its funding basis. They put in the numbers and the recipients, and Wikidata had the country and the geolocation in your language. And that was hilarious. So everybody had only to do half of it, and so it all came together.
Eleanor Drage (04:06)
What kind of data did you see on there? What new stories and content really stood out to you?
Friederike von Franque (04:12)
Yeah, what stood out is, for example, Frankfurt am Main, which had a huge Jewish community before World War II, is adding Shoah repositories and adding some, you know, some Nazi victims' databases to Wikidata. So we kind of bring back the memories of these forgotten people with their names and their fates, and this really rang a bell with me.
And what is the motivation of an institution that is adding its database to Wikidata is that they want to make the work easier and they want to find duplicates. And together it can become beautiful because you will have, for example, a writer's database, and then you have the Shoah Reminiscence database, and then you can put them together, and then you have the picture that is more colorful than that of this specific person.
Eleanor Drage (05:17)
Incredible. Thanks so much for telling us that, and I will investigate. So we've talked about the content, but you're an expert really in the behind-the-scenes stuff. So, where is Wikimedia hosted? Where can we find the compute power, the data centers? Where are they?
Friederike von Franque (05:37)
Yeah, Wikipedia is being accessed by over a million people worldwide every day. We even have noise music to signify the pings that every edit adds to Wikipedia. So we need databases that can manage workloads, and we are with hyperscalers.
We are mostly with certain American hyperscalers that are promising to be very environmentally friendly because we, of course, have this common good perspective. But still, the main load is in the United States, and we have mirror servers on different continents to have less ping time for the editors, making the work easier. Yeah, less time. If you make an edit, you want to see what you write. And if this takes too long, then people get disturbed and also change; you want to see it at once. And if you have too much latency, that can be annoying in the day-to-day work. Therefore, we have servers on different continents.
Eleanor Drage (06:33)
Less ping time, did you say?
Friederike von Franque (06:55)
Ease the workload for the service and to make it faster for the editors and the volunteers.
Eleanor Drage (07:01)
Yeah, that makes sense. You want to see your edits immediately. It must be really difficult then to kind of balance the desire to have environmentally friendly datasets with the need for commercially available spaces and data centers that are really effective and cheap. Is that an ongoing negotiation that you face at Wikimedia?
Friederike von Franque (07:24)
Yes, we have this ongoing conversation, and of course, we want to lower the environmental footprint from Wikipedia as much as possible. I think it is both; it is a legacy from the past. No one even dreamed about becoming the size we have now. So hyperscalers were convenient to upscale little by little, and they also could promise the stability that we needed. But we are looking into other opportunities.
Let's say the server management is done by the Wikimedia Foundation in San Francisco. And with Wiki, you are talking to Wikimedia Germany, which is one of the many outlets that the Wikimovement has. In Germany, we are busy with Wikidata, so that's our specialty. That's why we are so big. But we are not really meddling with where your servers are and what you do with your servers.
However, as we are a democratic movement, this perspective is part of our discussions. But what can you do? We are already at a hyperscaler that has the most promising environmental footprint. We cannot promise that they are. It's not a price question. You mentioned price. We say, "Okay, then we are willing to pay more because we have a global conscience." I don't know what we are going to do in the future.
But what can we contribute to the idea of Commons data centers? We can contribute our learnings from decentralized collaboration. And we believe in collaboration. We think that makes the world a better place. So you probably don't need to have huge data centers. They have their advantages, but you can do other things with data centers. We know about transparency, monitoring, and oversight. We think that this is necessary. Every edit in Wikipedia is being logged, and we can say that this is a cornerstone of why Wikipedia works.
If we are thinking about a Commons-oriented data center management, then I guess that transparency, monitoring, and oversight are very important. It's not control. It's just that you trust the system and the processes, and that you can see the processes and make them better because everyone can look into them, and maybe someone has an idea and says, "Why aren't you doing this and that in that way?" And then that makes it better. So that's what the Wikipedia model, let's say, can contribute to this discussion.
Eleanor Drage (10:19)
Based on what you've just said, then, a good data center seems to me to be decided on based on collaboration and participatory design, and having these ongoing conversations. But I think more tangibly, what is a good data center, and equally, what makes a bad one?
Friederike von Franque (10:40)
I think a good data center has to fulfill certain values, and that doesn't mean that you don't have to make money out of it, because if you don't make money, probably your service will not be sustainable. But how you make that money is a different question. Then you should employ the best environmentally friendly technique you can have. You should try to use recyclable materials, of course, have good working conditions, etc., etc. And again, to control this and to have an oversight about how it is done, then it would be good if you would have external people looking into it, like everyday people. I don't know about the content, though. The content is often the secret of the clients. I would wish that data center monitoring could help with getting rid of bad content. I don't think that the system will change this aspect.
Eleanor Drage (11:48)
I first got in touch with you because I went to a panel discussion you were on at re:publica in Berlin on whether data centers could be decentralized, whether they could be run by municipalities instead of by conglomerates. And this idea, I think, has been, you know, when I wrote an article in The Guardian a while ago about frugal tech, one of the things people were most interested in is how brilliant and surprising the possibility of being able to manage your own data centers locally is. So tell us about that, because I think it seems to me quite a European idea. In the UK, I think we're a bit more top-down, or we're less involved in deliberative democracy.
Friederike von Franque (12:30)
Really? Hmm.
Eleanor Drage (12:41)
We often don't understand what we vote for; we're sort of out of touch with politics and these local decisions. So maybe this is really unfair, and if you're listening, write in and complain, but yeah, tell us what you think of these data centers.
Friederike von Franque (12:56)
Well, the reality is that the state already has half of the decentralized data centers just because they were growing in them. Start with the universities, for example, they have great knowledge and performance in data centers. But also, little, like state institutions, they needed to store the data somewhere, and so they did in the next room. So the small things then became bigger and still owned, and they found out that they don't really have oversight over it.
And then you have the hyperscalers serving more or less the commercial area, targeting the public sphere to "let's move your stuff over to us, and we are much safer." And I would agree. If you don't have good management of your data center, then security is bad. The recycling and the modernization of these data centers will be bad unless you have a focus on that.
But my model was Stockholm. Stockholm builds optic fiber cables in the city, and they say, "We want to have glass fibers everywhere, even in the most faraway home that wouldn't be served by a commercial provider." And they did. And then they rented out the infrastructure to commercial vendors, and the result is that Stockholm has great and fast internet everywhere. And it's very cheap because the market really functions there. And this is not the case in so many other areas.
So I would wish for having the same system in, maybe, some data centers. I don't believe in the state doing commercial business. That's not the core issue of the state, and sometimes they are quite bad at it. But if you can manage to have the criteria that we just mentioned, please have environmentally friendly aspects, please have workers' ethics, etc., etc., and then we would like you to run our data center for high security, high performance, whatever you need, because it's the public realm. We can also afford it, and then the margin is not that important; that would be really great.
And I also believe that there will be commercial providers that would be willing to provide that service. Plus, we already have that in Berlin. There's a small data center provider that's run by a community. We have a new initiative, also from startups and providers, that want to have such a Commons-oriented data center service. And I believe that there can be much, much more. And if the state would provide, as the first client, like seed funding, that would be great.
And then if you have a community where you have dialogue, the client, the state as a client, can say, "This is not really working. We would rather like, you know, having this and that. And don't you think this would be..." so you can have a conversation with your service provider as a client. And as a client, you would also have a say in what you get. And then there would be external monitoring. So I think this whole package would be great. And hyperscalers can still exist outside or next to and parallel to this model.
Eleanor Drage (16:37)
Well, I'm sold. This is how you make it happen. Why didn't it turn out like this in Frankfurt, where you said there was a data center boom and loads of data centers were constructed with very little decentralized community governance that you like, you've just described? What went wrong?
Friederike von Franque (16:58)
Well, also a development issue. I mean, who can have oversight? This is the local community that can have monitoring of that. And they didn't. All of a sudden, within eight years, much, much more data centers have been built, plus all data centers have been made bigger.
So I learned about this because citizens complained about the huge blocks that were growing in the backyard. And so the municipality said, "Yeah, but these are, in general, outfits; we allow them to be built in the backyard in this area. This is the zone that we are allowed this for." So they came up with another idea of, okay, data centers are a special commercial service and we have a special law, and the data center community complained. So why? Et cetera, et cetera.
So this all takes time; it has to be seen. Plus, why should we have oversight of a commercial service? These are all new ideas, and that's what happened. But in Frankfurt, it's a small city; it's not even a million citizens. And in this city, the European internet is located, and tons of data centers, with, I don't know, how many terabytes we are serving. So Frankfurt is a good example of how things can go wrong, probably, or how things can develop, or that there is even an end to growth because there is no electricity left, and not enough water left, and not enough space. Where there may be some more space, but we still, I mean, the city still wants to have car repair shops, for example, and they also need space.
Eleanor Drage (18:51)
And may this be a lesson to, you know, the UK, which has grand plans under Keir Starmer to build way more data centers. We have 500 data centers in the UK currently, and there are 5,000 in the US, but you know, a data center can get lost in the US in the outback in a way that it can't in the UK. So, of the images or analogies you've used to describe data centers, this glut, this internet knot that you describe in Frankfurt, is it gambling? That data centers should abide by stipends, by stipulations, in order not to proliferate in the same way they have in Frankfurt. Can you tell us about that analogy? What do data centers have in common with gambling?
Friederike von Franque (19:41)
With gambling?
Eleanor Drage (19:43)
This is what you said last time; there was a gambling analogy, too, or maybe to casinos, or I don't know. They should abide by the stipulations.
Friederike von Franque (19:54)
I think I remember. Well, if you look at the UK has a good starting point because you have the sea and you have green energy. You can put data centers in a windmill, for example. London should be careful with having too many data centers, and cities that are developing new housing should integrate data centers because they can use the warmth that they would need.
Frankfurt, the gamble is that more AI will be used worldwide and that more data centers and data capacity will be needed. Let's say statistics are pointing in that direction. But I'm not so sure. If the AI boom is dwindling, then you would need less data center capacity. And by coincidence, the AI companies are also selling data centers and capacity and making lots of money out of the data centers, less so with their AI models. So how come?
So I have the feeling that we have a combination and a very good PR effort, also, that somebody who wants to sell us GenAI everywhere is looking at the data center money that they are making out of it. So that has exaggerated my theory on it. And I would be happy if the GenAI hype were less and systemic and symbolic AI hype were more substantial, and if people were looking into what they really need. And then probably they would need less data center power, but still, they would need data center power. So I think that the UK will see lots more data centers, but it's not that the UK will lose out in the future because it does not have enough data centers. I don't believe in that.
Eleanor Drage (21:59)
And symbolic AI, just so that people know the difference between that and generative AI. Can you give us some examples?
Friederike von Franque (22:06)
Yes, we have Wikidata, for example, and if you say "orange," this could mean the color, the company, or something else, you are not sure. So in a database, you would have a number that specifies what you mean. So if you just point to the number, then the machine will know the color is meant. And then you will have, like, a combination between "orange equals color equals whatever you need." So if you ask them questions, you would always get the same result. If you ask questions like "what colors in the world are there," then your data set will give you these examples.
If you have GenAI, then the mechanism of GenAI is: "I read every book I can get my hands on, or my machine eyes on, and find words or the combination of letters equaling orange." And then I give you every result with "orange" I have, which also could be a result with a company, based on the probabilities of what your question means. And because the probability changes through time, you will never get the same results, but you will get a plausible result, probably out of it.
So the best GenAI is working with the most data because then your statistics become better. But you need to update your data all the time. And this is like, you know, a huge shotgun that is trying to scramble in everything, and I believe that more, let's say, more geological aspects of symbolic AI will serve the same question but way better, because you always get the same results, which is, in certain circumstances, the thing you need in most circumstances. And in some circumstances, you need GenAI because you can get a better speech from it, for example.
Eleanor Drage (24:14)
Thank you. That was a really thorough, brilliant description, much better than I could do. What can other countries like the UK or, you know, those that are thinking of ploughing their way into building more data centers learn from Frankfurt? And, you know, what would you say to Keir Starmer to encourage him to get it right when, you know, mapping out how to build these huge, energy-intensive...
Friederike von Franque (24:36)
Yeah.
Eleanor Drage (24:43)
...blocks of concrete?
Friederike von Franque (24:44)
Yeah, that's what you said. You should make a map, a whole-of-Britain map, and find out what data centers already exist and where hyperscalers or where data centers want to be built.
Then you should ask them to locate their business headquarters in your country so that you get the money from it, at least so that they pay taxes. You should be aware that there will not be much staff. I think the DE-CIX is working with 20 people. So it's not really a lot of staff. You should maybe combine it with recycling companies, which can do the recycling of all the server racks, et cetera. And you should be aware that there will be loads of trucks coming in and out relatively often.
You should ask for the emergency engines that are getting tested once every month, normally with fossil fuels, that they be replaced by something else. And you should try to, as I said, try to locate them in areas where a little warmth is needed or a little cold is needed because you can change that energy to cold as well; for example, for vegetable markets that would need a cold storage. And also for a little warmth; that's equally possible. And you should ask them to finance that infrastructure that is needed to transport the warmth or the cold to different places because you're all profiting from it.
Yeah, and that is basically it. Probably, I mean, for your citizens, you need greenery outside and you need maybe also a dialogue. It's very annoying if people have to call in, I don't know,Texas, and then a machine is answering you, "Please call again in half an hour," and then you get shit answers. So that will be also quite important, I guess.
Eleanor Drage (26:55)
That's terrific. Thank you so much, Friederike, for coming on and giving us all this amazing information that people don't yet have access to. Much appreciated.
Friederike von Franque (27:04)
Very welcome.
