The Battle Over Data Centers with Tara Merk

How can communities take back control of the digital infrastructure that powers everyday life? In this episode, Eleanor Drage speaks with Tara Merk about how community-owned data centers could transform digital ownership and challenge the dominance of Big Tech.

The conversation explores alternative models of internet infrastructure that prioritize local empowerment, sustainability, and cooperative governance over corporate control. Drawing on examples from Germany’s renewable energy sector and community-led initiatives, Merk reflects on how decentralized ownership models can create fairer and more environmentally responsible technological systems. The episode also examines what it would mean to design digital infrastructure around collective needs rather than profit, reimagining the internet as a shared public resource.

Tara Merk is a researcher and advocate focused on digital infrastructure, sustainability, and community-led technology systems. Her work explores how cooperative and locally governed models can reshape the ownership and operation of data infrastructure, with particular attention to environmental responsibility, democratic governance, and digital equity.

Reading list:

• Tara Merk’s LinkedIn

• Community Data Centers  

• The making of critical data center studies 

• Data infrastructure studies on an unequal planet  

• Getting into fights with data centers: Or, a modest proposal for reframing the climate politics of ICT

• Data Center Opposition Report 

Transcript:

Eleanor Drage (00:53)

I'm really pleased to be speaking with you today. This is a really exciting episode for me because we've already done part one of this episode, thinking about the democratization of data centers. Today I'm with Tara Merk, who's a researcher and PhD candidate in Political Science at CNRS/Panthéon-Assas University Paris II, studying exit-to-community strategies for collective digital economy ownership. You will tell us all about what that is over the course of this episode. So just to begin, Tara, who are you? What do you do? And what brings you to the subject of good technology?

Tara Merk (01:34)

Yes. Hi, Eleanor. I'm really grateful to have the opportunity to be here. I'm Tara. I'm actually a postdoc by now at the CNRS in Paris. I'm also a research associate at the Weizenbaum Institute in Berlin and a research fellow at the New School in New York. What brings me to the topic of good technology is that I've always loved the internet, but there are so many things that I can't change or determine that sort of get handed to me. For me, it was always the idea: how can I tinker in mainstream platforms and mainstream spaces in a way that reflects my values and the values of the people that I hang out with? I am always asking how we can make technology serve us, with myself in mind as well as other people, and trying to figure out ways to make that happen.

Eleanor Drage (02:32)

Fantastic, and congratulations, obviously, on your PhD. These old university websites often keep old information. We are the Good Robot podcast, or I am now, because Kerry's gone, very sadly. So, what is good technology to you? Is it possible? How can feminist or pro-justice ways of thinking help us get to good technology?

Tara Merk (03:00)

I think, relatively simply put, what makes technology good is technology that is designed for the people it serves, putting them first and being accountable to their needs. There should be a feedback mechanism through which those people can have a say in how those technologies work: the way data gets collected, the way certain features are implemented, and how accessible it is to that user base. It is about working for the people it serves and being accountable to those people as well.

Eleanor Drage (03:39)

It sounds so straightforward, and yet somehow it's just not happening. Before we talk about how to democratize data centers, it's important to ask the key question: What is a data center?

Tara Merk (03:52)

I think that's an interesting question because data centers have been in the news quite a bit recently. They're usually portrayed as these huge industrial concrete blocks somewhere outside of big cities that have thousands of servers and blinking lights, looking very "Matrix" or futuristic. Actually, there's no one definition for what a data center can be or should look like. A data center, in essence, is just a building or a room housing some sort of IT infrastructure. This could be servers hosting websites, storage systems for certain data, or networking equipment. It's not strictly defined how big a data center needs to be, what needs to be in it, or what that IT infrastructure does specifically. At its core, it is a building or physical space with IT infrastructure, but everything else is pretty much up to the specific case we're talking about.

Eleanor Drage (04:54)

Okay, so it could be completely different. Let's keep that in mind for the rest of the episode. What is the data center market like at the moment, and how has it become so homogeneous? When we think about a data center, it's become this concrete flat building, one story just above the ground. If you look for them on Google Earth, which you can do with Anne Pasek's incredible tool that helps people look for the data centers supporting the websites they frequent, whether they're in Arizona or in Dublin, they all look pretty similar. They're not signposted; they're gated off and designed to be invisible so that we don't notice them. And yet, they have a huge effect on local infrastructure. Again, why is the data center market so homogeneous if a data center doesn't have to be this single thing?

Tara Merk (05:51)

I think that's a really interesting question. There are two parts to this that I want to explore. The first is that, in reality, if we look at the data center market to include smaller boutique operations, they do not have to be big concrete slabs. I started Googling and found blogs showing data centers in churches, on swimming rafts, in toilets, underwater, and in mountains. I think those could be mentioned by the news a little more often, but there are data centers everywhere, and they can look very different.

The version that gets picked up in the news most frequently is the very big concrete slab with thousands of machines. One reason for this is the way we measure data centers. Usually, we don't count the number of individual data centers in a country; we look at capacity. Capacity is usually measured by how much energy can go into it at a given moment. You might say, "I have a 100-megawatt data center" or a "10-megawatt data center." The energy input is important because it determines, to an extent, how much computing capacity or how many servers can be hosted there.

As we talk more about AI systems and their computing requirements, this size really matters. The AI systems we're using today require very high computing capacity. Consequently, everyone is on the lookout for data centers that can handle a throughput of hundreds of megawatt-hours every day to train and operate these models. There's a focus on who has the biggest one. Those that use the most energy are prominent in the news, both as a challenge, due to environmental impacts, and as an "ideal" model of what a data center should be.

Currently, the US leads the market, followed by China and the European Union. However, with the growing demand for AI, we're seeing a data center boom worldwide, including in the Global South. This is happening because land or real estate is cheaper, energy might be cheaper, or governments are willing to subsidize these projects in anticipation of job growth or other economic impacts that usually don't materialize.

According to McKinsey stats, 30 to 40% of the market is controlled by cloud service providers like Google Cloud, Amazon Web Services, and Microsoft. Another 20% is controlled by smaller cloud providers or AI companies like OpenAI and Anthropic. This means more than half of the world's data centers are controlled by these entities. The remainder consists of co-location facilities, where one company owns the building and provides electricity and security while renting out space to other companies for their servers. These co-location providers are usually owned by real estate investment trusts, large funds that are publicly traded and predominantly accountable to shareholder interests.

A smaller remaining part, around 10%, consists of independently operated data centers, such as those for universities, museums, or small businesses. One interesting thing, and a little scary, is that with the AI boom, market share for big providers like AWS and Google Cloud is predicted to grow to more than 60% by 2030. That dominance gives those companies a lot of power, not just over compute capacity, but power to decide how the environment is impacted, who can access computing resources, and how AI systems are going to be governed and developed. The important trend is that the market is already homogenized to an extent. While there is diversity, the trend toward further homogenization is dangerous, and it would be beneficial to think about alternatives to that model.

Eleanor Drage (11:58)

Yeah, absolutely. It's the form of Big Tech dominance that people aren't talking about, for sure. I'm so glad that you told us about these different forms of data centers that exist and that recognize that hyperscalers are not desirable for everyone. Not everybody wants or needs that much compute capacity. Indeed, this is why we need different forms of AI that aren't so energy-consumptive and that actually respond to more localized demands for energy. We'll be talking to a number of different people on this podcast about low-resource technologies that respond to these kinds of environmental pressures. What could a data center be otherwise? What are the alternatives out there? What is a community-based data center, for example, that I heard you speak so eloquently about in Berlin last year and got everyone in the crowd excited? What is that?

Tara Merk (12:58)

That has been a question sitting at the heart of some of the research I've been working on recently, both with the New School and with some colleagues at the Weizenbaum Institute. We’ve been asking ourselves: what do community-owned and governed data centers look like? One of the things we figured out in speaking to each other is that while I had a very specific idea in mind, which I'm not going to promote here, I realized that a lot of people have ideas, but they're all completely different. Just as wide as the term "data center" is, the term "community-owned and governed data center" might be a bit of a narrowing down, but it's still pretty broad.

Seeing that we all had these different visions, we decided to host a series of workshops, which we've been doing, to get other people involved. This includes researchers, academics, activists, artists, and civic voices. We're also trying to speak to political leaders and people in business who see themselves as entrepreneurs in that space and ask them: What does a community-owned data center look like to you? We’ve been running these workshops as collaborative experiences, so it’s not any single person's vision, but the idea of a group. We've been getting these really cool drawings out of them, which are difficult to show on a podcast, but there are some interesting preliminary high-level themes.

There are a lot of different ideas, but some common themes are that people think these data centers could be more transparent. One group actually came up with a data center built from glass. A lot of people talked about initiatives to bring student groups into having a school class set up their own server. The idea is that the data center isn't just a service provider, but also an educational institute or infrastructure within cities or even rural communities that people can visit. During these workshops, many people noted that they don't think about data centers much in their day-to-day life because they are so closed off and hard to get into. Just providing that accessibility could create a community aspect around the data center.

Apart from that, we're seeing people think about cooperative models or federated networks of data centers. How can a network of smaller facilities in different places share compute capacity? Using the cooperative model, people who live close to a data center could become cooperative members and shareholders, thereby getting a say over strategic decisions for that facility.

In these workshops, people usually run into the issue of financing because high-capacity data centers are not cheap; they're very resource-intensive. Often, people suggest getting the public sector involved. We've been doing these workshops predominantly in Germany or a European context, where there is a lot of talk about sovereign AI, sovereign clouds, and the "Euro-stack." The idea is to take that public discourse and materialize it in these sorts of projects by involving the public sector.

Lastly, there's the aspect of what services the data center provides. Is it mining Bitcoin, training new AI models, or hosting them? Many have mentioned that if a data center is in your vicinity, it should host services for the people who use it. For example, what if everyone in a certain zip code got access to an email or a domain hosted in that data center that they could configure to their needs? Or what if a municipality could use those facilities to host its own public services?

There is also the question of how it gets powered. People think about where the energy comes from, and while we'll probably still need to rely on traditional resources, I've heard opinions like: "What if we just switch it off?" If it's completely solar-powered, why not switch it off when the sun goes down? It depends on what gets hosted, but there’s this idea of how we deal with the inputs and outputs, the energy input, but also the warmth and heat it emits. People imagine it not just as a museum and meeting space, but as a place with a pool, a sauna, or greenhouses on top of the data center.

Through these workshops, I can say there are a lot of "cool and why not do it" ideas out there for community-owned and governed data centers. There is real interest once you speak to people about it. As for the single best answer, I'm curious to learn more through this workshop series, as the key themes shape people's thinking, so it gets decided by community input rather than just me speaking here.

Eleanor Drage (19:26)

The point is that people have opinions on what their data center should look like. These are people who may never have thought about what a data center is, where it is, or what it does. There's no reason why the everyday person, even those with no contact with large AI devices, can't make an impact on how data centers are governed in their local municipalities. So it's context-specific, right? Not all data centers should look the same. People have different ideas about how to decentralize them. Tell us about the German context, since you're based in Berlin. What does decentralizing data centers look like there?

Tara Merk (20:11)

I want to double down on what you said; context matters a lot. It matters in two ways. The first is how people think about what "community" could be. One of our collaborators with the Weizenbaum project is based in Mexico and has been working a lot with indigenous land rights and governance. He and his communities will have a very different way of thinking about these things compared to me in a German context.

The second way is the culture and legal system. At the end of the day, we’re talking about land, buildings, and electricity. These are highly regulated infrastructures, and we have a different legal toolkit to pick and choose from depending on the context. In Germany, I think three things are particularly interesting.

The first is our very strong federated structure. I did my PhD and postdoc in France, and the historical context there is quite different. In Germany, we have experience doing things at the state level rather than just the national level. While piecing things back together can be annoying, I believe it's a good starting point for thinking about localized digital infrastructures.

The second aspect is the idea of why communities would own infrastructure. Over recent years, as Germany has transitioned toward renewable energy, there’s been a rollout of wind turbines across the country. We’ve heard stories where turbines are built next to small towns, and residents initially object, saying they’re ugly or ruin the view. However, there have been several cases where residents became co-owners of the wind turbine. They get a share of the financial benefits and, in some cases, can buy the energy it produces. This changes the way those communities approach the infrastructure. We’ve seen cases where people are proud of "their" turbine; it becomes part of the local scene and makes the town unique. This experience with renewable energy infrastructure can really inform how we think about data centers. They might not be the most beautiful addition to a landscape, but they provide services, so why not apply similar thinking?

Lastly, there is the historical learning that defines how legal tools can be used. As part of my fellowship project at the New School, I've been conducting a historical case study of a cooperative data center group that operated in Germany for over 50 years. It served a network of cooperative banks, the Volks- und Raiffeisenbanken. However, in 2015, it merged with a non-cooperative data center group doing the same thing in southern Germany and subsequently lost its cooperative legal status. Through my research, I'm trying to figure out why that happened and if there were specific policies or ecosystem levers that could have maintained that cooperative entity. Seeking out these historical precedents for alternative ownership of data centers helps us understand what was possible before and how we might make it possible again.

Context really matters, geographically, historically, and culturally, in bringing those pieces together.

Eleanor Drage (25:06)

Governance and the law must play a vital role in ensuring that municipalities and citizens aren't just offered a symbolic share but are actually given a good deal. How do we govern these massive physical structures effectively? Does it involve managing real estate and land? I know there are many tried-and-tested ways of managing locally run land and pastures. What does that have to do with managing data centers?

Tara Merk (25:42)

That is a key piece. The experience people have in co-managing and co-governing, and "community" is a big word, so it’s a matter of defining who that community is, which depends on the legal tools available in a given jurisdiction to collectively manage resources like housing, real estate, and land.

The interesting connection here is to my other research, which focuses on community ownership of virtual platforms, protocols, and data, the virtual parts of the internet stack, rather than the physical. When we think about data centers, we can lean on the fact that for hundreds of years, people have lived in co-op houses, squatted buildings, managed land in indigenous ways, and created complex but useful agreements. This isn't too far removed from our shared knowledge. Most people likely have a friend who lives in a co-op or has been part of a community garden. We already know how to do this.

This is different from virtual governance, using crypto tokens or platform co-ops, which can get very abstract. We often have to "unlearn" how we deal with digital platforms today, where we accept huge companies making unilateral decisions, like stopping fact-checking or allowing racial slurs. To be successful in virtual co-governance, we have to unlearn the mentality that our only option is to "exit." I believe community-owned data centers have an advantage because of that shared history of managing physical resources. This is just another industry vertical where we should apply that experience.

Eleanor Drage (28:27)

Exactly, we've always been able to do it, so there’s no reason we can't do it here. Tell us about the internet, because we don't often talk about ownership of the virtual realm anymore, at least not since the conversations about cybernetics in the seventies. Why do people seem to care more about physical infrastructure, and who should own the internet?

Tara Merk (28:54)

I'm not sure people necessarily care more about physical infrastructure, especially if you ask the Web3 and blockchain communities. However, with the crash of the NFT market, we see that physical ownership still has a strong appeal. People cashing out Bitcoin to buy cars signals that we intuitively care about the physical because it is tangible.

Nevertheless, virtual ownership is becoming increasingly important, especially regarding data. The question of who owns our data, using "ownership" broadly, is crucial because whoever owns that data determines how it is used. In both the virtual and physical spaces, ownership is about the rights associated with it: the power to decide how something is used, who can access it, and who cannot.

As for who should own the internet, it goes back to my definition of good technology: technology that serves the people affected by it and is accountable to them. That feedback loop of accountability can be anchored in ownership rights. I believe that ownership should include both the physical assets that make up the internet and its virtual elements. While the practical implementation will look different for each, moving in that direction will hopefully yield more "good technology" in the long run.

Eleanor Drage (31:35)

Fantastic. You heard it here first! I also want to give a shout-out to the indigenous groups doing incredible data sovereignty work. I’ve worked with Jason Lewis from Indigenous Protocol, and I know the Native Nations Institute does a lot of cool things. These groups are really pioneering the way we think about data sovereignty. Tara, I could speak to you all day, but I have to bring this to an end. Thank you so much for introducing everyone to the idea of democratized and decentralized data centers. For many, this will be new, and I hope people will shrug off that apathy and feel that participation is truly possible. Thank you.

Tara Merk (32:17)

Thank you so much.