The three of us have been writing an article together called “The Making of Critical Data Center Studies,” in which we track its development over the past decades to establish it as a field in its own right.1 In planning these Five Things about Critical Data Center Studies, we thought carefully about different ways to approach this assignment: was it better to talk about news headlines, or go by new concepts in emergent scholarship, or more effective to draw out important tensions in the field?
We settled on an elemental approach, which allows us to do all of these things while also specifically addressing urgent social, political and environmental concerns that make this relatively new field ‘critical.’ As Nicole Starosielski (2019) suggests, “over the past decade, media studies has become elemental. By this, we mean that the field has become attuned to constituent parts, especially to the substances and substrates that compose media.” Within such an understanding of media, researchers can begin articulating the constituent parts of data centers—the (rare, and otherwise) earths found in the hardware components of servers, the water that is used for cooling the servers, the labor that goes into training the data located on the servers, the power and energy that surges through data centers, the heat exhaust that is amassed from server computation, and so on. As media studies scholars bring data centers into their modes of analysis, political questions emerge about the rapid and ongoing rise of data accumulation in nearly all sectors of public and private life.
By simply thinking of the data center, we have already placed ourselves. We are here, at the center of data, where data gather, where we gather with them. What a limiting spatial concept. The center is only a temporary rest stop on the way to another center. What about the journey? While data centers are situated in specific locales, the insides of the data center (components within computer servers as well as components within the energy infrastructures that power data centers, such as solar panels and windmills) are mined, smelted, refined, and transported from geographies all over the planet. Cooper (2021) draws our attention to data peripheries, a term developed as “an attempt to denaturalize the data center as the primary place where the materiality of Big Data storage is analytically accessed.” In other words, data peripheries attempt to move data center studies “into the undertheorized world of the material supply chains undergirding Big Data, attending more deeply to the industrial mobilities of materials and labor that precede the data center, much of which come out of the Global South.” These data peripheries include the myriad logistics spaces (Cowen, 2014) entangled in the manufacturing of the servers and hard drives of the data center, as well as the global sites of resource extraction that Sy Taffel (2021) characterizes as a planetary-scale metabolic rift in service of digital life.
Earth extraction leaves much in its wake — scars on the planet, hazardous waste, dangerous and unfair labor practices, ecological devastation, and more. The ongoing growth and maintenance of data infrastructures is intimately tied to our current epoch of environmental catastrophe (Parikka 2015). To this end, earth extraction is tangled up with histories and ongoing practices of (settler) colonialism in the forms of land dispossession and violence against people and the more-than-human world (Yusoff 2019). Many industries are prioritizing settler futures through extraction practices aimed at growing green economies where metals and rare earths are deemed “strategic” or “essential” materials (Riofrancos 2021), and technology companies are a major player in this ongoing project (Hogan 2018). As data centers continue to expand by both adding more servers and renewable energy sources to their portfolios, the question of earth extraction — and the peripheries of the data center — will remain a central one for critical data center studies.
Central for cooling rows upon rows of servers, water is an important infrastructural element in the making of data centers. Data center scholars have delved into the environmental politics of the National Security Agency’s (NSA) Utah data center where the United States government uses millions of gallons of water daily to cool its servers (Hogan 2015). They have examined Facebook’s water usage in New Mexico, USA, where the company has secured water rights to pull millions of gallons of water per day from desert aquifers along the troubled Rio Grande (Edwards 2020). And they have taken us to Google’s South Carolina data center where water has been the subject of debate, controversy, and “agri-cultural” contestations of power (Gilmore & Troutman 2020). The entanglement of water and digital culture has increasingly captured the attention of the news media, communities, and other industries vying for water rights. As one sign read at a protest against the expansion of Facebook’s New Mexico data center, “Farmers need water. You can’t eat data.”
Water is poised to be an ongoing concern for the data center industry. Many tech companies — such as Google/Alphabet, Meta, and Microsoft — are making commitments to be “water positive” in the near future. While such commitments circulate as flashy headlines, a number of questions follow about what such commitments look like in practice. Who is defining what water positive means in practice? And what accountability metrics are in place to verify such commitments? Thus, a new wave of critical data center studies scholarship might investigate what is actually meant by “water positive,” paying close attention to how the concept is constructed by the industry and monitored through third-party verification firms. More work might also embark upon field studies that observe, document, and, if possible, verify, big tech’s involvement in water restoration projects.
The data center industry makes big promises: to transform local markets, to draw in talent, to modernize economies, and so on. But rarely, if ever, does this prove to be true (anymore) beyond low level job creation within the data center itself. More often than not, the industry gets tax breaks and exploits local resources; none of it benefiting the site, city, or community where the data center is built (Pickren 2017; Burrington 2015). As noted by A.R.E Taylor (2019), “Architects, construction contractors, cleaners, facility managers, disaster recovery officers, maintenance workers, security guards, service technicians, sales teams and secretaries, are just some of the people that are essential to running the data center.” In this way, Vicki Mayer (2020) invites us to consider the political and cultural conditions that help explain the attraction to the tech industry and its appeal by governments and local authorities to create mini silicon valleys everywhere they can, as though there is something inherently vibrant and good about new technology (and this despite their poor labor and environmentally destructive practices). Mayer attributes this in part to the ‘aura’ of data centers, which extends to tech culture more generally (NVIDIA, Google, Box, Meta, etc.) and speaks to the less tangible affective qualities of the industry that invite, trick, or compel people to labor for them.
Labor is elemental to the data center story (Monroe 2022). Whether discourse about retrofitted tech industries generating jobs locally (Jacobson & Hogan 2019; Amoore 2018; Mosco 2014), or the automation and fantasy of ‘depeopled’ data centers (Gonzalez 2019; Taylor 2019), or the idea of data itself laboring on servers (Hu 2015), its various instantiations gesture to assumptions about the future – one that encapsulates and is emboldened by the idea that technology is progress, and that the accumulation of data gives way to some kind of universalised human intelligence (i.e. big data means closer to Truth). The focus on labor, as elemental, is therefore important for the way it draws attention to those assumptions, but also to actual human bodies in tech spaces, as well as questions of politics, ownership, and power – back from the materiality of the infrastructure and from commercialized images of data centers as ‘depeopled’ and purely technological spaces (Hogan & Vonderau 2019).
The International Energy Agency estimates that data centers currently consume between 1-1.4% of the world’s energy (IEA 2022). This number cannot possibly encapsulate the sheer rapidity at which data/energy relationships are evolving. Data center companies are fast becoming some of the largest energy buyers on the planet, and there is increasing vertical integration of data centers and energy infrastructures. Facebook/Meta has invested directly in numerous renewable energy projects across the United States. Data centers are not simply energy consumers, but are emerging as major players in the energy industry themselves. There are even increasing calls to treat data centers, and their unique demand/response affordances as “virtual power plants,” allowing, among other things, the return of allocated energy to the grid during peak hours. In this way, data centers can, at least rhetorically, become energy producers.
Shifting data/energy relationships are complicating the very concept of the data center itself. As we’ll discuss in our longer article, the 1-1.4% figure depends heavily on how one defines “data center” in the first place. There is much structural diversity in the data center space, and current quantitative measurements of net energy consumption do not reflect this diversity. For example, if cryptocurrency mining was considered then the above figure would double. As the cryptocurrency mining industry has professionalized with an ever-expanding infrastructural footprint, it has built new kinds of energy relationships that are now filtering into the more “traditional” data center industry. Iceland’s entire data center infrastructure was built on the back of cryptocurrency mining. Until 2020, up to 90% of all clients in Iceland’s data centers were cryptocurrency miners. We should be asking deeper questions not just about data center energy consumption, but about the evolution of data/energy relationships. As the data center takes on more varied forms, and becomes more and more difficult to define, focusing on these relationships will be key to better understanding how to regulate data center energy consumption more broadly.
At all points in the computing process, the system struggles, and ultimately always fails, to contain the inevitable heat it produces. Starosielski (2021) reminds us that the history of digital technology can be understood as the “coldward course of media.” Keeping media cold is the losing battle data centers continually face, and climate change is exacerbating these losses. Earlier this year, in the UK, multiple data centers quite literally “melted down” in the face of extreme temperatures. More recently, a raging fire took out two South Korean data centers.
It is always somewhat surprising when heat sneaks into the narrative, but it shouldn’t be. Understanding heat helps us understand the material foundations and limitations of digital data in the first place. Heat helps us feel computing at a visceral level. Heat has always been at the center of digital data management. In fact, the US legally regulates data centers both as information processing and thermal control entities. Current legal definitions of data centers in the U.S. do not deal with the data, or the labor of processing and distributing it. According to Title 42, Section 17112 of the U.S. Code (2007), a data center must have “environmental control equipment to maintain the proper conditions for the operation of electronic equipment.” The US Code requires a data center to have sophisticated temperature regulation in order to be legally classified as a data center. Thus, a data center isn’t a data center unless it has air conditioning. “Climate bunkers” (Moro 2021) indeed.
The regional politics of data centers are shaped in part by climatic politics. The endless battle between heat and data projects what Starosielski (2022) calls thermopower across geographies. As a result, the greater Arctic has become a key site for data center development. Asta Vonderau (2019) has written extensively about the data center development in Sweden, and how Sweden’s Arctic climate (in particular its cold air) is directed, managed, and branded in service of the burgeoning data center industry. Furthermore, Alix Johnson (2019) illustrates how the branding of data center development in Iceland encourages the assumption that the Arctic, due to its reliably low temperatures, is, quite simply, the “natural home” for data centers. This is an aesthetic move, one that renders enterprise-scale data as a process embedded in and, in effect, a part of, the natural ecology in regions with cold climates. Data Centers by Iceland, a government organization that promotes Iceland’s data center business, leans heavily into this aesthetic, producing all sorts of content that juxtapose high-tech data centers with pristine waterfalls and dancing aurora, while often repeating the decade-old refrain, “Iceland is the coolest place for data centers!”
Heat, or the lack thereof, defines how data centers look, how they are understood, and also reveals their social, political, and ecological limitations.
— Mél Hogan, Dustin Edwards, and Zane Griffin Talley Cooper