In this literature review from the Minderoo Centre for Technology and Democracy, Julia Rone outlines the key trends and logics behind the boom in data centre construction across the globe.
Hamlet: Do you see yonder cloud that’s almost in shape of a camel?
Polonius: By th’ mass, and ‘tis like a camel indeed.
Hamlet: Methinks it is like a weasel
Polonius: It is backed like a weasel.
Hamlet: Or like a whale?
Polonius: Very like a whale.
The cloud – this fundamental building block of digital capitalism – has been so far defined mainly by the PR of big tech companies.
The very metaphor of the ‘cloud’ presupposes an ethereal, supposedly immaterial collection of bits gliding in the sky, safely removed from the corrupt organic and inorganic matter that surrounds us. This, of course, can’t be further from the truth.
But even when they acknowledge the materiality of the ‘cloud’ and the way it is grounded in a very physical infrastructure of cables, data centres, etc., tech giants still present it in a neat and glamorous way. Data centres, for example, provide carefully curated tours and are presented as sites of harmoniously humming servers, surrounded by wild forests and sea. Some data centres even boast with saunas.
Instead of accepting blindly the PR of tech companies and seeing ‘the cloud’ as whatever they present it (similarly to the way Polonius accepts Hamlet’s interpretations of the cloud), we should be attuned to the multiplicity of existing perspectives on “the cloud”, coming from researchers, rural and urban communities, and environmentalists, among others.
In this lit review, I outline the key trends and logics behind the boom in data centre construction across the globe. I base the discussion on several papers from two special issues. The first one is The Nature of Data Centres, edited by Mél Hogan and Asta Vonderau for Culture Machine. The second: Location and Dislocation: Global Geographies of Digital Data, edited by Alix Johnson and Mél Hogan for Imaginations: Journal of Cross-Cultural Image Studies. I really recommend reading both issues – the contributions read like short stories and go straight to the core of the most pressing political economy problems of our times.
The ‘nature’ of data centres
Data centres as key units of the cloud are very material: noisy, hot, giant storage boxes containing thousands of servers, they occupy factories from the past or spring up on farm land all over the globe. Data centres are grounded in particular locations and depend on a number of ‘natural’ factors for their work, including temperature, humidity, or air pollution. In order for data centres to function, they not only use up electricity (produced by burning coal or using wind energy, for example). They also employ technologies to circulate air and water to cool down and emit heat as a waste product.
But data centres are not only assemblages of technology and nature. Their very appearance, endurance and disappearance is defined by complex institutional and non-institutional social relations: regions and countries compete with each other to cut taxes for tech corporations that promise to bring jobs and development. Some states (e.g. Scandinavian states) are preferred over others because of their stable institutions and political ‘climate’.
No blank slate
To illustrate, the fact that data centres are built in Sweden’s Norrbotten region has to do a lot with the ‘nature’ of the region conceptualised reductively by tech companies as cheap energy, cheap water, cheap land and green imagery (Levenda and Mahmoudi, 2019, 2). But it also has to do a lot with the fact that Norrbotten is filled with the “ruins of infrastructural promises” (Vonderau, 2019, 3) – “a scarcely populated and resource-rich region, historically inhabited by native Sami people, the region was for a long-time regarded as no-man’s land” (ibid). Not only is Norrbotten scarcely populated but it also has an “extremely stable and redundant electricity grid which was originally designed for […]‘old’ industries” (ibid, 7).
A similar logic of operation could be discerned in the establishment of a data centre in the Midway Technology Centre in Chicago, where the Schulze Bakery was repurposed as a data centre (Pickren, 2017) Pickren was told in an interview with a developer working on the Schulze redevelopment project that “because the surrounding area had been deindustrialized, and because a large public housing project, the Robert Taylor Homes had closed down in recent decades, the nearby power substations actually had plenty of idle capacity to meet the new data centre needs” (Pickren, 2017). As Pickern observes, “there is no blank slate upon which the world of data simply emerges”(ibid.) There are multiple “continuities between an (always temporary) industrial period and the (similarly temporary) ascendancy of digital capitalism” (ibid).
Extraction and the third wave of urbanisation
What the examples of Norrbotten in Sweden and the redevelopment of Chicago by the data industry show is that despite a carefully constructed PR around ‘being close to nature’ and ‘being green’, decisions on data centre construction actually depend on availability of electricity for which depopulation is only a plus. Instead of ‘untouched’ regions, what companies often go for are rather abandoned or scarcely populated regions with infrastructure left behind. Data centres use resources – industrial capacity or Green energy – that are already there, left from previous booms and busts of capitalism or from conscious state investment that is now used to the benefit of private companies.
“Urban interactions are increasingly mediated by tech and leave a digital trace – from paying your Uber to ordering a latte, from booking a restaurant to finding a date for the night.”
Both urban and rural communities are in fact all embedded within a common process of a ‘third wave of urbanisation’ that goes hand in hand with an increase in the commodification and extraction of both data and ‘natural’ resources (Levenda and Mahmoudi, 2019). What this means is that urban interactions are increasingly mediated by tech and leave a digital trace – from paying your Uber to ordering a latte, from booking a restaurant to finding a date for the night.
This urban data is then stored and analysed in predominantly rural settings: “[T]he restructuring of Seattle leads to agglomerations in urban data production, which rely on rural data storage and analysis” (ibid, 9). Put simply, “[J]ust as Facebook and Google use rural Oregon for their ‘natural’ resources, they use cities and agglomerations of ‘users’ to extract data”.
Ultimately, data centres manifest themselves as assemblages for the extraction of value from both people and nature.
As if in a perverse rendition of Captain Planet, all elements – water, air, earth, human beings and technology – unite forces so that data centres can function and you can upload a cat photo on Facebook. In this real life data-centre version of Captain Planet, however, all elements are used up, extracted, exhausted. Water is polluted.
People live with the humming noise of thousands of servers.
Taxes are not collected and therefore not invested in communities that are already deprived.
What is more, data centres often arrive in rural regions with the promise to create jobs and drive development. But as numerous authors have shown, actual jobs created by data centres are less than what was originally promised, with most jobs being precarious subcontracting (Mayer, 2019). As Pickren notes, “If the data centre is the ‘factory of the 21st century,’ whither the working class?”
Data centres do create jobs but predominantly in urban areas. “[W]here jobs are created, where they are destroyed and who is affected are socially and geographically uneven” (Pickern, 2017). Where value is extracted from and where value is allocated rarely coincide.
And if from a birds view perspective, what matters is the total number of jobs created, what matters in Sweden’s Norrbotten or The Netherlands’ Groningen, where data centres are built, is how many jobs are created there and furthermore, what types of jobs (Mayer, 2019). In the same way, while from an abstract point of view tech companies such as Microsoft might be “carbon neutral”, this does not change their questionable practices and dependence on coal in particular places.
The Introduction to the “Location and Dislocation” Special Issue quotes a classic formulation by Yi-Fu Tuan according to whom “place is space made meaningful” (Johnson and Hogan, 2017, 4).
“Whenever we hear big tech’s grandiose pledges of carbon neutrality and reducing carbon emissions, we need to understand that these companies are not simply “green-washing” but are also approaching the problem of global warming “in the abstract””.
One of the key issues with tech companies building data centres is the way they privilege space over place – an abstract logic of calculation and global flows over the very particular local relations of belonging and accountability.
In a great piece on “fungible forms of mediation in the cloud”, Pasek explores how the practice of big tech companies to buy renewable energy certificates does more harm than good, since it allows “data centre companies to symbolically negate their local impacts in coal-powered regions on papers, while still materially driving up local grid demand and thereby incentivizing the maintenance or expansion of fossil energy generation” (ibid, 7).
The impact for local communities can be disastrous: “In communities located near power plants, disproportionately black, brown and low-income, this has direct consequences for public health, including greater rates of asthma and infant mortality” (ibid).
So whenever we hear big tech’s grandiose pledges of carbon neutrality and reducing carbon emissions, we need to understand that these companies are not simply ‘green-washing’ but are also approaching the problem of global warming ‘in the abstract’, at the global level, paying little attention to their effect in any particular locality.
As Pasek notes, this logic of abstraction subordinates the ‘urgencies of place’ to the ‘logics of circulation’.
Unsurprisingly, it is precisely the places that have already lost the most from previous industrial transformations that are the ones who suffer most during the current digital transformations.
Invisibility and hypervisibility
What makes possible the extraction practices of tech companies is a mix between how little we know about them and how much we believe in their promise of doing good (or well, not doing evil at least).
In her fascinating essay The Second Coming: Google and Internet infrastructure, Mayer (2019) explores the rumours around a new Google data centre in Groningen. Mayer explores how Google’s reputation as a leading company combined with a the total lack of concrete information about the new data centre create a mystical aura around the whole enterprise: “Google’s curation of aura harkens back to the early eras of Western sacred art, during which priests gave sacred objects their magical value by keeping them ‘invisible to the spectator’” (Mayer, 2019, 4).
Mayer contrasts a sleek Google PR video (with a lone windmill and blond girls looking at computer screens) with the reality brought about by a centre that offered only a few temporary subcontracting jobs. The narrative of regional growth presented by Google unfortunately turned out to be PR rather than a coherent development strategy.
Furthermore, in a fascinating essay on data centres as “impertinent infrastructures”, Velkova (2019) explores the temporality and impermanence of data centres that can be moved or abandoned easily.
How could such impertinent structures provide regional development?
What is more, even if data centres do not move, they do reorganise global territories and connectivity speeds through the threat of moving: “data center companies are constantly reevaluating the economic profitability of particular locations in synchrony with server replacement cycles and new legislative frameworks that come into force”.
Data centres are above all impermanent – they can come and go. Rather than being responsible to a particular locality, data centres are part of what Pasek called a “logic of global circulation”.
“Should tax regulations, electricity prices, legislation or geopolitical dynamics shift, even a hyper-sized data center like Google’s in Finland or Facebook’s in Sweden could make a corresponding move to a place with more economically favourable conditions within three years” (Velkova, 2019, 5).
So data centres are on the one hand, hypervisible through corporate PR. On the other hand, they are invisible for local communities that are left guessing about construction permits, the conditions of data centres arrival, their impact on the environment and the economy.
But ultimately, and this is the crucial part, data centres are above all impermanent – they can come and go. Rather than being responsible to a particular locality, data centres are part of what Pasek called a ‘logic of global circulation’.
Holding each node accountable
Big tech’s logics of extraction, abstraction, invisibility, hypervisibility and impermanence are driving the current third wave of urbanisation and unequal development under digital capitalism.
But it is possible to imagine another politics that would “hold each node accountable to the communities in which they are located” (Pasek, 9).
The papers from the two special issues I review here provide an exhaustive and inspiring overview of the ‘nature”’ and imaginaries of data centres.
Yet, with few exceptions (such as the work of Asta Vonderau), we know little about the politics of resistance to data centres and the local social movements that are appearing and demanding more democratic participation in decision making.
Would it be possible for us – citizens – to define what the cloud should look like? Not sure. But this is a crucial element of any project for democratising digital sovereignty. And this is what I work on now.