Scaling the Cloud: Making State and Infrastructure in Sweden

ABSTRACT

Popular representations imagine the internet as being immaterial and fluid; hidden from the public eye are the industry and complex infrastructure securing the functionality of the World Wide Web, as well as this industry’s social and environmental effects. Focussing on the implementation of a Facebook data centre in the Swedish city of Luleå, this article investigates how the global cloud is localised within a specific historical and social context. It shows how this new industrial development becomes a part of state-making and regional identity-building processes by triggering the re-scaling of territories and shaping new geographies in relation to expanding cloud infrastructures. Tracing those infrastructure-making processes reveals some of the key dynamics between the Swedish state, its regions and the global IT economy.

KEYWORDS:

• Cloud computing
• data centre
• infrastructure
• Facebook
• Sweden

Introduction

In 2014, Volvo published a cross-promotional video entitled Swedish Thinking: Journey to the Node Pole on its YouTube channel.¹ The video carries its viewers to Luleå, a small town high up in Sweden’s North where Facebook had recently begun building its first European data centre, arguably also the largest and most innovative to date. Acting as the main protagonist in the video’s narrative, municipal commissioner Karl Petersen relates Luleå’s story as that of a city that has always managed to make productive use of its cold climate and peripheral geographical position, in this case by cooling servers and securely storing the world’s Big Data. Strolling through his snow-covered hometown, Petersen suggests:

In Sweden we make the most out of the cold. When I became mayor, I saw the potential for change. The world needs a lot of Big Data in the future. […] Data centres need cooling. We use the arctic cold to get more from less. In Sweden we have a climate for innovation.

As Petersen expresses his pride over the fact that Facebook has decided ‘to come to Luleå,’ the video shows images of large cooling fans and endless rows of server racks, followed by time-lapsed images of Luleå’s streets that evoke a luminous stream of data to accompany his observation that ‘now the world’s data is flowing through Sweden’. The video concludes with advertising copy appearing on the screen – ‘Swedish thinking gets more from less. That’s why Volvo’s Drive-E powertrains get more power from less fuel’ – linking the preceding discussion to Volvo’s well-known Swedish brand. The video thus makes Facebook’s data centre appear as one of the most prominent Swedish technological achievements of recent times, and also as one reflecting a specific national mindset or determination – ‘getting more from less’ – which is similarly reflected by Volvo.

Such imaginings of the North as a site for industrial progress, scientific discovery and innovation date back to eighteenth- and nineteenth-century narratives that were crucial for the development of the Swedish nation-state and its technological modernity (Sörlin 2002). However, these visions relate to more than just imagined geographies of innovation, as the video’s title readily indicates: The Node Pole is a public–private company for data centre industry development, which was established in the wake of Facebook’s decision to locate its data in Sweden. The Luleå-based company sells land, cold northern air and cheap hydro power to data centres that provide the infrastructure for the global cloud.²

This article is based on one and a half years of research on the emerging data centre industry in Northern Sweden. My examination of Facebook’s data centre implementation in Luleå reveals how this industrial development has triggered a re-scaling of territories, shaping new geographies in relation to expanding cloud infrastructures (Brenner 1999: 436). In the first part of this article, I will analyse complications that typically mark the relation between national territories and border-crossing IT industries and infrastructures. In the second part, specific practices and strategies of scale-making and their effects are discussed (Barry 2006; Kaijser et al. 2015 ).

Without making an a priori distinction between ‘global forces and local places’ (Tsing 2000: 352), I understand scale-making practices as collaborative and contested undertakings conducted by multiple actors which have different interests and are situated in diverse places and social contexts. Adopting a term coined by ‘tech visionary’ Richie Etwaru that is circulating within the field, I describe such practices as clouding. Etwaru suggested to think the cloud as a verb and urged to rescale and extend the principle of IT-mediated effectiveness beyond the realm of data storage (which just offers infrastructure as a service), aiming at an ultimate state of efficiency and profitabilty which he calls ‘everything as a service’ (Etwaru 2014; Hu 2015). Describing practices of scale-making as clouding, I thus highlight their relatedness to such powerful technological visions of frictionless profit flows which repeatedly have been described as being characteristic for contemporary (digital) capitalism (Ferguson 2005; Hu 2015). However, clouding, as I use the term here, also refers to the changing modes of the cloud’s visibility – to the strategic and experimental foregrounding and obscuring of certain infrastructural relations, materialities and effects which are initiated by differently situated actors without necessarily harmonising with corporate ideas of frictionless profit flows. The new geographies that emerge as a result of scaling and clouding are not just outcomes of formal political programmes, planned industrial developments and engineered technological innovations. They rather come into being as a result of a mutual dynamics between planned and unplanned alliances, engineered and non-engineered human and non-human relations and their foreseen and unforeseen effects. Geographies of the cloud are thus dynamic and experimental (Jensen & Morita 2017: 619).

Tracing the emergent geographies of the cloud, this study uncovers the ways in which the global cloud depends on, and is shaped by, local infrastructural sites, and how it at the same time shapes local politics and power relations. It demonstrates, first, how corporate ideologies of an immaterial and global cloud relate to the Swedish government’s attempts to make the cloud national. Second, it shows how the globally encompassing data centre industry’s dependency on regional resources contributes to the changing relations between the state and its peripheral regions. Third, it asks how this industry builds on, and integrates into, the natural and industrial environment of Luleå and Norrbotten, thus transforming urban landscapes and local future imaginaries.

The research presented here is situated in the field of infrastructure studies which brings together scholars from various disciplinary backgrounds such as Human Geography, Sociology, History, Science and Technology Studies and Anthropology (Ruhleder & Star 1996; Barry 2006; Bowker et al. 2010; Kaijser et al. 2015; Harvey et al. 2017). Anthropology contributes to this research field with its situated attention to the processes of infrastructuring, that is, with ethnographic studies on how infrastructural implementations generate effects and continuously reproduce and transform socio-technical relations in different local contexts (Harvey et al. 2017: 5). Such studies call for a methodological retooling and the development of an ‘infrastructural toolbox’ which offers both new possibilities for investigating complex technological systems and for theorising key anthropological issues about the making of states and markets (Larkin 2013; Appel et al. 2015). They also highlight the emergent and experimental character of infrastructures and investigate their capacities to generate new configurations of the world, to give ‘form to culture, society and politics’ (Jensen & Morita 2017: 617).

In contributing to anthropological infrastructure studies, this article foregrounds the material downside of seemingly just virtual ‘information highways’ (Ruhleder & Star 1996; Star 1999: 379). It points to the backstage infrastructural work that enables global data flows and renders their local entanglements visible (Bowker et al. 2010). Infrastructure here is understood as fundamentally relational – situated not only in regard to diversely located human and non-human actors who are connected through infrastructuring processes, but also in terms of temporal dependencies and an embeddedness of ‘new’ infrastructures into historical technological and social contexts and natural environments (Hu 2015; Starosielski 2015). ‘Cloud infrastructure’ I define as a socio-technical configuration that is material and expansive, open-ended, ontologically vague and experimental.

Accordingly, Facebook’s data centre in Luleå is seen as a site where local regional and national actors meet border-crossing technological developments, and where the state’s space ‘as material rational and abstract potential’ is constituted (Harvey 2012: 77). Rather than attempting to grasp cloud infrastructure as a whole, Facebook’s Swedish project is used as a lens to study infrastructuring – a continuous process of ordering, of relating and mediating between places, technical structures, moral values, organisational resources and communities of actors (Niewöhner 2015).

Following such processes, this article provides insights into what Larkin (2013) has called the ‘politics of infrastructure’, that is, political rationalities and practices of government which emerge and become observable in the course to infrastructural implementations. Demonstrating how those rationalities are assembled and come into being, I take into account both infrastructures’ planned effects as well as its unintended consequences, such as unexpected human and non-human alliances, for instance, technological problems or frictions between different communities of practice (Cubitt et al. 2011; Dourish & Bell 2011). Rendering infrastructuring (dis-)orders visible, I am going to show how local infrastructuring processes form part of the border-crossing technological zones of digital capitalism (Ferguson 2005; Barry 2006; Chalfin 2015).

The analytical perspective chosen here acknowledges that infrastructures have no single and static scale. They generate scales through the work of ordering, extending, configuring and foregrounding (Lea 2017). This perspective allows for studying through different contexts, beyond global/local or virtual/real divides, making scale-making itself an object of investigation (Tsing 2000; Thrift et al. 2014). Investigating scale-making thus means to analyse ‘through what social and material processes and cultural commitments localities or globalities come […] into being’ (Tsing 2000: 348), and how the geographies of the cloud are made real. Such an investigation needs, according to Tsing, to trace the rhetorics of scale – ‘cultural claims about locality, regionality, and globality’ (Tsing 2000: 347), as well as the changing definitions of interests and identity that both allow and result from collaborative activities of infrastructuring (Tsing 2000: 352).

The argument proposed here is based on repeated periods of fieldwork (ten months in total) mainly conducted in Luleå and its surroundings, but also at related sites in Stockholm, London, Monaco and Hannover, where I visited industrial associations, attended international trade fairs, conferences and professional development courses. My empirical material consists of more than 50 interviews and conversations with different actors involved in the implementation of the Facebook data centre in Luleå and in Norrbotten’s data centre industry. The interview subjects included politicians and decision-makers active on both regional and national levels, such as representatives of Luleå’s Municipality, Business Agency, Norrbotten’s County Administrative Board, Invest in Norrbotten, Business Sweden and The NodePole; architects, city planners, ecologists, technicians and other municipal employees; local journalists; private subcontractors and consultants such as NCC, a construction company; data centre engineers and managers; and IT scholars at the Cloudberry Data Center project at Luleå’s Technical University (LTU). The interviews provided me with information about different actors’ interests and motivations for engaging in the development of the new industry. In addition to interviews, my material includes first-hand observations of data centre sites and related industrial, academic and public events such as, for instance, the book launch of Hello Mister Mayor, Facebook is Calling (2014) in Luleå, or The Node Pole reception at the Data Cloud Europe 2015 conference in Monaco. Through observation, I was able to grasp the diversity of representational and maintenance work conducted in the process of the cloud infrastructure implementation in Luleå and Norrbotten. Finally, I have extensively used and studied steering documents such as Norrbotten’s Data Center Strategy or Sweden’s Digital Agenda, alongside current and historical press reports on local industrial developments. Those documents were instrumental for contextualising the Luleå Facebook project in a broader political context, especially within official strategies of branding Sweden as an IT-nation.

In order to underline the methodological choices made in this research, and the constraints my research inevitably is subjected to, it is important to note that industrial-scale data centres are usually closed to the public. Globally operating corporations such as Google or Facebook are especially careful in guarding their infrastructure and in channelling all communications through central PR departments. In addition to the secrecy surrounding them, data centres are automated spaces designed to host technologies, not humans. Only a very few people actually work inside industrial-scale data centres, while most planning and operation processes are organised and regulated from outside the data centre building. Therefore, direct contact with Facebook data centre representatives in Luleå provided only limited information and necessitated a search for alternative possibilities for access to the data centre industry. Among those alternatives were smaller data centres and data centre construction sites which often are less securitised and more easily accessible, as well as industrial associations and trade groups such as DataCenterDynamics. Subcontracted companies and local authorities involved in the construction or maintenance of data centres opened up additional possibilities for acquiring knowledge about the cloud’s technological and social organisation. ‘Being in the field’ while investigating cloud infrastructure thus necessarily means polymorphous engagements (Gusterson 1998) – a multi-sited interaction distributed along and around cloud infrastructures in different local and transnational contexts, engaging with diverse expert cultures, materialities and stocks of knowledge.

The Intricacies of Infrastructure

In order to understand the ways in which the infrastructuring of the cloud may trigger a rescaling of national territories and shape new geographies, it is necessary to describe data centres in terms of their specific materiality and topography as well as in terms of the imaginaries and expectations that dominate the popular perception of the cloud. Data centres are not only at the core of cloud infrastructure, but they also represent one of the fastest growing industries worldwide and serve as archives of the digital age. While the scope of data centre operation is potentially global, it is also dependent on local resources and energy grids. Analysing those local entanglements, therefore, reveals frictions and intricacies which characterise the relation between data centres as a vital part of global cloud infrastructure on the one hand, and the state with its various regions and authorities as well as diversely located subcontracted private companies and their commercial interest on the other.

The Materiality of the Cloud

Data centres are industrial-scale organisations offering the storage and delivery of data via the internet. They represent the fastest growing user of IT energy, consuming 3 percent of all global electricity, with this share growing by 12 percent a year (Cook & van Horn 2011; Bawden 2015). Cloud computing – the online storage of and real-time remote access to data – on which Facebook and other social networks and online services are based, has particularly contributed to the enormous growth of the IT industry and its energy use. Facebook’s data centre in Luleå alone requires one terawatt hour of electricity per year, equalling the total electricity usage of Sweden’s sixth biggest municipality, with a population of 146,000 (Eriksson 2013). With IT firms constantly looking out for cheap energy, remote and energy-rich regions around the world compete as potential sites of cloud infrastructure in the hope of solving local demographic and economic problems. The city of Luleå is a case in point. After undergoing a lengthy and highly competitive selection process among more than one hundred places in Sweden and Europe, Luleå won the bid due to its cool climate, low electricity prices and geopolitical security.

Equally crucial, however, were attractive deals with regional authorities and companies regarding the data centre’s energy and water supply systems, cheaply sold land, and not least state subsidies amounting to 100 million Swedish Krona in total, as the Head of Luleå’s City Planning Office and other interlocutors later admitted.³ According to those interlocutors, other Swedish municipalities were even offering land for free in order to gain the lead in the competition, and Facebook’s lawyers continuously played competing sites against each other, for instance by trying to make Luleå commit to take financial responsibility for possible losses caused by eventual shortage in the data centre’s water supply (Interview, November 2014).⁴ Winning such a tough competition raised optimistic expectations among regional officials that the new industry would help to create jobs and to provide alternatives to established local industries such as forest, steel and mining for both Luleå and the surrounding Norrbotten region, Sweden’s largest and northernmost municipality. Those expectations were supported by a rising awareness among local business developers that data centres, often characterised as ‘factories of the twenty-first century’ (Cook & van Horn 2011; Levy 2012; Siemens 2014), indeed resemble older process industries such as steel factories in terms of their energy needs, security requirements and engineering complexity, and thus may make use of existing local technological competencies.

While acknowleging data centres’ resemblance with other process industries, local media and politics tended to promote server farms as being different and better than ‘traditional’ factories, especially in regard to their environmental impact and resource demands. Such perceptions were mostly motivated by referring to different outer appearances of ‘old’ and ‘new’ industries in Lueå. The city’s steel plant,⁵ for instance, once celebrated as a symbol of Sweden’s industrial modernity, is visible from all over the city with its large, dark buildings and smoking chimneys. Visitors to the plant are surrounded by industrial noise and may catch glimpses of the industrial production process. Controlled explosions are regularly heard throughout the city, and the local press occasionally reports industrial accidents. In addition to daily sensory experiences and public awareness about working conditions or environmental damages, collective memories play a major part in shaping public perceptions of the local steel plants and mines. In Luleå, Sweden’s so-called steel city, individual biographies and family histories have been intertwined with the ups and downs of national steel production for generations (Öhman 2015). By contrast, the Facebook data centre is hardly visible and difficult to decode. Taking the shape of a large grey box, it lacks any signs of productivity apart from construction workers and a security firm’s cars driving around the centre’s premises. Imperceptible to outsiders are the typical roaring sound of cooling fans evoking the imagined noise of information highways, or the specific smell of industrial air (Blum 2012).

As has been repeatedly pointed out by scholars, wide-spread ecological or biological metaphors such as the ‘cloud’ contribute to the relative invisibility and limited public awareness of IT infrastructures and industries as material entities that have local social and environmental effects (Mackenzie 2011; Levy 2012; Carruth 2014). In a similar vein, ‘server farms’, or data centres, are often described as ‘modern, effective, clean and environmentally friendly’ alternatives to the ‘dirty, toxic and ugly’ factories of the twentieth century. As mentioned above, Facebook’s data centre in Luleå is no exception. It runs on renewable hydro power and marks a shift in Facebook’s energy policy that occurred after the IT giant had been heavily criticised by Greenpeace for its coal-based energy use.⁶ As a consequence, the data centre is promoted by Swedish national and regional media as contrasting with allegedly more problematic, traditional industries such as Norrbotten’s steel plants or mines. What is not taken into consideration in this respect, however, is the fact that after completion, Facebook’s data centre will consume more energy than Sweden’s largest steel plant. A local journalist described how Luleå citizens perceived the difference between ‘old’ and ‘new’ industries:

This town is used to pollution by smoke and other harmful substances from the SSAB steel plant, even the city’s most popular living area is close to the plant. And here, in Facebook’s data centre, it’s just a large building, there is not even a chimney, just electricity being used. (Interview, April 2014)

As I will show later, IT companies such as Facebook have a vested interest in dematerialised and lightweight representations of their infrastructures which help to ensure constant growth of internet usage and to downplay its environmental footprint (Cubitt et al. 2011; Maxwell and Miller 2012). From the viewpoint of IT marketing, the idea of sustainability, evoked in ecological metaphors so widely used to characterise the internet and its infrastructures, is thus an additional brand asset which helps to cloud the data centre’s actual resource and energy needs and problematic environmental effects.

The Topography of Data Processing

The difficulty of grasping the cloud in material terms and pinpointing its local embeddedness becomes even more obvious when considering the data centre industry’s main ‘product’, Big Data, and the topography of data processing that maps how data cross territorial and administrative borders from server farms to users. In the case of Facebook’s Luleå centre, over 120 different companies were involved in its construction, not to mention countless governmental agencies, regional development boards, and international consultancy experts. The local technological complexities remain invisible, however, to the majority of cloud users who are far removed geographically from the Luleå site which stores the data of 800 million Facebook users from Europe and other parts of the world. This remoteness of cloud infrastructure from the very users producing the data contributes to the appearance of data centres as merely neutral providers, responsible only for the processing of information (Jaeger 2009; Hogan 2015). Yet this obscures the fact that apart from economic interests inscribed into cloud infrastructure itself, processes of data storage are hardly neutral in regard to data security and privacy, among other issues. As Sandra Braman has noted, the cloud created a regulatory paradox when growing into an ubiquitous computing infrastructure. ‘The relation between globally dispersed and dynamic data flows, local infrastructural materialities and national regulations came apart, and bringing it all back together currently makes an overarching policy problem’ (2015: 277). This includes the question of whether the cloud ‘will be considered to legally be in one designated location (and therefore beholden to the laws, policies and regulations of one place), or in every location that has a data centre that is part of the cloud’ (Jaeger 2009: 7). An equally important issue is if data centres will be considered as just a neutral data processors or as responsible data controllers (Braman 2015: 282). The new EU General Data Protection Regulation,⁷ which comes into force in May 2018 defines such legal obligations related to data privacy and security on data processors. Industry experts are convinced that this regulation will lead to considerable financial and organisational efforts by data centres. How exactly particular cloud infrastructure providers will be putting their new obligations into practice remains, however, still unclear (Donnelly 2015).

Regulatory issues are relevant also in relation to the legal intricacies that frame Facebook’s data centre operation in Sweden. During the NSA (National Security Agency) scandal in 2012, Facebook was often discussed in the context of other American technology firms that had given state surveillance direct access to their servers. Accordingly, Swedish government representatives attending the official opening of Facebook’s Luleå data centre in summer 2013 eagerly confirmed the security of private data, claiming that Facebook’s facility was ‘under Swedish law’ (SVD Näringsliv 2013). My investigations reveal, however, that the data centre’s relation to national legislation is not that simple. This relation in fact epitomises a major complication which shapes the relation between nation-states and IT markets in general, and between global social networks such as Facebook and their local infrastructural sites in particular.

Facebook’s data centre operates as a covert subsidiary named Pinnacle Sweden. As an industrial site, the centre is under Swedish law and obliged to confirm to Swedish environmental and city planning regulations. When it comes to the data that are stored in and processed by that data centre, however, it is less clear whether Swedish law applies or not. In response to my inquiries, representatives of the Swedish Data Protection Authority (DPA)⁸ admitted to being unsure if an abuse of private user data related to Facebook’s Swedish data centre would be investigated by Swedish authorities at all. Since the company’s European headquarters are located in Ireland, such a case of abuse might be regarded as falling under Irish legislation. Furthermore, depending on the case, it would be necessary to clarify which of its aspects would be subject to national legislation, and which aspects fall under European legislation. The DPA admitted to not having been in contact with Facebook’s data centre, because from this institution’s perspective, the data centre is primarily a neutral data processor that does not bear direct responsibility for the data stored. In consequence, DPA representatives were not able to explain what happened with the more than 40 inquiries it received from Swedish citizens related to the possible role of the Luleå data centre in the NSA case, about which Swedish media had reported previously (Thorell 2013).⁹ DPA’s uncertainty regarding the boundaries of its own responsibility and possibilities of control illustrates the complicated relation between state territories and data streams which continuously cross those territories and their legal frames. When it comes to Luleå, these complications were obscured in political or public discussions during the data centre implementation process. In official communications of this new industry, future-oriented scenarios focussing on the assumed positive effects of locating cloud infrastructure in Luleå prevailed. In what follows, I will analyse those scenarios as scale-making strategies – that is, as experimental projects of ordering, relating, foregrounding and obscuring.

Configurations of the Swedish Cloud

The infrastructural intricacies described above reveal Facebook’s data centre project in Luleå as a site of agency and power where state-making, nation-branding and regional development strategies converge with digital economy trends and border-crossing technologies, a site where (national) centres and peripheries are contested and redefined, and where new subjectivities, forms of politics and future visions emerge. In the course of this process, various modes of the cloud’s infrastructural visibility come into being and its material forms and symbolic meanings are negotiated. Analysing these configurations as scale-making practices make visible how localities and globalities come into being and how new geographies of the Node Pole are made real (Tsing 2000: 348).

Corporate Disentanglement

The construction of Facebook’s vast data centre operation in Sweden’s North has avidly been followed by both global and local media, with IT experts publicly praising the data centre’s technological advancement, and environmental organisations such as Greenpeace welcoming Facebook’s turn to renewable energy use. Sweden’s newspapers have pictured the country as an ideal place for global business, while local journalists nurtured expectations regarding investment, new jobs and other positive effects of the Facebook project for Luleå and the Norrbotten region. Luleå’s officials have expressed delight over this exceptional publicity, and Facebook’s representatives frequently stressed the corporation’s commitment to its new environmental goals.

Contrary to such official statements, however, my observations in Luleå revealed that the visibility of local infrastructure was strategically avoided by Facebook. Designed as a ‘white label’ storage site that appears unrelated to Facebook’s brand and core operations, the data centre does not invite local collaborations. Facebook’s industrial research and software development are confined to its American headquarters, and public relations are strictly regulated through a PR agency located far away in Sweden’s capital, Stockholm. Public appearances of Facebook’s brand, such as an ice sculpture of the ‘Like’ thumbs-up in Luleå’s city centre, or a public event celebrating Facebook’s presence in the city, were initiated by the local municipality and business agency, not by Facebook. Although the data centre is featured in Luleå’s technical museum, and although Facebook runs a Community Support Program that finances cultural and educational initiatives, the IT giant is clearly less engaged locally than SSAB, for instance, Sweden’s largest producer of steel. As one of my interviewees related, ‘Facebook would never put its sign on an ice-hockey team, since Facebook already is a winner and does not need that kind of visibility, especially since a hockey team also might lose’ (Interview, September 2014).

My conversations with municipal employees as well as construction and other companies made it clear that Facebook had followed a strategy regarding the data centres’ public invisibility from the very start. Initial meetings with Facebook were organised without the press and did not even mention the company’s name. Only a few people in Luleå’s municipality and business agency knew about the ongoing negotiations at all. The company’s actual identity was kept secret until the very last minute, despite the involvement of architects, environmentalists, lawyers and other consultants in later planning phases that required considerable knowledge transfer and project information. As interviewees told me, an evocative working title – ‘Project Gold’ – and the high priority assigned to the project by Luleå’s municipal commissioner were the sole indications of the investments that were at stake during the data centre planning process. According to these respondents, not knowing the company’s name or the exact purpose of the project while still having to deliver substantial information, and privileging an anonymous customer, caused frustration and difficulties, especially in regard to environmental and building permits, core documents needed for industrial construction projects in Sweden.

The data centre’s restricted accessibility poses a difficulty even today, especially for journalists who are regularly denied direct contact with the data centre. All the subcontractors I contacted for interviews, including the local event agency that organised the data centre’s opening event, are required to ask for permission from Facebook’s American headquarters before communicating to anyone about issues related to the data centre’s implementation process. Some of these companies admitted they were uncomfortable with this obligation to secrecy as it hindered dissemination of information about their performance; somewhat ironically, the imposed secrecy even applied to information which could easily be found on the internet.

While limiting the visibility of its local infrastructure, Facebook simultaneously initiated the online Open Compute project in order to inform both the public and the IT community about new technologies and to invite contributions to their development.¹⁰ In a similar vein, Luleå data centre manager Joel Kjellgren highlighted Facebook’s global dimension while downplaying the role of individual data centre sites – as, for instance, during his talk at the DataCloud Nordic 2014 event where he presented statistics on global network use and on the geographical scope of Facebook’s operations, stressing the fact that even the loss of a mega data centre, like the one in Luleå, would not disrupt the network’s global functionality or even be noted by its users.

The contrast between Facebook’s efforts to highlight global virtual visibility while downplaying the role of local infrastructural sites is not due to security issues. Rather, infrastructural invisibility is key for promoting Facebook and the cloud as global entities freed from local materialities or frictions, as ‘an ostensible non-place obscuring agency and power inherent within it’ (Bowker et al. 2010: 6).

IT companies such as Facebook work towards a disentanglement – or at least controlled entanglement – of the cloud from its infrastructural localities, scaling the cloud as a global and virtual technological zone. Marketing strategies as those described above, for instance, make data centres invisible as material and local sites, with the aim of minimising friction and contestation – or community and environmental overflows (Appel 2012: 639) – which could disturb free data and profit flows. Such corporate strategies of scaling are nothing new. They very much resemble the logics of contemporary ‘zonal capitalism’ as described by Hannah Appel (2012), James Ferguson (2005) or Brenda Chalfin (2015) in their analysis of offshore oil companies’ modes of disentanglement from their local contexts.

Keeping the cloud virtual and global requires, however, continuous work since the functionality of the cloud depends on diversely situated and unstable alignments with multiple actors in local infrastructural sites (Ong 2007: 7). Facebook’s corporate secrecy and local invisibility in Luleå is not easily maintained. In the process of locating a data centre, the cloud’s modes of visibility and scales of presence are constantly renegotiated in historically and culturally specific ways. As Jensen and Winthereik note, the local does not necessarily operate as expected when it comes to connecting actors involved in infrastructure-making processes (2013: 12), which is why the corporate ‘global’ geographies of the cloud remain fragile and experimental.

National Embedding

While countries and regions compete over sites for data centre industry development, technology firms such as Facebook or Google also depend in many ways on the natural and social resources those regions may provide for meeting their infrastructural and energy needs. Luleå’s Facebook data centre is a key example of how the infrastructure of the global cloud integrates into national territories and narratives and plays a role in state-making processes. Northern Sweden’s cool climate and cheap hydro power are not the only reasons why this region is a perfect place for locating cloud infrastructure. Of equal importance is the IT industry’s integration in and dependency on already existing local infrastructures that were established in the course of the country’s industrial development over the last two centuries. Norrbotten counts as one of Sweden’s most technologically advanced regions, one that hosts a ‘technological mega-system’ (Hansson 2004): an extremely stable, multi-layered industrial infrastructure connecting forest, steel, mining, and other traditional industries with hydro power stations, railways and harbours, as well as large military defence facilities designed to withstand a nuclear explosion. These older infrastructures in the country’s Northern regions contributed to the economic foundations of the Swedish nation-state, laid the ground for Sweden’s self-image as an aspiring industrial nation and became essential in homogenising state-making practices (Sörlin 2002, 2013). Since Sweden’s political and academic elites conceived of the country’s Northern regions for a long time as empty landscapes without a history of their own, a function of earlier large-scale, industrial-infrastructural projects was to properly connect these remote, resource-rich and culturally diverse regions to nation-state territory, thereby transforming them into domestic colonies. This transformation consequently included the exploitation of natural resources and also a strategic disinterest towards the needs of the native Sami people. Accordingly, until recently only industrial infrastructural projects, scientific discoveries and technological innovations were included in the official version of Northern Sweden’s history (Harvey 2012). The building of Facebook’s data centre and the deployment of cloud infrastructure in Sweden’s North can also be regarded as state-making practices that connect the country’s technological mega-system to another global mega-system: the internet.

As a media event, Facebook’s construction and opening drew enormous national and global coverage, comparable only to the Swedish royal wedding. The Facebook project arguably has put Sweden ‘on the map’ of global IT competence, and was crucial for achieving the goal, defined in Sweden’s Digital Agenda, to make the country a ‘leading IT nation’ in the world (Government Offices of Sweden 2011). In accordance with this goal, national media have promoted Luleå’s data centre as providing a uniquely Swedish cloud: unique in being innovative, green, safe and trustworthy, as opposed to the allegedly dirty clouds over the United States and other parts of the world. Official presentations of the Facebook project were embellished with national symbols; the grand opening event of the data centre, for instance, was decorated in national colours, with photographs of ‘typical’ Swedish landscapes, and by the slogan ‘midsummer deluxe’ in reference to Sweden’s national holiday.

Such nationally framed representations contradict Facebook’s detached attitude towards infrastructural locality as they situate the Luleå data centre in relation to the global social network and even rescale the company’s infrastructural presence on Swedish territory as a form of national belonging that would imply commitment and obligation. Yet even if such representations – in their function as state-making and nation-branding instruments – efficiently link national territories to global IT economies and their markets, such emergent new geographies are far from frictionless or easy to maintain. What materialises in these ‘national’ geographies of the cloud, is the problematic role of the state which

is called upon to regulate the activities of corporate capital in the national interest at the same time as it is forced, also in the national interest, to create a ‘good business climate’ to act as an inducement to trans-national and global finance capital. (Harvey 1989: 170)

Through such nationally framed strategies of scale-making, the state’s territorial imperative is modified to fit the shape of the cloud. The regional dependencies on IT investitions and national authorities’ limited possibilities of control are obscured (clouded) and the national geography of the cloud is claimed by those authorities as being ‘under national law’ and regulation. However, the cloud rather represents a lightly regulated zone of ‘extrastatecraft’ where multiple forms of governance and legislation co-exist, and where state and non-state actors collaborate in search of a profitable win-win situation (Easterling 2014; Chalfin 2015). Furthermore, these national geographies do not remain consistent but are constantly renegotiated and rearranged from within by various local actors.

Regional Relocation

Infrastructures such as transport, tubes, or cable systems enable flows of goods, data, or water. However, these ‘infrastructural pneumatics’ may also include people, social or administrative entities, and cultural imaginaries related to broader processes of social change (Larkin 2013; Gregg 2014) as is the case with Facebook’s Swedish data centre.

In the process of implementing cloud infrastructure in Luleå, other regionally framed scales and geographies of the cloud became prominent beyond those already mentioned. Luleå’s data centre has turned into a symbol not only for national IT development, but also for the future of the city of Luleå and the Norrbotten region. As regional officials claimed, including the city’s municipal commissioner, the Facebook project not only introduced a new industry but it even brought a new emotional atmosphere and mindset to Luleå:

It is a similar effect to when we built the Culture House [Kulturhus] eight years ago. Back then we could change people’s minds and make them proud of their city. And now I think Luleå’s citizens are proud of Facebook. They [Facebook] are here, they are only here in Europe. They left America in order to go to Luleå. That’s great. (Interview, September 2014)

While it is obviously doubtful if such official statements represent any citizen’s attitude towards the new industry, the symbolic power of Facebook’s brand image here becomes instrumental for regional institutions and communities of actors who strive to redefine the city’s and region’s peripheral position. Facebook’s local presence is taken up and instrumentalised for research, business and cultural activities by The Node Pole, for instance, who promote the region to potential investors. LTU also makes use of Facebook in order to receive funding for developing IT-related study programmes and for establishing another data centre, SICS North Swedish ICT,¹¹ which will serve as a national institute for industrial and academic research on cloud computing. Even the municipality of Luleå uses Facebook’s presence as an opportunity for public celebrations and events, in order to foster its citizens’ agreement with regional development strategies, and to rebrand Luleå and Norrbotten as a hub of IT competence. One such event took place in winter 2013, when the city’s municipality and the local Business Agency organised 2500 locals into a gigantic human ‘Like’ in order to receive a mention in the Guinness Book of World Records and, consequently, to create higher visibility for investors and tourists.

In addition to these events, publications such as the book Hello Mister Mayor, Facebook is Calling (Vidén 2014), written by former municipal commissioner Karl Petersen, or the Report on the Effects of Facebook’s Establishment for Luleå Science Park as a Regional Node (Granberg 2014) contributed to interpret the data centre implementation process as a regional project – as ‘bringing Facebook to Luleå’ – for turning this process into a promotional narrative of locality. These publications described the knowledge and experience gathered in the process of establishing Facebook’s data centre as a regional competitive advantage. Indeed, newly acquired data centre knowledge directly affected local political and industrial communities by generating new expert elites.

In Luleå’s case, this elite is a small group of people working for the municipality and the local Business Agency. It includes independent consultants acting as brokers between regional and national institutions, on the one hand, and global business on the other. Keeping the ‘secret’ of Project Gold, these individuals took on the major task of communication and knowledge transfer. As those ‘elites’ would often state in interviews, they had never experienced a comparable loyalty and solidarity between Luleå’s business and municipal administration and between the municipality and Norrbotten’s County Administrative Board before or after working on this project. This experience of empowerment through solidarity for secrecy motivated those actors’ conviction that Facebook’s data centre even marked the beginning of a new industrial era, and a new period of regional development followed by social and cultural change: ‘After we’ve managed this [brought Facebook to Luleå], we believe that we are able to manage everything’ (Interview, May 2014). Indeed, also many of my other interlocutors – among them local academics, journalists, architects and environmental experts working in the administration or municipality – believe that the Facebook project initiated a broader social transformation in Norrbotten, freeing the region from its burdened image as a national periphery and rescaling it as an IT competence hub. Challenging private companies and state authorities to keep pace with a globally operating technology firm and to comply to its needs, this project forced change upon the rigidly regulated, old-fashioned political and organisational structures and even upon individuals:

If they [Facebook] have a question on Friday evening, they expect to get an answer on that very evening and not next Monday morning. Its business and if they don’t get what they want, they will just walk away. People are needed who are ready to work on weekends and holidays. (Interview, April 2014)

Statistical facts such as rising student admission numbers or real estate prices were often seen as evidence for cloud-related visions of regional social change. According to the widespread interpretation of such figures, the data centre industry would produce new mentalities, working cultures and subjectivities, thus transforming the burdened image of the former ‘steel city’ and the remote, predominantly working-class region. My interviewees described such new subjectivities in marked contrast to ‘traditional’ working-class Luleå citizens who, in turn, historically have been stereotyped in the figure of the masculine steel plant worker interested in just sports or hunting. IT employees, on the other hand, were imagined as culture- and knowledge-oriented and mobile given their allegedly frequent travels across the cloud’s infrastructural geographies, thus bringing transnational experience and lifestyle to the city. As an IT scholar from Luleå Technical University described the anticipated change:

When I was young I left this region because I was not interested in working in the mines with those dirty, stubborn men. […] Even the mining industry makes efforts to change the old-fashioned industrial image of Luleå. They would like to hire qualified people and if those experts cannot find jobs for their wives, or if there is only sports to do after work, then they say ‘no, thank you’. (Interview, October 2014)

The idea that data centres represent a ‘different’ industry may be related to the dominant perception of IT as a particularly ‘cultured’ technology, as described by Löfgren and Willim (2006). However, there are also specific historical and social reasons for such local expectations. Regional narratives produced by the local press, in regional development reports, and related by my interviewees assert that the city had managed to attract the American tech firm without the state’s help. This idea of regional independency from the state is an attractive narrative for local politicians, industry representatives and the broader community. The popularity of these visions is grounded not so much in the actors’ technological optimism as in Norrbotten’s and Luleå’s complicated historical relations to the Swedish state. Over the past centuries, national industrial projects have dominated the development of Northern Sweden with lasting effects on people’s everyday lives, and at least some of those industrial projects are still rooted in the city’s and region’s collective memory. This includes, for instance, family stories stretching back to the times when the dams for the water power stations in the Lule river were built, forcing many inhabitants who lived at the river’s shores to resettle elsewhere (Öhman 2016). Luleå citizens also remember the Stålverk 80 project, a governmental plan to build an enormous steel plant that dates back to the 1970s. Many then felt attracted to the city, and large areas were turned into building sites; streets were broadened, new living areas planned, and yet the project was never realised due to the oil crisis (Leijonhufvud 1976). Regional newspapers from that period (1973–1976) document the great expectations related to the steel plant project, and its sudden clash with reality. During the oil crisis, inhabitants of Norrbotten – the most resource-rich region in the country, equipped with the most stable energy grid and other infrastructures – experienced a scarcity of gas, electricity and even hot water in the middle of the harsh Nordic winter. Accordingly, local memories related to former national industrial projects primarily reveal traumatic experiences of being repeatedly dependent on state subsidies during such times of crisis. Several of my interlocutors who moved to Luleå as youngsters because their parents were hoping to get jobs in the new steel mill experienced those hopes and difficulties.

In such a historical context, the relocation of the city from the national periphery to the centre of the global cloud becomes particularly enticing, as do visions of the region as an independent player on the global IT marketplace. Data centres are expected to become more than just new factories exploiting local labour. In its dependence on local climate and cheap hydro power, cloud infrastructure rather is expected to lay the grounds for the very first industry in the region that would make local use of the region’s natural resources, instead of shipping those resources southwards, as had been the case with most of Luleå’s traditional industries. The relocation of Luleå from periphery to centre is also a topos highlighted in regional development strategies, such as The Development Strategy for a World Wide Leading Technological Region for Climate Smart and Effective Data Centers in Norrbotten (2014), and in media reports, and by renegotiating meanings of the cold. Given the growing data centre industry and its cooling needs, ice and cold – symbols of Northern Sweden’s remoteness and harsh climate – turn into assets, necessary to save the data of the world and to make the region profitable. Comparing Luleå’s low annual average temperature to that of the largest urban IT hubs such as Frankfurt, London or Amsterdam, representatives of The Node Pole proudly explained to me that ‘packaging’ the cold with low electricity prices and excellent regional infrastructural preconditions would result in a unique local product that can be sold on the global IT market, making Luleå a competitor to those hubs.

As Easterling notes, ‘fiber-optic cable buried in the ground gives land a new value much like a highway or railroad’ (2014: 12). In a similar vein, through the implementation of cloud infrastructure Luleå and its surrounding region attain new value. The city and region are put on the map as hubs within an infrastructural space of data processing, enabling regional actors to act and imagine themselves within new contexts and in relation to other than just national scales. While foregrounding regional interests and competencies, such scale-making strategies, however, obscure the less positive consequences of the data centre industry’s implementation. This includes, for instance, the very limited number of jobs available in this highly automated industry.

Regional scale-making practices and versions of the cloud’s infrastructural visibility exist in parallel and in (conflicting) relation to national and corporate ones. These practices not only illustrate how the border-crossing infrastructure of the cloud integrates into and forms social and industrial context locally, but also how the globe ‘as a region made large’ (Tsing 2000: 348) is brought into being.

Conclusion

The implementation of the data centre industry in Northern Sweden triggers social change and has spatial effects. In the process of cloud infrastructuring, territories are rescaled; new subjectivities and forms of governance emerge; and socio-technological entities such as the ‘global cloud’, or the Swedish ‘IT nation’, as well as competence hubs of Norrbotten and Luleå come into being and form translocal geographies of ‘The Node Pole’.

This article has demonstrated how these geographies are made relevant and real through diverse strategies of scaling. I have characterised such strategies as ‘clouding’, in reference to changing modes of the cloud’s visibility and to the practices of packaging, foregrounding and obscuring infrastructural materialities. As I have shown, Facebook tries to limit its own local visibility and to disentangle its infrastructures from local social contexts by framing the social network as merely virtual and global, and by downplaying the role of specific infrastructural sites, in order to ensure the most effective and frictionless flows of data and profit. At the same time, the Swedish state – as a stakeholder in cloud infrastructuring process – seeks to benefit from Facebook’s infrastructural presence on state territory by interpreting this presence as an obligation to national belonging. State institutions and representatives form the cloud as a national domain through nationally framed representations and official claims about the data centre’s legal commitments to Swedish law. The national geography of the cloud is also shaped by expanding the state beyond its actual territory, connecting the national infrastructural grid to the worldwide internet, and redirecting public and media attention towards Sweden as the world’s ‘leading IT nation’. However, the national cloud is not stable but constantly contested from within by regional and local actors. The data centre’s actual dependency on local resources and infrastructural materialities makes Facebook’s integration into Luleå’s industrial and natural environment unavoidable, and supports the city’s and region’s aspirations of sovereignty. By claiming Norrbotten’s and Luleå’s role as independent technological hubs within a global IT economy, local institutions and experts reimagine and reshape the region and the city as being part of the Node Pole, a border-crossing cloud infrastructural space which (at least partly) escapes national regulations and enables regional independency from the state. Through such local framings, the historically hierarchical relation between Sweden’s national centre and periphery is being contested, and Luleå becomes a ‘transterritorial “centre”’ (Sassen 2005: 37) of the data economy.

Being constantly negotiated, the geographies of the cloud are not stable, but dynamic and multi-scalar, composed of changing forms of governmentality and modes of regulation and sovereignty within which

the world capital and national capital can […] shadow each other, alternately exhibiting a regional cultural ethos, national pride or global ambition. State and non-state actors use each other as proxy or camouflage as they juggle and decouple from the law in order to create the most advantageous political or economic climate. (Easterling 2014: 27)

The Facebook data centre project seems to be particularly effective in this respect. It creates a win-win situation for politics and business and even integrates local aspirations for sovereignty. This motivates local actors to perceive and represent their region and city ‘as a service’ and to stay ‘in a permanent state of selling’, as my interlocutors describe their everyday work of infrastructuring Luleå as a global IT hub (Interviews; Granberg 2014; Vidén 2014). The collective aspiration for effectivity and regional relocation directs public attention away from problematic aspects of the new industry, such as its enormous energy and water needs, potential long-time environmental consequences, or the limited possibilities of regulation and control. Clouding thus also means obscuring ‘the invisible trouble’ that, as Star and Ruhleder state, is inherent to infrastructuring processes with their power inequalities, exclusions and unintended effects (Lampland & Star 2009). Uncovering these troubles sheds light on the fact that the emerging geographies of the cloud not only shape new centres and competence hubs, but also new peripheries: places to which infrastructural materialities and environmental effects of IT networks and services are now outsourced.

My ethnographical research of local infrastructuring processes and practices of scaling demonstrates that clouding is a logics of key importance for developing and maintaining the data flow and technological zones of digital capitalism. Such research not only investigates the politics of infrastructure – the forms of governance and political rationalities – but also takes into account the ‘sticky grip’ (Tsing 2005) of the local. It demonstrates how seemingly just virtual and global data clouds are entangled within local social and natural environments and how they connect with collective and individual aspirations and ways of knowing about the future.

Disclosure Statement

No potential conflict of interest was reported by the author.

Additional information

Funding

This work was supported by Riksbankens Jubileumsfond, The Swedish Foundation for Humanities and Social Sciences [grant number P13-0205:1].

Notes

1. https://www.youtube.com/watch?v=YwqtBJMZJIU (Accessed 12 June 2017).

2. http://thenodepole.com (Accessed 12 June 2017).

3. This sum was made available through European Regional Development Funds designated for boosting regional competition, modernisation and cross-border co-operation.

4. As the Head of Luleå’s City Planning Office stated in the interview, the city declined this requirement in the negotiations with Facebook lawyers, risking to lose the contract. The city’s position was accepted by the IT firm, however.

5. The Luleå steel plant is owned by Swedish steel processing company SSAB which has operated the plant since 1951.

6. In 2010, Greenpeace launched the Unfriend Coal Campaign, with Facebook users sending Facebook messages to the company urging it to change its energy policies. http://www.greenpeace.org/international/en/campaigns/climatechange/cool-it/ITs-carbon-footprint/Facebook/ (Accessed 12 June 2017).

7. https://www.eugdpr.org.

8. http://www.datainspektionen.se/in-english/ (Accessed 12 June 2017).

9. According to media reports, the Swedish Data Protection Authority promised to investigate those inquiries. The authority was not able to confirm, however, that this investigation actually had taken place. I was told that such inquiries were most probably decided to be unjustified and not taken up for further investigation.

10. http://www.opencompute.org (Accessed 12 June 2017).

11. https://www.sics.se/groups/sics-north-swedish-ict (Accessed 12 June 2017).