HOW DATA CENTERS HAVE COME TO MATTER: Governing the Spatial and Environmental Footprint of the ‘Digital Gateway to Europe’
The authors wish to thank the four IJURR referees who reviewed an earlier draft of this article for their detailed and constructive comments. The support and valuable insights provided by the various interview partners in the Amsterdam Metropolitan Area are greatly appreciated, as is the language editing by Joy Burrough.
Abstract
Data centers, the material backbone of smart cities, power the digital economy and advanced digital services. Metaphors of ‘the cloud’ and ‘cloud computing’ obscure the massive computing and storage infrastructures, the resource flows and the land uses they mediate. To date, smart city research and policies have been concerned less with the materiality of enabling data infrastructures than with the material effects of increased datafication and digitalization of urban services. Only recently has urban data infrastructures’ rapidly expanding spatial and environmental footprint pushed their materialities to the forefront of public and academic controversies. Building on recent research on cloud geographies and ecologies, this article traces the politicization and emerging regulation around data centers in the Amsterdam Metropolitan Area, the self-proclaimed ‘digital gateway to Europe’. Here, after years of unconditional political support and regulatory passivity, a cascade of policy reforms has been introduced to confine data center growth. Nevertheless, severe urban governance challenges remain in mitigating data centers’ massive electricity, resource and land demands, and in exploiting their residual heat. We thus advocate broader dialogue across the affected policy fields and broader publics about which political objectives merit prioritizing given the constraints of available electricity and land.
Introduction
Data centers are the physical infrastructures of cloud and Internet services, which include data processing, storage and dissemination. These enabling infrastructures provide advanced digital services upon which urban economies and urban living increasingly rely. Often located out of public sight in suburban or rural regions, these highly securitized facilities hide burgeoning material infrastructures for storing and processing increasing amounts of data on our emails, video uploads, Facebook posts and Google searches (Hogan, 2022). Real-time data processing, in particular, has exploded, enabling urban services for Uber and AirBnB, ‘smart city’ solutions for mobility, electricity grids and buildings, mass surveillance and monitoring, e-commerce and the implementation of the Internet of Things, artificial intelligence and virtual and augmented realities (ibid.). Many companies, non-profit organizations and administrations have radically outsourced their computing and data infrastructure to urban data centers that provide redundant data security, storage, power, cooling and other network services to their decentralized customers (Narayan, 2022).
Metaphors of ‘the cloud’ and ‘cloud computing’ obfuscate data centers’ materiality: their massive computing and storage infrastructures are powered by local utilities, diesel generators, battery banks and cooling systems connected to Internet Exchange Points through transcontinental and transoceanic fiberoptic cables. These infrastructures’ growing demand for energy, water, critical minerals and land—with 90% of data having been created in the last two years (Hogan, 2022: 273)—is attracting increasing concern in critical infrastructure debates and raises complex governance challenges. This is particularly true for the Amsterdam Metropolitan Area (MRA), the self-declared ‘digital gateway to Europe’. MRA's importance regarding Internet services and traffic is now global, comparable to that of the Port of Rotterdam for maritime shipping traffic or Schiphol for the aviation sector. Considerable investment in digitalization has made it an attractive location for data center clusters.
With average annual growth rates of 17% in the 2010s, the data center industry has become an important economic backbone and taxpayer in the MRA. For over 15 years, all levels of government in the Netherlands have unconditionally supported the digitalization and datafication of urban processes through ambitious smart city strategies, platforms and investment incentives, as evidenced by Amsterdam's multiple awards for being one of the most successful smart cities in Europe (Mora and Bolici, 2017). However, the hubris of digital innovation has become increasingly visible, and the region is wrestling with data centers’ growing energy, material and spatial footprint. Prior to a freeze on building permits imposed in 2019, developers could not be refused a building permit in business zones even though electricity grids in the MRA were under severe strain. Hence, municipalities and grid operators in the region expressed concern that data centers’ demands on the local power grids and land requirements were spiraling out of control. In 2019 the Amsterdam and adjacent Haarlemmermeer municipalities therefore announced a moratorium on data center construction and called for stricter regulations (Gemeente Amsterdam, 2019). However, as we show in our case study of Amsterdam, effective regulations to govern the energy (re)use and land-use planning of international private tech companies remain a major challenge.
To date, urban studies have been concerned less with the materiality of the enabling infrastructure than with the material effects of the widescale digitalization and datafication in the emerging ‘digital’, ‘smart’, ‘computational’, ‘platform’ city. Accordingly, urban scholarship has focused on the social, economic and political implications of smart cities or of platform urbanism, how they transform other networked infrastructures and urban services, change the logics and techniques of urbanization processes and urban planning, and create new forms of digital capitalism and dominance by large tech companies (Pickren, 2018; Caprotti and Liu, 2020; Luque-Ayala and Marvin, 2020). Only recently, and based on earlier work (e.g. Rutherford, 2011), has a growing strand of research redirected attention to the geographies and ecologies of data centers (e.g. Diguet et al., 2019; Brodie, 2020; Munn, 2020a; Furlong, 2021; Libertson et al., 2021; Ortar et al., 2022; Turnbull et al., 2023). However, much of this ‘material turn’ in data center studies research is concerned with the hyperscale data centers of the largest global tech companies in rural or periurban areas and largely neglects the more mundane colocation data centers that enable the increasingly common practice of many companies, universities, government agencies, etc., to outsource their servers and networking services to urban colocation data centers. To date, empirical urban research on data centers' ecologies and spatialities has been scarce, and the land-use and environmental politics tethered to their material infrastructures remain occluded. Yet data center infrastructures and their material metabolism are not neutral; they are developed and built through inherently sociopolitical processes, implying power struggles, and winners and losers from processes of infrastructural reconfiguration. Further research is therefore needed on the urban politics of data center growth and the policy responses to their spatial and environmental externalities.
In this article, we explore the materialities of digital infrastructure and their impact on land use and energy infrastructure in the MRA. More specifically, we aim to unearth how digital infrastructures’ ecological and spatial footprint has long been obfuscated but is being increasingly politicized. Building on a literature review, we highlight gaps in research on smart cities, data centers’ geographies and ecologies, and the governance of data infrastructures. In our empirical case study, we first explore how the data center industry emerged in Amsterdam below the radar of public control and then discuss the mounting controversies and policy responses that resulted in a data center moratorium. We examine how governance of data center planning has evolved since the moratorium and conclude by arguing that, although the increasing politicization of data centers through new regulations has increased these digital infrastructures’ visibility, daunting governance challenges remain. Additionally, by elucidating the materiality and geographies of data centers, we question the value propositions of smart cities and tech-enabled urban solutions.
Our research approach is qualitative. As we could hardly build on secondary empirical studies (exceptions are Stratix, 2018; Masson et al., 2022), we analyzed a broad range of documents, including reports and policy papers issued by the national, provincial and municipal governments, annual reports and publications of the Dutch Data Center Association. To trace the politicization of data centers in the MRA and north Holland, we conducted a content analysis of newspaper coverage of discourses on data center development in the region. When necessary, the policy documents and news articles were translated using DeepL translation software.
Between fall 2021 and 2023, semi-structured in-depth expert interviews were conducted with 16 persons representing the most relevant stakeholders in the MRA data center ecosystem: municipal and provincial governments (n = 3), researchers and consultancies (n = 4), data center logistics and operations (n = 4), data center associations (n = 2), energy and water utilities (n = 2) and real estate experts (n = 1). To select interviewees, we first mapped the stakeholders relevant to the operation of data centers, their resource and land use and their governance. Participatory observations at an online panel discussion ‘Data Centres: Taking the Bitter with the Sweet’, hosted by the Amsterdam Smart City initiative in 2021, brought us into initial contact with some key stakeholders. From here, we used a snowballing method and online research to come to our full list of interviewees. We developed a semi-structured guide for our interviews, focusing mainly on stakeholders’ perspectives on the regional development of the data center industry in the MRA and its land and resource use, their views on the increasing politicization of data centers, their assessment of national, provincial and municipal policies adopted to mitigate the infrastructural, spatial and environmental impacts of data centers and what they saw as the biggest challenges and opportunities for governing data centers. The interviews were then transcribed verbatim and coded manually using inductive and deductive methods. Only four interviews were in-person, the remainder were conducted online because of Covid-19 pandemic restrictions. Complementary to the expert interviews, participation at the national data center days in 2022 and 2023 allowed us to observe data center facilities and ask additional questions regarding the business models, operational requirements and regulatory frameworks of data centers. Spatially, our empirical study focuses primarily on Amsterdam and Haarlemmermeer municipalities, which have experienced the greatest influx of data centers.
Governing the material infrastructures of smart cities
Datafication of cities: ICTs, smart cities and ubiquitous computing
Like many of the shifts in urban development during the twentieth century, technology is seen as the catalyst for ‘smart’ urbanization. Worldwide, cities increasingly rely on information and communication technologies (ICTs) to solve urban problems: from balancing energy supply with demand and mitigating transportation congestion, to enabling citizen participation, renewing planning procedures and surveilling urban processes. To compete in the race for digital investment, urban governments embrace data-driven and algorithmic planning, promote smart city strategies, and develop locational strategies to attract ICT companies. Often, promoting ICT investment has been discussed as an uncontested technological fix for boosting economic competitiveness and fostering social welfare, political decision-making and environmental sustainability.
Since the mid-1990s, research in geography, urban planning and science and technology studies on ICTs has articulated growing interest in and concern about the impacts of computing and electronic devices on city development and urban infrastructure (Graham and Marvin, 1996; Rutherford, 2011). Scholars have critically examined how digital technologies and computational logics become intertwined with everyday life, infrastructure networks and the wider sociopolitical context of their deployment. Challenging a techno-optimistic reading of the smart city paradigm, smart urbanization has been criticized as a corporate-driven strategy dominated by large tech giants promoting neoliberal agendas (e.g. Söderström et al., 2014, Das, 2020; Guma and Monstadt, 2021), for its ‘technofix’ approach and ‘digital solutionism’ (Kuntsman and Rattle, 2019; Turnbull et al., 2023), celebrating technology and automation as neutral tools for solving complex socio-ecological problems (Kitchin, 2013). Other studies engage with large corporations’ data politics and their largely unregulated collection and economic exploitation of data produced in cities. Often governments lack the specialized technical expertise and institutional capacity required to define effective rules for data sharing and use, data privacy, data trust and data security (e.g. Acuto et al., 2019, Micheli et al., 2020).
Moreover, scholars have problematized smart city initiatives as increasingly evolving on the periphery of traditional planning departments and being led by economic development, innovation and technical standards agencies (Joss, 2018: xvii). Finally, the rise of data-driven, algorithmic planning and urban platforms and operating systems has sparked concerns and critical debates about the covert politics of data creation, management, ownership, transparency and access that stratify urban life (Luque-Ayala and Marvin, 2020; Safransky, 2020). Debates have problematized how data-driven analytics and computational logics are depoliticizing planning processes by advancing particular epistemologies of the urban (Luque-Ayala and Marvin, 2020). For better or worse, widespread digitalization and datafication are transforming everyday living and changing the fabric of urban spaces worldwide. The fast-paced advances in smartphones, social media, cloud storage, automation, artificial intelligence and the Internet of Things, and the sudden shift to remote work and e-commerce during the Covid-19 pandemic have accelerated data traffic growth.
Much research on the opportunities and pitfalls of ICTs, and on big data and ubiquitous computing for developing and planning cities implicitly treats digital technologies and data flows as immaterial while overlooking the significant material resources required to sustain computing and data infrastructures. As Pickren (2018: 26) argues, for a long time, researchers ‘have largely eschewed a deeper investigation into the production of the networked infrastructures that make these new ways of living possible in the first place’.
Rematerializing the cloud: geographies and ecologies of data centers
- Edge data centers are smaller, decentralized facilities providing computing and storage near where data is generated and used. Recent studies show that the industry is moving towards processing data closer to the source because transmitting increasing amounts of real-time data to centralized data centers is becoming unviable (Munn, 2020b).
- Hyperscale data centers, which are generally operated by the world's largest tech companies, including Microsoft, Amazon, Meta, Google and Apple, can, in principle, be located anywhere. However, due to their scale and size, their data storage facilities are often built in more rural locations with preferably some of the following characteristics: cool climate (reduces the need for cooling), sufficient water for cooling, and sufficient connection to fiber optic and high-voltage electricity networks.
- Colocation or multi-tenant data centers typically have multiple clients, each renting a portion of the server space. These are usually companies, universities, government agencies, etc., that outsource their server space to a data center that provides all the necessary connectivity, energy, cooling, data security and space for them to install their own servers. Unlike hyperscale data centers, colocation facilities tend to be in (sub)urban areas near robust telecommunications infrastructure, with connectivity to undersea cables and Internet exchange points (DDA, 2020). Access to an affordable and stable electricity supply is crucial. Since data is increasingly being exchanged directly between data centers, their proximity contributes to the stability and high speeds of data traffic. Urban agglomerations such as the MRA thus provide locational advantages for hyperconnectivity clusters of colocation data centers (Masson et al., 2022).
Data centers’ immense resource use and emissions have led to their major environmental and spatial repercussions becoming a key concern in data center studies. First, their equipment and digital technologies generally require a wide and expanding range of materials, some of which are critical raw minerals (Al Kez et al., 2022)—often mined, recycled and disposed of under precarious labor and environmental conditions in the global South. These technologies are periodically decommissioned and disposed of when technical warranties expire, and units’ performance falls below high reliability and redundancy standards. Despite ongoing efforts to extend the lifespan of units and repair and recycle old equipment, only a small share of this e-waste is recycled (ibid.). Moreover, data centers can become obsolescent or be shut down and abandoned. As recent studies show, their afterlives can cause pollution, waste or deserted buildings (Brodie and Velkova, 2021). Second, many data centers require huge volumes of cooling water, which is then discharged into water bodies and impacts biodiversity (Ristic et al., 2015). Third, data centers require megawatts of electricity to cool equipment and power facilities and servers. Cooling systems often represent 50% of data centers’ electricity consumption to counteract the heat generated by the servers and computing equipment (Diguet et al., 2019). Any loss of power can damage equipment, severely compromise data center operations and lead to significant financial loss. To ensure continuous energy for data communication, facility cooling and backup servers, data centers are designed to protect themselves from power outages or other eventualities, using uninterruptable power supply batteries together with diesel generators as emergency power sources if the power grid fails (Munn, 2020a). While ‘geo-replicating’ or ‘mirroring’ data in multiple locations and doubling energy storage and cooling equipment provide greater redundancy against local outages (ibid.; Farmer, 2023), it increases their material footprint. Fourth, data centers require affordable land, both for their warehouse-sized buildings and, equally importantly, for the infrastructures fueling their operation, for example, sites of electricity generation or use of residual heat, transformer stations, subterranean telecommunications, electricity, heat and water networks, or waste disposal sites. Finally, the material consequences of urban datafication also involve the associated technologies of data use and generation, such as smartphones, smart vehicles or electricity meters, or novel sociotechnical practices of food delivery, automated driving etc. (cf. Caprotti and Liu, 2020).
Despite the increasing concern about data centers’ materialities and the attention given to infrastructural solutions to their increasing resource and land use (e.g. Wahlroos et al., 2018; Al Kez et al., 2022), research on how data centers’ ecologies and spatialities are embedded in and co-evolving with urban political economies and on the governance of the enabling material infrastructure of datafication remains patchy and little advanced.
Governing urban data centers
For a long time, governance and planning debates have primarily focused on the opportunities and pitfalls of proliferating digital technologies in surveilling and governing social, economic and ecological processes in cities and facilitating ‘smart’ policy interventions in real time (see above; for an overview: Turnbull et al., 2023). However, in recent years, scholars have stressed the urgency of developing more effective regulatory frameworks and planning approaches to mitigate the data center industry's extractive practices. Backed by global capital and creating limited job opportunities, the industry threatens to bleed into and displace local interests, Farmer (2023) argues, as it competes for electricity, water and land with local communities (see also Libertson et al., 2021). One of the problems is that marketing schemes and limited transparency obscure data centers’ materiality. Hence, they operate in a regulatory blind spot beyond the remit of cybersecurity, land use and energy policy frameworks (ibid.). Pickren (2018) and Diguet et al. (2019) point out that data centers’ operations are, furthermore, disguised and remain below the radar of urban policy due to their location in suburban business zones or in converted industrial buildings. Holt and Vonderau (2015: 74) note that, although cloud companies use marketing schemes to demonstrate commitment to climate mitigation, environmental consciousness and transparency, they often ‘work to conceal or obscure less picturesque dimensions of cloud infrastructure’ (ibid.). Brodie (2020) argues that data center operators in Ireland benefited from generous tax and business deals with governments to entice tech investment but are concomitantly given free rein in their spatial and environmental practices. Bridges (2024) points to a disconnect between the sociotechnical imaginaries of digitalization strategies and the materially constrained capacities underpinning these strategies, including competing demands for electricity, water and land. Amoore (2018) suggests that the opaqueness of cloud infrastructure has allowed the ICT industry to bypass legal and bureaucratic regulations. One political hurdle is thus to ‘wrest the cloud back into a form over which one can have oversight … to make it comprehensible and accountable in democratic forums, and to render the cloud bureaucratically and juridically intelligible’ (ibid.: 9). However, data centers have received limited attention from urban and regional planners, and governmental oversight and regulations lag far behind the dynamic data center development (Amoore, 2018; Diguet et al. 2019; Libertson et al., 2021: Farmer, 2023; Bridges, 2024).
To summarize, thanks to recent debates, the focus has broadened from the urban impacts of digitalization toward empirical scrutiny of the materialities of the Internet. Whereas recent research on cloud geographies and ecologies has increasingly engaged with how data centers operate in specific urban contexts and how they impact resource and land use, empirical studies on the policy and planning responses to the extractive practices of the data center industry are rare. We thus argue that a priority is to reterritorialize the abstract cloud as an intelligible and governable infrastructure that provides critical services but whose negative externalities require closer scrutiny.
Based on the research gaps outlined above, our subsequent case study provides a brief overview of data center development in the Amsterdam Metropolitan Area (MRA) and the politicization of data centers’ spatial, infrastructural and ecological impacts. Building on this, we introduce into the municipal and provincial policy innovations to confine data center growth in the MRA and problematize the urban governance challenges in keeping those impacts at bay.
The promise of the ‘digital gateway to Europe’: data center networks in the Amsterdam Metropolitan Area
In 1988, thanks to the existence of a cable permitting data exchange between Dutch and North American laboratories, Amsterdam's Science Park (home to the National Research Institute for Mathematics and Computer Science) became the first European location to be connected to the Internet. This not only marked the start of the European Internet but also symbolized Amsterdam's pioneering role in Internet developments. Thus, ‘.NL’ became the first root domain outside the United States (Nevejan and Badenoch, 2014). In 1989 Stichting NLnet, the first independent provider in the Netherlands, was established. It offered affordable local dial-in and ISDN infrastructure for the entire country and helped Amsterdam become the best-connected exchange point for European Internet traffic. As a catalyst for regional growth, the establishment of the Amsterdam Internet Exchange (AMS-IX) in 1994 propelled Amsterdam's development into a small but growing network of national, international cross-continental Internet connections, triggering an exponential increase in the need for data processing, storage and network capacity (Masson et al., 2022). In addition to providing landing points for new international undersea cables, an important locational advantage was the liberalization of the Dutch telecom market six months earlier than the rest of Continental Europe. The presence of the AMS-IX, Amsterdam's dynamic digital culture with several research institutes, skilled labor and strong political support created flywheel effects, resulting in geographical concentrations of ICT-related or ICT-facilitated activity in the MRA: from 1996 various new international players with network operations and data center facilities settled in Amsterdam, thereby creating demand for further expansion of digital infrastructures (Stratix, 2018; Masson et al., 2022).
The first bottleneck in the electricity supply in the Amsterdam region emerged in 1999 (Stratix, 2018). Data centers’ projected demand for electrical power rose to 600 MW and could not be accommodated without strengthening the high-voltage grid. A first moratorium on new data centers was announced in the MRA, which initially led to investments being shifted to other regions. However, the Dotcom crash in 2000, the collapse of the telecom market in 2001, and the bankruptcy of several data centers slowed market growth and eased the power bottlenecks for a few years (ibid.), so the moratorium was not implemented.
Driven by ICT developments such as the spread of smartphones and cloud computing, the number of data centers increased sharply from 2005. Figure 1 shows the current situation. Three major hyperscale data centers were built, one being Google's center in Groningen near the landing point of an undersea cable and offshore wind parks. In 2014, Microsoft built a data center campus in Hollands Kroon, 25 km north of the MRA, soon followed by Google. Nearby, a fourth 50-hectare Microsoft hyperscale center was approved and is currently under construction after a protracted permitting process involving legal disputes between the municipal and provincial governments and fierce opposition from residents and farmers. Investment plans for hyperscale data centers within the MRA—one of 166 hectares for Meta near Almere, another instigated in De Kwakel by an anonymous investor—were approved by the respective municipalities. However, local protests led the Dutch Senate to embargo the sale of state-owned land, and the province objected to the plans. Despite this check to investments in hyperscale centers in the MRA, the region developed into a hyperconnectivity cluster of colocation centers which benefited from their proximity for data traffic exchange via direct connections between them, the proximity of large Internet exchange points (Masson et al., 2022), and access to fiber optic cables. Additionally, the lower price of Dutch electricity compared with prices in European cities, such as Frankfurt and London, makes the MRA attractive to international colocation companies (Operations Managers, Iron Mountain, interviewed on 1 December 2021).
In the first wave of data center growth, it was standard practice to convert existing warehouses and industrial buildings and to collaborate with real estate firms to find space and negotiate leases (Account Manager, TenneT, interviewed on 2 March 2022). More recently, however, several dedicated colocation data centers have been designed and built, with Equinix's windowless 70m high tower and Digital Realty's 11-story, 72m tall building in the Amsterdam Science Park being two examples (Stratix, 2018; see also Figures 2 and 3).
As the exponential growth of data continues, the MRA has consolidated its position as one of the key nodes for colocation data center operations in Europe. In 2022 it hosted 71% of the multi-tenant market in the Netherlands (DDA, 2023), with 55 colocation facilities plus 48 data centers. Colocation data centers are clustered in five campuses: Amsterdam Science Park, Amsterdam West, South-East, Schiphol Rijk and Almere (see Figure 1). A few colocation data center companies are slightly outside the hyperconnectivity zone in areas where land prices are lower because their customers’ requirements for latency are less strict (Executive, Sustainable Digital Infrastructure Association, interviewed on 20 January 2022).
Data center infrastructures in Amsterdam have become a symbol of the country's advanced digital basis and digital ambitions. National, provincial and municipal policies have highlighted Amsterdam's strategic international role as the digital gateway to Europe and the city's and country's commitment to sustaining the digital economy and digital futures:Once you have a certain number of customers, we consider our data centers to be kind of the marketplace. Also, for customers to interact with each other. They all need each other. That's how clusters of data centers start to develop over the years, and why it also makes sense for us to build our next data center not somewhere in a remote area in north Holland, but in proximity to the existing data centers in the region of Amsterdam (Manager, Interxion, interviewed on 24 March 2022).
As the economic policy advisor argues, data centers have for a long time been associated with ideas of progress, modernity, and sustainability (ibid.). They have thus been supported unconditionally by local governments and business associations.If you don't have that infrastructure, you don't have the Internet with the speed we have right here. You don't have the creative ecosystem doing all kinds of smart stuff, good stuff for the city and the future of the city. We wouldn't have the position we have here in the sense of livability of the city or the talents wanting to be and live here (Deputy Head, Economic Policy Advisor, City of Amsterdam, interviewed on 21 February 2022).
The end of the data h(e)aven? The increasing politicization of Amsterdam's data centers and first policy reforms
For nearly a decade, data center growth was largely unnoticed by the broader public and below the radar of environmental and spatial planning authorities. But, since the late 2010s, discourses framing data centers as positive economic forces stimulating enhanced social welfare, sustainability, and ‘smartness’ have increasingly encountered fierce criticism from local citizens’ initiatives, environmental groups, energy and water utilities, farmers and spatial planning authorities. Even though the materiality of datafication processes was spatially and discursively concealed by ‘hiding’ data centers in inconspicuous buildings with high-security fences and intensive surveillance in suburban peripheries, and foregrounding their enabling function, data centers—particularly the large hyperscale ones—moved to the center of public controversy.
Notwithstanding these developments, the Dutch government has repeatedly emphasized that the planning of data centers needs to be organized at municipal or provincial level (Masson et al., 2022: 44). This policy has created friction, since data centers usually fall into a relatively low environmental category (‘category 2’) under existing zoning regulations, due to their limited immediate environmental impact in terms of their emission of pollutants, noise, odor, or light. Consequently, municipal governments have been unable to refuse an environmental permit in a commercial zone if data centers complied with building regulations (van Oosten et al., 2021).
However, the operations of hyperscale data centers north of the MRA and plans by Microsoft, Meta, and one anonymous investor to build new ones have raised considerable resistance among residents acting in a coalition with electricity and water utilities, local farmers, local newspapers and politicians (see Masson et al., 2022; van Es et al., 2023). Hyperscale data centers have not only spawned local NIMBY conflicts typical of proposed large infrastructures, they have been framed as ‘water- and electricity-guzzling’ infrastructures exacerbating water shortages during hot summers, increasing the need for large-scale wind parks, and degrading valuable farmland, while generating little employment and regional economic added value because they serve Internet users abroad (Masson et al., 2022; Ekker 2020; Vuijk, 2022).
In 2018, concerns about electricity grid capacity across the MRA spurred investigations into resource use by colocation data centers. An article in a prominent Dutch newspaper accused important stakeholders such as the Ministry of Economic Affairs and Climate, the Central Bureau of Statistics and the Netherlands Environmental Assessment Agency of not monitoring data centers’ energy consumption (Bakkeren, 2019). Even Alliander, the umbrella company of Amsterdam's distribution grid operator Liander, reported being ‘regularly surprised’ by requests for power at a certain location. In a press release, Alliander noted, ‘the regular spare capacity at transformer stations is usually exhausted after the connection of one average data center’ (Alliander, 2019, own translation).
The national government has responded to increasing concern about industry needs and energy bottlenecks with a national data center strategy involving a partnership between three ministries, several provincial and municipal governments, regional economic boards and the Dutch Data Center Association (DDA). The most important framework is the spatial strategy for data centers launched in 2019: a roadmap for the growth of data centers that specifies the economic development strategy to strengthen the international competitiveness of five urban regions in the Netherlands through targeted expansion of the digital infrastructure (BZK, 2019). The spatial strategies for colocation data centers lay the foundation for the expansion of colocation clusters in the MRA and for an evaluation of additional opportunities in Zuid-Holland province.
Soon after, the National Strategy on Spatial Planning and the Environment (NOVI)—part of the new Environmental Act (Omgevingswet) that came into force in January 2024—outlined the spatial and environmental planning responsibilities of different levels of government. The recommendations on data center spatial strategy, based largely on the roadmap, highlight that data center growth should be clustered where: (1) energy demand can be sustainably met via current or future electricity grids; (2) residual heat from data centers can be recovered and fed into district heating networks; and (3) market players’ requirements for digital connectivity can be met (van Oosten et al., 2021). Additionally, the NOVI stressed that local governments are mainly responsible for spatially integrating data centers (BZK, 2020).
Despite the NOVI and roadmap guidance on locational requirements for data center growth, electricity providers and municipalities in the MRA were concerned about the national government's reluctance to adopt more restrictive regulatory frameworks. Given the considerable spatial concentration of data centers, the key concern was that the capacity of existing electricity grids and green electricity generation would not be able to meet data centers’ rocketing demands and simultaneously accommodate the rapid electrification of mobility and heating, the increasingly prevalent use of air conditioning in summer, and the MRA's ambition to build 250,000 homes by 2040 (Hut, 2021). Exponential growth in the number and size of data centers caused electricity demand in Amsterdam and Haarlemmermeer alone to increase by 65% between 2017 and 2019 (CBS, 2021). In 2021, the electricity consumption of data centers in Noord-Holland was equivalent to the average consumption of approximately 630,000 households—a demand that is expected to increase fivefold within a decade (Alliander, 2021).
Energy grids’ logistics and spatial requirements make grid operation complex. The construction of a new Liander substation in the MRA can take between five and seven years (Gemeente Amsterdam and Liander, 2021) and requires an average of 3,200m2 of land. The combined width of the cables leaving the substation is about 27m, requiring below-ground spatial evaluations (Gemeente Amsterdam et al., 2022), but subterranean space for future grid expansion in Amsterdam is scarce. Thus, planning and approval procedures for network extensions often take several years, partly because they have also to be coordinated with other network operators to avoid recurrent construction works (ibid.; Regional Lead, Liander, interviewed on 16 March 2022). Furthermore, grid operators must ensure data centers’ non-discriminatory access to their grids. While there is great need for investment in grids, their operators must recover their investment in grid extension to avoid being financially penalized by the Authority for Consumers and Markets (ACM) for faulty investment decisions. The significant demand for electricity, together with regulatory and market uncertainties restricting long-term data center planning, translates into major challenges for grid operators:If there is a data center which asks for 80 MVA, the whole transformer is 100% for one customer. So, does the public have to pay for the transformer? (Regional Lead, Liander, interviewed on 16 March 2022).
There was little scope for municipal governments and utility companies in the MRA to challenge data center development. They lacked effective zoning regulations or policy instruments to control the location of data centers and their energy use, ‘because they almost always fit in with the zoning plans and TenneT and Liander are obliged to supply electricity’ (Gemeente Amsterdam, 2019, own translation). Responding to increasing local concerns, Amsterdam and Haarlemmermeer municipalities announced a one-year moratorium in June 2019 to give local governments more time to consider data center construction, reevaluate the bottlenecks on the energy grid, and plan a data center sustainability strategy.They [the ACM] are monitoring our investments. And we are being benchmarked with all the other grid operators. And if we invest millions, and the data centers won't come, then we get a financial penalty. It's really the way the government watches us that forces us to adopt a defensive investment strategy. And if we aren't sure the benefits will come from the investment, this is logical because it's not our own money, it's money from the people … That holds us back, for now, from investing really far upfront. The real question of data centers is a question of power. So, for data centers, we basically just wait till they start asking (Regional Lead, Liander, interviewed on 16 March 2022).
Thereafter, Amsterdam and Haarlemmermeer municipalities and Noord-Holland province adopted a raft of policies to control future data center growth and align it with their sustainability ambitions (Gemeente Amsterdam, 2020a; Gemeente Haarlemmermeer, 2020; Provincie Noord-Holland, 2021). Whereas the provincial strategy focused mainly on hyperscale data centers in the north of the province, the spatial clustering of data centers and voluntary sustainability initiatives, the two municipalities introduced multiple requirements to confine the future growth of data centers and their energy and water use and to improve their architecture: new energy regulations limited the annual growth of energy demand to 137 MVA in Amsterdam and Haarlemmermeer municipalities and stipulated that spatial clustering of data center growth should be in line with the available grid capacity. Henceforth, new colocation facilities with electricity demands exceeding 80 MVA must construct their own 150 kV purchase stations and connect directly to TenneT's transmission grid to reduce bottlenecks in distribution. Long-term power purchase agreements should stimulate investment in large-scale renewable energy, and new data center buildings should have solar roofs and façades. Additionally, the reuse and storage of new data centers’ residual heat were made mandatory, and in emergency power outages data center operators are required to use non-fossil fuel sources, not diesel generators. Furthermore, the amount of drinking water required for cooling water should be drastically limited, and non-potable water sources or air-based cooling systems be explored. Finally, future data centers should be multi-story buildings with multifunctional use of their ground floors to reduce their land consumption, improve their architectural quality, and facilitate mixed-use neighborhoods (ibid.). Although existing data centers remain largely unaffected by these post-moratorium regulations in both municipalities, they provide more concrete guidance for future growth.
Complementing the two municipalities’ initiatives, which go far beyond the province's policy, the Amsterdam Economic Board spearheaded a large partnership with multiple private and public stakeholders called the Lower Energy Acceleration Program to explore options for increasing energy efficiency and regulatory reforms at the national and European levels. The City of Amsterdam and its grid operators Liander and TenneT have released a spatial framework for the expansion of electricity infrastructures until 2035, including the planning of multiple substations and grid expansions, based on a projected tripling or quadrupling of electricity demand in Amsterdam by 2050 (Gemeente Amsterdam et al., 2022).
The data center policies of the municipalities of Amsterdam and Haarlemmermeer brought more clarity and identified locations for possible data center development. However, an unintended effect of the land-use restrictions was that real estate prices skyrocketed. While land intended for data center development was bought in 2017 for a price between €200 and €400/m2, square meter prices rose to between €1500 and €1700/m2 in 2022/23 (CBRE, 2023: 13). In comparison, land for logistics and industrial developments has yielded an average price of €29/m2 (ibid.).
In late 2023 the municipality of Amsterdam introduced another moratorium on further data center growth. Despite fierce opposition from the data center association, it declared that it would no longer approve the construction of new data centers within its boundaries unless they directly benefited Amsterdam and met strict environmental standards.
Governance challenges and the politics of urban data center growth in Amsterdam
Given data centers’ increasingly evident ecological and spatial footprint, municipalities in the MRA—and, to some degree, the provincial government—are among international pioneers in curtailing uncontrolled data center growth. While the data center industry has developed security standards and classification systems for its industrial stakeholders, our analysis shows that local governments remain ill-equipped to keep abreast of dynamic technological and market changes and their spatial and environmental implications. Despite progressive policy interventions after the moratorium, all active data centers in the MRA anticipate growth or even strong growth (DDA, 2023). Thus, considerable governance shortfalls and challenges remain.
Decision-making based on incomplete information and metrics
To stay competitive, data center operators protect information on data flows, energy and resource use. Information on their environmental performance is, consequently, not readily available to local governments. Simultaneously, reliable data on employment opportunities or other economic benefits derived from data centers is often lacking or contested (Masson et al., 2022). Council members in the MRA regularly complain about inadequate access to information, and political representatives and citizens express annoyance with the secrecy surrounding investment by tech companies that thereby avoid public scrutiny of their plans (ibid.). In its first report on data center energy use, the Netherlands Central Bureau of Statistics noted that ‘the exact capacity, efficiency, and capacity utilization rate of most data centers are unknown’ (CBS, 2021). Interviewees confirmed that it is difficult to get specific information from data centers on future development plans and power and water usage.1 Simultaneously, data center operators criticize governments’ failure to keep pace with ICT innovations and their implementation of regressive policies (Strategic Advisor, Lumen Technologies, interviewed on 1 December 2021).
Data center operators and industry representatives insinuate that some regulations (including moratoriums) and accusations of resource use are guided by false information or political motives. For example, data center operators and DDA representatives argue that criticism of data centers’ spatial and energy footprint is often disproportionate and ignores their criticality for digital services (Employee in Communications, Dutch Data Center Association, interviewed on 18 January 2022). They claim that the industry is a frontrunner in supporting the energy transition and renewable energy investment (Strategic Advisor, Lumen Technologies, interviewed on 1 December 2021; DDA, 2023).
Another interviewee stressed that lack of information had made effective interventions contested:So, the PUE, for example, has different categories. And at this moment, nobody is saying, okay, I am using category one, two or three … so you don't know which category. So, it needs to be more transparent (Program Manager, Equinix, interviewed on 22 March 2022).
The opacity of data centers’ environmental performance means local planners have to make decisions based on incomplete information.When looking at planning and when looking at the sustainability impact of data centers, the current level of insight and transparency that decision-makers, that regulators, that politicians have, is falling absolutely short of what is needed to be able to make an informed decision. At the moment, we as a society cannot properly assess the carbon or environmental impact’ (Executive, Sustainable Digital Infrastructure Association, interviewed on 20 January 2022).
Fragmented planning procedures for electricity grids and data centers
Overcoming bottlenecks in renewable electricity generation, power grids and substations is essential for the future of data centers in the MRA (Gemeente Amsterdam, 2020; Gemeente Amsterdam and Liander, 2021). Before the moratorium, municipalities had few means to curtail data center construction through zoning and environmental planning, but developing renewable power plants, electricity substations and grids requires complex, controversial and lengthy planning and approval procedures. More restrictive municipal policies have changed this somewhat, but horizontal coordination between energy and data center policies, inter-municipal coordination and vertical coordination remain challenging.
Moreover, data center spatial policies at the municipal, provincial and national levels differ in their political priorities and approach (Economic Policy Advisor, Province of Noord-Holland, interviewed on 14 December 2021). The spatial roadmap, most municipalities in the MRA, and the province aim to facilitate ‘selective’ growth in the data center sector, whereas Amsterdam municipality has introduced far-reaching restrictions. In municipal policies, this selectivity is mainly determined by constraints such as grid bottlenecks and land scarcity. The implication is that more growth is desirable—even if impossible within a municipality's boundaries (Masson et al., 2022: 61f). National policies, however, highlight the locational competition in international data center markets, so that local demands cannot be decisive for the extent of national data center growth (ibid.). In the MRA, the construction of new data centers hinges on developing a hyperconnectivity cluster in Almere to mitigate spatial and energy stress in Amsterdam and Haarlemmermeer after 2030.We predicted what is now happening in Amsterdam: the grid is full, and we aren't able to expand the grids because there isn't any space … The grid operators can't say: ‘Move your buildings, we need space’. For [the municipality of] Amsterdam, it's also not easy … to do, so probably we started too late (Regional Lead, Liander, interviewed on 16 March 2022).
Responding to these planning decisions, grid operators started drafting plans to expand the grid to accommodate future energy needs in Almere. However, whether Almere municipality will allow the hyperconnectivity cluster is now uncertain (Regional Lead, Liander, interviewed on 16 March 2022). Although the municipal government initially supported the project after consulting with energy grid operators and approved construction on nearly 200 hectares of land, in January 2022, it decided to pause construction pending clearer guidelines on data center construction from the national government (Clahsen, 2022). The decision, however, impacts the MRA's trajectory in data center spatial planning and energy grid expansion; it may make it necessary to revise existing spatial strategies (ibid.) and abandon energy grid expansion.
Despite institutional fragmentation and conflicting priorities, several multi-stakeholder working groups in the MRA are trying to resolve the challenges of energy use and carbon emissions, collaborating across the institutional boundaries of policy levels, policy fields, infrastructure domains and involved local jurisdictions. However, the energy use and carbon footprint of decentralized ICT customers who store their servers in colocation centers remain entirely unregulated, as do the distributed users of these servers. Many data centers offer their services for monthly flat rates (Manager, AtlasEdge Data Centres, interviewed on 13 June 2023), disincentivizing ICT customers from saving energy. Regulatory gaps for the individual data center tenants include their servers’ energy efficiency, e-waste disposal, reuse or recycling, and regulations forcing tenants to switch their servers to standby mode when not in use. So, colocation center operators can only advise their customers (ibid.). Even more decentralized and beyond any demand-side management are ICT users’ data use and traffic, and practices such as deleting redundant data to save electricity.
Thus, even under tighter environmental regulations and effective spatial planning, data centers’ growing demands will exacerbate scarcity of critical materials, green electricity, grid capacity and land.
Unfulfilled promises of data centers as a residual heat source
Amsterdam's vision of urban heating transitions redefines data centers as low-carbon thermal infrastructure, as many of them rely on renewable electricity and could provide non-fluctuating and stable heat (Gemeente Amsterdam, 2020b). One of the cornerstones of the post-moratorium agreements is the requirement for new data centers to recover their residual heat and feed it into heating networks for use in nearby households, offices and other facilities. Agreements on using residual heat must be stipulated in contracts between the developer, operators of new data centers and heat providers. Although our interviewees—including data center operators, government leaders and utilities—see this as an opportunity for data centers to increase their sustainability, many also expressed concerns regarding the technical, financial and governance restrictions.
Data center operators consider the provision of residual heat to be beyond their core business. One noted that investors would not construct a data center in a specific location merely to accommodate residual heat infrastructure, since proximity to residual heat networks ‘is definitely a second-tier consideration’ (Strategic Advisor, Lumen Technologies, interviewed on 1 December 2021). Moreover, due to the IT industry's high innovation dynamics, its investment cycles and amortization periods are typically much shorter than those of other networked infrastructures. Data center operators are thus reluctant to conclude long-term heat supply contracts (Regional Lead, Liander, interviewed on 16 March 2022). Yet these are needed for district heating providers, given their long-term planning horizons, the high sunk cost of infrastructure investment and long-term return on investment. Hence to date only individual pilot projects have been realized (e.g. the heating of 1,400 student homes and a swimming pool).[The municipality] can allow it [the network], but then the question arises who should pay for a network fed with residual heat, so, if you let it be built by a third party, that party must have some assurances that it can make a feasible business case (Economic Policy Advisor, Noord-Holland province, interviewed on 14 December 2021).
Non-transparent and exclusive decision-making processes
As major public interests are involved, the externalities of material data infrastructures raise multiple questions regarding their democratic control and participation, the distribution of responsibilities across different policy levels and fields, and accountability for data center planning. Currently, neither the key responsibilities for different government levels, neighboring municipalities and infrastructure providers, nor the power relations between them, are clearly defined. Examples are: the negotiations between Hollands Kroon municipality and Noord-Holland province about the contested authority to grant permits for hyperscale centers; the responsibilities shared between operators of data centers, electricity and residual heat providers in reducing electricity use and developing data centers as thermal infrastructures; the accountability of land-use planning in confining data center growth; the coordination between neighboring municipalities in data center planning; and the responsibilities of ICT customers in colocation centers in reducing electricity use.
As a result of non-transparent procedural norms and organizational responsibilities in data center policies, the ability of citizens and their elected representatives to exercise sufficient democratic control remains limited (Masson et al., 2022). Colocation data center operators were able to operate below the radar of the broader public and local politics for a long time, e.g. by ‘hiding’ their materialities discursively (‘cloud’ metaphor), spatially (suburban business zones, converted industrial and/or fenced buildings), institutionally (simplified approval procedures for new centers), and ecologically (by marketing their investment in green technology). Hence, despite their considerable externalities, the growth of colocation centers has mobilized criticism and resistance only from concerned energy utilities, land-use planners, researchers, and local politicians, and hardly any from residents, journalists, or civil society organizations.
But, even in the case of the large hyperscale data centers, not all interested parties have been equally involved in the various phases of the decision-making process (Masson et al., 2022): parties with a primary interest in a positive decision (the technology companies and the landowners, and/or developers involved) had privileged access to regulatory authorities, whereas network operators or concerned residents were informed late (ibid.). Also, information about intended investors, e.g. the anonymous investor in a hyperscale center in De Kwakel (see above), is often not public, which reinforces citizens’ and political representatives’ suspicion that their interests are not taken seriously and that transparency of and democratic control over the decision-making process are not ensured (ibid.). As a result, the projected or realized hyperscale centers in Hollands Kroon, Almere and De Kwakel sparked considerable civil society protests.
Conclusion
This research adds to critical data and smart city studies by focusing on the materialities of urban data center infrastructures serving as the backbone for ubiquitous computing and smart cities. To date, smart city research and policies have primarily focused on the benefits and pitfalls of the increasing datafication and digitalization of urban services, how large tech companies and computational logics shape urban development and planning, and the wider sociopolitical impacts of smart city development. However, much of the scholarly and policy work on smart cities obscures the infrastructural materialities of data flows, their situated geographies and ecologies, and the sociopolitical negotiations surrounding their development. Rematerializing smart city research by focusing on urban data centers—one of the material manifestations of cloud infrastructures—thus provides a valuable entry point for exploring the environmental and spatial impacts of increasingly digital urban worlds, and for uncovering the controversies surrounding their regulation and planning.
Our case study on the MRA reveals that data center investors long operated in a regulatory void with hardly any public oversight of their hidden material infrastructures. Instead, municipal, provincial and national governments succumbed to the promise of global investment in digital infrastructure in consolidating Amsterdam's position as the digital gateway to Europe and facilitating digital transformations. Thanks to the low visibility of data centers’ materialities, the centers’ extractive relationships with local communities resulting from their high demands for energy, water and land, while, concomitantly, creating limited jobs and local economic value were long unnoticed by policymakers or the broader public. However, while public protests by concerned residents, farmers, NGOs and critical journalism centered mostly on hyperscale data centers, multiple expert reports and increasing complaints from water and energy utilities pushed the materialities of data centers and their massive environmental and spatial impacts onto the political agenda. Amsterdam's subsequent pioneering regulatory and planning efforts accelerated the adoption of a cascade of policies to confine future data center growth and align it with the capacity of the electricity grid and scarce land resources in the region.
Responding to growing public and political pressure, Amsterdam's data center operators have increasingly attempted to ‘green’ their infrastructures, primarily by purchasing renewable energy. However, efforts to improve the energy efficiency of their facilities and their initial embarking on local circular economy initiatives cannot conceal the environmental and infrastructural rebound effects, as continued data center growth in the MRA will aggravate future energy and land demands. Whereas Amsterdam can be seen as an international frontrunner in adopting new regulations and planning instruments to confine data centers’ unfettered growth, our analysis points to multiple gaps in regulatory frameworks that cannot be solved by municipalities alone. Recent policy reforms may, at best, facilitate spatially confined growth patterns for data infrastructures and better align them with the capacity of electricity grids. Yet, regulating an industry whose operations, land and resource use remain opaque, whose investment cycles and regulatory processes are misaligned with those of electricity and heating infrastructures, and whose international investors have privileged access to policymakers, remains a challenge for future sustainability policies.
Increasing the strain on electricity grids and on scarce renewable energy and land in the MRA raises complex distributional conflicts and questions of contested territorial and energy futures that reach far beyond the institutional boundaries of digital policies and municipal jurisdictions. Data center growth engenders unintended consequences for other significant policy objectives, such as decarbonizing and electrifying mobility and heating, the affordability of energy and land, and regional growth plans to build 250,000 homes. Their implementation strongly depends on having sufficient urban grid capacity and land, scarce renewable energy resources and critical raw minerals. Instead of broadening dialogue across the affected policy fields and the broader public about which political objectives deserve priority given the scarcity of electricity and land, policies mainly focus on accommodating data center growth.
The materialities of data infrastructures thus raise important and as yet unsolved questions for urban policy and scholarly analysis about the value propositions of smart cities and digital urban solutions: more attention needs to be given to digital sufficiency and how ICTs could become an essential part of broader environmental transformations aiming for fewer devices, software designs that reduce data traffic, users applying digital devices frugally, and digitalization supporting a transition to economies that recognize environmental limits to the ever-growing demand for energy and critical resources (Santarius et al. 2022). Concomitantly, more attention in data center development and governance should be given to the broader publics who have a stake in the hidden resource, land, and energy politics of the digital economy: citizens and ICT users, climate activists, critical scholars, utility companies, and environmental and land-use planners.
Biographies
Jochen Monstadt, Department of Human Geography and Spatial Planning, Faculty of Geosciences, Utrecht University, Princetonlaan 8a, 3584, Utrecht, CB, The Netherlands, [email protected]
Katherine Saltzman, Department of Human Geography and Spatial Planning, Faculty of Geosciences, Utrecht University, Princetonlaan 8a, 3584, Utrecht, CB, The Netherlands, [email protected]
References
- 1 Researcher, Waternet, interviewed on 6 December 2021; Economic Policy Advisor, Province of Noord-Holland, interviewed on 14 December 2021; Regional Lead, Liander, interviewed on 16 March 2022.
