This open access book assembles cutting-edge research in media and communication exploring AI infrastructures and sustainability. It builds upon and expands perspectives on media infrastructure, addressing critical issues such as environmental impacts, economic concentration, and social justice – areas that have not been comprehensively examined under the umbrella term “sustainability” within discussions on AI. The authors explore the complex social, economic and material processes through which AI formations take shape, and offer novel perspectives on the entanglement of digital media, AI and sustainability discourses as they manifest in the shaping of futures with AI.

This chapter explores the relationship between energy, sustainability and digital infrastructure for AI from the perspective of infrastructure misalignments. It focuses on a misalignment between the temporalities and pace of scaling of cloud computing infrastructure, and those of energy grids, showing how this temporal misalignment results in the crowding out of industries and in putting pressure on public values of equal access to electricity. Using examples from Sweden, the chapter further discusses collective efforts by regional and national actors to repair the situation by upscaling electricity grids and change planning practices that in turn create further misalignments and societal conflicts. I suggest that infrastructural misalignment as a perspective on digital infrastructure and sustainability invites a shift in focus from questions about the carbon emissions of computing infrastructure towards concerns with resource access, public values and the common good. 

This commentary is a response to Thao Phan’s article that tracks the relationship between contemporary digital test beds and histories of colonialism, racism and empire. Building on Phan's arguments, I suggest two possible further expansions: one is to link drone experiments with global logistics and militarism, and track how the end of experiments enables a slippage between military and commercial targets of emerging technologies by iteratively producing new forms of targets. The other one is to ask how testbeds could also operate as sites for scholarly experiments and interventions on the test, to different ends and to a different success. 

This chapter argues the need to expand media and communication research in two ways. We first indicate the necessity of broadening current media and communication approaches to AI and automation to include analyses of their material underpinnings. We then propose expanding discussions of AI to examine it as an analytical phenomenon along its infrastructures, encompassing the manifold relations that actors have with AI from mineral mining to the AI commodity. Situating AI infrastructures as a key concern for media and communication research will help account for the inequalities, economies and societal conflicts that emerge within AI-saturated futures. In other words, AI must be explored at the intersection of infrastructure and sustainability.

This chapter discusses what a good digital infrastructure for AI could mean, and for whom. It explores the frictions that emerge when communities, industries and public institutions collectively try to imagine and materialize multiple versions of the “good”. Drawing on the concept of “friction” and valuation studies, we analyse the establishment of a data centre for AI systems with a sustainability profile in Östersund in Sweden. We show how such digital infrastructure emerges from the friction between differing local understandings of liveable futures and between quantifiable and non-measurable qualities of the good, such as sense of place, belonging, trust and living a life with nature. Ultimately, we propose an expansion of the ongoing discussion on the sustainability of digital infrastructure and a reorientation of attention from questions of quantifiable and measurable harms, such as carbon emissions, to concerns with power and valuations through which the different qualities of “good” infrastructure come to fruition and contention.

LIFE TODAY is data driven – and the basis of data collection. Industries and institutions gather more digital data for their potential to generate value and action than ever before. These range from health data (Stevens et al. 2018; Vezuridis & Timmons 2021) to environmental data (Deitz 2023; Vardy 2020) to “alternative” data which are cleaned, repackaged and sold for use by industries in need of predictive models, like finance (Hansen & Borch 2022; see also Plantin 2019).

This special issue provides a window onto the meeting of critical data studies and sociology in Scandinavia to explore the situated, localised modes of “being and beco-ming” with data (Lupton 2018:9) in a particular moment in time, a time when in the Northern European context, public administrators, institutions, software developers and scientists are busy imagining and trialling arrays of data-driven solutions to au-tomate public services, intellectual work and infrastructure operations. Data, in these contexts, often figure as an element of already ongoing, much older practices of clas-sifying, categorising, quantifying and producing actionability on the social. However, data bring to these the novel element of the imagined possibility for feeding these practices into machine learning, automation algorithms and computational models to fulfil ideas of efficiency, increased productivity or predictive potentials. Data become, in this context, part of a landscape of instrumenting actions and socio-technical struc-tures that order social life at a distance and according to formal rules (Savolainen & Ruckenstein 2024).