From the outskirts of Athens to the centre of the Po Valley and the woods of North Eastern Finland, the European Union’s (EU) AI race-to-the-bottom is taking place all over Europe in areas which are already under significant environmental pressure. Beyond the promises of innovation, economic growth and digital sovereignty, the relentless expansion of AI infrastructure in the form of data centres comes with a significant consumption of natural resources and risks deepening already existing social inequalities. As the EU and its Member States accelerate investments in AI infrastructures, an unavoidable question emerges: who benefits from this development – and who bears its energy-related costs?
Contradiction of the twin transition
The ‘twin transition’, as envisioned in the EU Green Deal and the EU’s industrial strategy, implies that the transformations leading to a climate-neutral and digital economy and society are meant to reinforce each other.
This digital-green twinning becomes prominent in specific sectors, such as that of energy, where the urge to shift from fossil fuels to renewable energies is presented as intertwined with the digitalisation of the energy system, through smart grids, and artificial intelligence technologies. At the same time, the reality of the extremely significant resources’ consumption of AI infrastructures cannot be ignored. While policymakers slowly start to recognise the problem and adopt (mild) policy measures on energy consumption, Big Tech is experimenting with nuclear energy and revisiting their net-zero carbon goals, with much opacity and a major influence exerted on EU institutions to favour supposed confidentiality needs over transparency.
Ultimately, the EU twin transition narrative simply recognises some ‘externalities’ of digitalisation – mainly in terms of energy consumption – while digitalisation objectives take precedence over truly ecological transformation. This is profoundly different from acknowledging that some digital technologies can boost some changes necessary to move towards ecological sustainability.
Materiality of AI infrastructures
AI-enabling infrastructures are not only power-hungry, but their materiality exerts pressure on people and environments in many other ways. This includes reliance on critical minerals, land grabbing, need for clean water, disposal of e-waste and exploitation of individuals and collectives, co-opted by the various needs of the AI lifecycles.
The hyper focus on energy consumption can be seen as a fixation of the Global North, which perceives and frames it as the priority, when it is only one tiny piece of a much more complex puzzle. Still, unpacking the energy-related costs of AI deployment remains a salient starting point to critically look at AI’s infrastructures against EU and national regulatory approaches.
Energy justice meets AI’s materiality
The concept of energy justice provides a useful framework through which we can look at the unprecedented pace at which AI infrastructural expansion takes place, with a particular focus on data centres’ energy consumption. Emerging from environmental justice scholarship, energy justice seeks to address not only the distribution of the benefits and burdens associated with energy systems, but at the same time, the fairness of relevant decision-making processes themselves and the recognition of the people impacted by such decisions. Having gained academic prominence in the previous decade, scholars have recently started to conceptualise energy justice in relation to AI and digitalised energy infrastructures.
The lens of energy justice helps us to critically and comparatively analyse the cases of Greece and Italy, two of the several countries in the EU currently experiencing a major growth of AI-related infrastructural projects and united by similar climate and energy conditions. While the Nordics are characterised by large production of renewable energies and therefore attract investments for renewable-powered and supposedly ‘sustainable’ data centres, Italy and Greece represent blank slates of available soil and resources for public and private investments in AI infrastructures.
Greece: Strategic investments in an energy-poverty context
In Greece, several major data centre projects are concentrated around the Athens region, particularly that of Spata, via investments by companies such as Microsoft, DAMAC/DEI, Dromeus Capital/Apto. These projects are facilitated by Greece’s national legislation on ‘strategic investments’ which simplifies and accelerates approval procedures for investments considered crucial for economic growth. Although environmental impact assessments are formally required, public participation opportunities prior to the approval of these projects are substantially inexistent. Locals’ upheaval has resulted in filing an appeal before the Council of State.
This is particularly striking given that Greece continues to face significant levels of energy poverty. With local electricity grids already under pressure, local authorities’ concerns regarding whether renewable energy commitments could realistically be fulfilled were left unanswered.
Italy: Socio-economic development vs. environmental protection
In Italy, investments in AI infrastructures are concentrated in the Centre-North, and are promoted by Microsoft and Amazon and by EU initiatives, such as the IT4LIA AI Factory and the government’s manifestation of interest to host an AI Gigafactory, with envisaged cooperation with Greece.
All this against a geographical context where the Po Valley is among the most air polluted areas in Europe; Lombardy’s region has the highest level of soil consumption in Italy and all the new requests for connection to the grid would account for a 70 per cent increase in regional energy demand and would necessitate material interventions to be accommodated. Citizens' committees intervened to denounce and monitor the location of the new data centres on agricultural lands or common spaces. A lack of territorial design causes municipalities to operate randomly, while data centres ensure high tax revenues and authorization processes are being sped up by ad hoc rules.
The regulatory framework still consists mainly of non-binding instruments, including guidelines from the Italian Ministry of Environment and Energy Security, while a legislative proposal is currently before the Senate. While the Government will be delegated to adopt specific rules, the current proposal lays down the general principles, which, on the one hand, qualify the projects as ‘overriding public interest’ – with a consequent ‘simplification’ of environmental impact assessments procedures – and, on the other hand, ‘prioritise’ the regeneration of decommissioned energy production sites and projects involving ‘clean energy solutions’. However, if these priorities do not become legal obligations, they will likely remain on paper, as demonstrated by Lombardy’s cases.
How can AI development be aligned with energy justice, and what can EU law do about it?
Whilst EU’s AI ambitions will undoubtedly intensify in the coming years, the current regulatory approach to the environmental footprint of AI is limited to (weak) reporting obligations, which do not substantially regulate the matter and are often overridden by confidentiality claims (e.g., Microsoft).
To align the twin transition and AI development with energy justice, questions about the socio-ecological impact of digital expansion, the allocation of natural resources, and the democratic participation thereupon cannot remain peripheral considerations or mere externalities. They must move to the centre of the EU's legislative agenda, as citizens in Greece and Italy begin to advocate with their resistance and their claims.
This research has received funding from the European Union's Horizon Europe research and innovation programme under grant agreement No 101189771 (DataPACT project) and Katholieke Universiteit Leuven's Research Fund (FlexIQ project C2M/24/028)
******
This blog post is based on the paper to be presented at the conference Sustainability Law: Law’s Response to the Convergence of Crises, 16-18 September 2026, at the University of Oslo
