Data centres, ‘AI’ and strategies for emancipatory resistance

1) The close connection between data centres and ‘AI’ 

The massive expansion of data centres and the hype surrounding ‘AI’ are directly linked. This is because the hype surrounding ‘AI’ mobilises the financial resources and unanimous political support required for an expansion of physical infrastructure that is unprecedented in both speed and scale. This connection encompasses technical, economic and political dimensions.

• The technical connection: Data centres are, in a sense, the factories of ‘AI’. While data centres have been around for some time (since the advent of the ‘cloud’ as centrally managed computing capacity), ‘AI’ requires many more and much larger data centres: Some data centres are used primarily for training ‘AI’ models (these require particularly expensive semiconductor chips and are often located in more rural areas), others primarily for inference, i.e. calculating ‘AI’ outputs based on trained models (these are usually located as close as possible to internet hubs, often in urban areas). Data centres are at the end of a complex global supply chain: Particularly evident is the increased demand for rare earths and metals, as well as for water and (fossil) energy, which are required at many points along this supply chain: in the production of lithography machines, which in turn are necessary to produce chips, which in turn are necessary to operate a data centre. The infrastructure surrounding data centres and ‘AI’ has been expanding significantly for several years and increasingly includes end devices whose functioning depends on a continuous connection not only to the internet, but also to ‘AI’ and the cloud (ranging from home appliances through industrial production to military jets).

• The financial connection: The financing of data centres is underpinned by an almost religious belief in the inevitability of ‘AI’ and speculation regarding its future profitability. ‘AI’ companies incur losses running into the billions annually because their expenditure – in the form of costs for computing capacity, energy and wages – exceeds their revenue. Data centres are the focal point for financial risks: chip manufacturing is (so far) profitable (see the financial statements of Nvidia, TSMC and SK Hynix), and the same applies to the provision of computing capacity as a service, which is offered primarily by the three hyperscaler companies—Amazon, Microsoft and Google. Data centres sit between chip manufacturing and the provision of computing capacity and are often merely leased by those who, in turn, use or sell computing power. Data centres are extremely capital-intensive (up to several billion euros per data centre) and highly speculative, so that even people within the data centre industry (anonymously/confidentially) say: ‘Anyone who understands the risk won’t invest.’ The risk (i.e. the probability of losses) stems, amongst other things, from the short lifespan of chips (one, two or a maximum of seven years, depending on the source), the long-term financing of data centre construction (often 15–20 years), a high degree of uncertainty regarding the availability and price of chips, energy and water, and an equally high degree of uncertainty regarding the demand for ‘AI’ computing capacity. These risks are currently becoming increasingly intertwined through complex circular purchasing arrangements between ‘AI’ companies, chip manufacturers and data centre operators, a situation which even financial analysts view critically. For several years now, analysts at Goldman Sachs, amongst others, have been pointing to the existence of a ‘bubble’.

• The political connection: The supposed inevitability of ‘AI’ is politically linked to a competition for geopolitical domination (particularly between the US and China), a quest for ‘sovereignty’ (e.g. through chip manufacturing and data centres in Europe) and the promise (or threat) of ever more powerful ‘AI’ models. Consequently, the question of whether and for what purpose data centres are needed is not debated in society; rather, their necessity is dictated by politics. Civil society organisations are currently, at best, asking whether a smaller scale or more sustainable operation might suffice. This is insufficient and prevents any engagement with the highly political implications of data centres and ‘AI’ as catalysts for existing crises:

  • Environmental destruction is on the rise, as along the supply chains (involving mines, fossil-fuel power stations and chip factories) entire regions are sacrificed to ‘AI’ and poisoned for all forms of life.

  • Work is devalued and intensified, as ‘AI’ automation forces workers in the Global South into ever more intense exploitation, while devaluing and disempowering workers in the Global North.

  • The capacity for collective organisation declines, as ‘AI’ models are designed to mimic human communication and supplant social interaction. Racist and anti-feminist forces are being strengthened, as ‘AI’ models reinforce white, cis-male power structures and reduce social diversity by exaggerating the statistical patterns of the supposed majority. This applies equally to other forms of exclusion and denigration such as ableism, antisemitism, …

  • Social inequality is on the rise, as society is expected to bear the brunt of austerity and deprivation to support a ‘bubble’ that socialises foreseeable losses.

  • Militarisation and violence are technologically amplified and trivialised under the guise of greater efficiency and precision.

  • Power and control are becoming increasingly concentrated, while accountability and responsibility are being technologically obscured and outsourced.

  • Fascist ideologies are being normalised through the pursuit of a supposed superintelligence, which is linked to eugenic concepts for sorting lives into those deemed valuable and those deemed unworthy.

2) Fighting together for climate and social justice, against fascism and militarisation

The multifaceted destructive power of ‘AI’ and data centres affects a wide range of political issues. This means that many movements can join forces in resisting the technological assault. ‘AI’ and data centres are currently among the most tangible manifestations of inequality, the pursuit of profit and economic power. As such, they also present an opportunity for movements to formulate and shape alternative visions for future societies, free from the market and private property. Many of these efforts are still at an early stage, but their interconnection could enable them to reinforce one another. We see potential starting points for struggles centred on data centres and “AI” in the following movements.

• Climate justice: Water consumption, fossil fuels, mining, land use and electronic waste are all potential starting points. In view of increasing drought across large parts of Europe, future struggles over water as a resource could form a possible focal point for struggles against ‘AI’ and data centres. Reformist cosmetic measures for greater efficiency or ‘small AI’ would merely stabilise and downplay capitalist destruction.

• Social justice: An alliance is conceivable and desirable between those who develop ‘AI’, those whose work it is intended to replace, and those whose education is rendered meaningless by ‘AI’. There are already initial expressions of resistance, such as collective refusal to allow data to be used for ‘AI’ training and strikes against ‘AI’. Developers can help to ‘reject and undermine “AI”’, and universities can be places of resistance through collective rejection of “AI”. The “AI”-accelerated automation of social cuts, which obscures accountability, also calls for opposition.

• Anti-fascism: The link between political and economic power and anti-democratic to fascist ideologies is evident not only in the US under Trump, with ‘AI’ advocates such as Musk, Thiel and co. Their impact is also felt here and is being being emulated, for example by tech billionaire Gotthardt, who spreads fascist-style propaganda via the ‘news portal’ Nius. Resistance is already emerging against Tesla and against Nius.

• An anti-fascist movement must address the unequal conditions of the political public sphere that favour right-wing propaganda and target not only Nazis, but also these very technologies.

• Anti-militarism: The arms race involving ‘AI’ promises more efficient killing and is encountering a newly forming anti-militarist movement. It can take its cue from the diverse processes and junctures where the civilian sphere becomes interwoven with the military. It is already well documented how supposedly social interactions feed into the training of military ‘AI’ and where military ‘AI’ is also gaining influence in Berlin.

To ensure that this resistance naturally offers as few points of connection as possible for right-wing ‘homeland defenders’, we should focus on clear emancipatory messages. In this regard, it can be helpful to build practical international solidarity at an early stage and to draw on the struggles for indigenous territories and those against ‘AI’-driven data processing.

3) Effective resistance through diversity and unpredictability

The strategy for resisting a major political and economic project must be chosen with care. Experience from the history of social movements suggests that diverse and unpredictable struggles can have the greatest impact.

We see an example of the successful interweaving of forms and locations of action in the anti-nuclear movement. For strategic reasons, the movement decided to focus its resistance on the transport of nuclear waste in order to prevent the use of nuclear energy: the spent fuel assemblies had to be regularly transported abroad to reprocessing plants and, in accordance with international treaties, returned for final disposal after reprocessing. The German government’s plans envisaged a central final repository for highly radioactive waste in Gorleben. This provided two focal points for action: firstly, the shipments of nuclear waste from the nuclear power stations to the reprocessing plants as anchor points for (supra-)regional resistance, and secondly, the return shipments to Gorleben as a focal point for nationwide and even international movements. These focal points saw the interplay of a wide variety of forms of action, which were not centrally controlled but organised on a decentralised basis.

Most campaigns today have lost or even deliberately got rid of their diversity and unpredictability. For example, ‘Deutsche Wohnen Enteignen’ and ‘Widersetzen’ have moved towards making the resistance scenario as predictable as possible. With pre-determined action scenarios and by restricting themselves to methods of (maximum) civil disobedience, organisers hope to attract more participants. As a result, at larger events, the authorities increasingly rarely lose control. This makes it difficult to open up practical spaces of possibility and, with them, new social spaces for deliberation. We also see this loss of impact clearly in the struggle against large housing corporations, when the political opponent simply undermines the legalistic campaign through new laws.

For a diverse and unpredictable movement against ‘AI’ and data centres, the following might be conceivable: a focal point could, for example, be a data centre of outstanding significance and/or ESMC in Dresden, Europe’s largest high-performance chip manufacturer. Similarly, the lithography machine manufacturer ASML in Eindhoven, with its globally unique position—both technically and geostrategically—as a bottleneck in chip production, could become a focal point for a pan-European movement.

All of us who want to put an end to coal-fired power generation in our local areas, who oppose the far right, who want to disarm Rheinmetall, who want to expropriate Vonovia/Deutsche Wohnen: let us stop being civil and obedient, and let us begin to offer unpredictable resistance. Let us join forces with those who organise in less visible contexts and with those who, for good reasons, have never been visible.

Let’s form gangs instead of just support groups; let’s combine autonomous spontaneity with campaign work; let’s be unpredictable and show solidarity with one another. Let’s work together to drive ‘AI’ out of our daily routines and our political practice, and let’s block its destructive materiality in data centres!

Further resources:

– Conference on 7 November 2026 on ‘AI’ and data centres (DE)

– Discussion paper by Aufstände der Allmende (DE)

– capulcu on criticism of ‘AI’ (DE)

– Structural Integrity on ‘AI’/inequality, ‘AI’ after capitalism and the ‘AI’ bubble (EN)

– Manifesto 2.0 from the Cables of Resistance conference (DE/EN)