Judy Kuszewski, Founding Partner
Almost as quickly as the generative AI industry has boomed, public (and increasingly official) opposition to data centre development has mushroomed into what may be the mass-resistance cause celebre of our time. There are several main reasons for this opposition, whether to a specific data centre or to the systemic compound impacts of the sector. But it is worth considering broadly what makes this opposition specifically appealing to so many people, and whether this is the most visible front in a wider war.
Communities around the world are pushing back against the march of DCs in their neighbourhoods:
- In Buckinghamshire, England, about 25 miles from where I live, a coalition of pro-environment, pro-social and tech sector activists have recently halted the development of a ‘hyper-scale’ DC, arguing that the government’s decision to force it through disregarded the energy and water requirements and climate impacts of such a project.
- In January this year, the town of Quilicura, near Santiago de Chile, gained notoriety for its campaign to discourage the unconsidered use of AI while their town – host to multiple DCs – was facing a severe water shortage. Their campaign offered would-be AI users the services of real-life skilled people from their community for a single day. Over 25,000 prompts were received. Flesh-and-blood researchers answered such deep questions as ‘why are hot dog buns sold in packs of eight and sausages in packs of 12? And living artists and designers drew portraits of pet dogs or created images of sloths playing in the snow.
- The city of Marseilles in the south of France has been the site of numerous locally-organised campaigns against the uncontrolled development of DCs, whose power and water use requirements have created significant environmental justice challenges for local people. The city is one of the poorest in France, and the DC industry has directly prevented the cleanup of serious air pollution by delaying the electrification of the city’s port.
What many of these and other examples of public opposition have in common is an emphasis on the direct conflict DCs pose between community needs and DC demands – especially in relation to energy and water. Because of the specific demands of markets for AI services, DCs are by definition concentrated in populated areas. This puts computational power near to industries such as financial services (which value low latency times), but in the process exacerbates demands on already-strained infrastructure, and in the case of metropolitan areas such as London, monopolises large amounts of physical space where housing stock is in short supply and precious shrinking ecosystems are under constant pressure.
Developers – and supportive governments – counter with promises to power DCs with renewable energy, to operate to the highest standards of efficiency, and to create jobs and tax revenue locally. In reality, all these claims are controversial at best, with renewable energy purchasing contracts simply monopolising existing renewable supplies and creating no new generating capacity; and public subsidies and tax breaks vastly outstripping any benefits in jobs and local revenues. An academic paper cited by the Centre for Inclusive Trade Policy found that U.S. data centre investments in receipt of at least $811 million in tax relief generated a measly 837 permanent jobs in total. That’s a million dollars of public money for each job created – for a collection of DC deals valued at a total of some $10 billion.
Moreover, we must confront the systemic impacts, risks and opportunities that DC (and more broadly AI) development brings. It is safe to say that a single data centre on its own is unlikely to raise many alarm bells outside the locality where it is proposed to be built (unless it’s Meta’s would-be hyper-scale Zeewolde campus in the Netherlands). But in some ways, this is exactly the problem: if data centres are only considered one-by-one – can a single instance be justified in its location, with its specific configuration, and its individual impacts and potential benefits? – it is inevitable that the wider, system-level trends will be missed until the impacts are much wider than the individual planning application.
Campaigners claim that data centres in the pipeline in the UK would altogether double the country’s energy requirements in a few short years. The grid was not created for this, we don’t have the sustainable energy supplies to provide for it, and anyway, we need lots more electricity for other things (to bring down the cost of living, to reduce pollution, and to transform major economic sectors for the future).
The UK government declared AI data centres to be critical national infrastructure in 2024. This means, among other things, that vital services including energy and water will be prioritised for the needs of DCs even in times of crisis (such as the severe drought affecting much of the country in summer 2026) – over and above the needs of hospitals, agriculture, transport, households, and any other activities that people in the community are likely to value more highly.
The energy use trade-off – that is, that any energy used to power DCs cannot also be used to power other services or endeavours – raises afresh the question that Amory Lovins first posed fifty years ago in an article in Foreign Affairs, and recently revisited by Joel Makower: we should be asking not how much energy do we need, but what do we need energy to do? The apparent doubling-down on polluting fossil sources to feed the ever-growing numbers of DCs and jettisoning of efficiency as an aim altogether represents a massive backward slide into unconsidered demand-side thinking about this issue.
There is a ‘data sovereignty’ argument – which is real and important – that deserves to be acknowledged. Regardless of anything else, governments often share the view that sovereign countries cannot afford to have major economic activities (like AI) dominated by foreign interests. Yet it is important to recognise that simply locating DCs in a certain jurisdiction does not in itself confer digital sovereignty – this depends on the whole ecosystem of infrastructure that underlies it, including ownership questions and binding energy and water standards. Indeed, digital sovereignty in itself deserves discussion and whether it is an important objective in its own right may be a subject for another day.
And like any technology, whether AI is ‘good’ or ‘bad’ depends on what it is used for. Whether any given data centre is used for ‘sovereignty’ (or for any of the other benefits backers claim, such as curing cancer or solving climate change) rather than for deepfake porn may be impossible to say, meaning there is a much more fundamental discussion to be had about public benefit, public acceptance and what it is founded on. And this is to say nothing of Anthropic’s IPO prospectus warning of ‘existential risks to humanity’ inherent in its AI systems while still expecting a valuation of over $2trn – is there any other industry that could get away with this? – again, a subject for another day.
There are policy fixes for these problems, addressing systemic impacts and prioritising strategic solutions – if communities and governments have the resources and capacity to consider them. For businesses – most immediately those businesses that own, use, promote, or finance DCs or the digital services they enable – the opposition is only growing, meaning the challenges will become more visible, not less.
The real danger is that the rush to develop DCs without due care and consultation may have hardened public opinion to the extent that community, NGO, and increasingly government opposition is now more categorical than contingent. The need for powerful trust-building and radical honesty around the future of this industry is now immense.
Does your business need a principled, positive and pro-social strategy? Let us help you work through the challenge.