The Coming Land Grab for AI

In the folds of North Devon, where pasture breaks towards river valleys and the pattern of fields still reflects an older economy, a proposal has arrived that seems to belong to another country entirely. Xlinks, the company once associated with the abandoned ambition of bringing solar power from Morocco to Britain through a vast subsea cable, now wants to build what would be Britain’s largest artificial intelligence data hub on 850 acres near Great Torrington. In financial terms, the scheme is pitched as a statement of national seriousness about AI. In physical terms, it would be a new industrial province laid across an agricultural landscape. For local people, those are not two ways of describing the same thing. They are the beginning of an argument about whether the race for computing power is being pursued with any real reckoning of what it consumes, where it belongs, and who is expected to bear the cost.

The bare dimensions of the project explain the alarm. Opponents say the site would swallow land equivalent to roughly four average British farms. Structures on the site could rise to around 40 metres, introducing a scale and mass more associated with logistics parks and energy infrastructure than with the quiet topography of rural Devon. Xlinks says it plans to invest £13.8 billion in a 1.5 gigawatt data centre, with an associated energy storage site, and argues that the development could create 650 jobs while contributing as much as £3.6 billion to the national economy. Those are not trivial promises, particularly in a region where secure, well-paid employment is not always easy to come by. Yet the company’s economic case has collided almost at once with a deeper suspicion now gathering around the infrastructure of AI: that its boosters prefer to speak the language of innovation while leaving others to calculate the price in land, power and water.

It is water, above all, that has turned local unease into organised resistance. Data centres may be discussed as if they were abstract assets in a digital economy, but they are heat-generating machines on a colossal scale. Cooling them demands enormous resources. Across Britain, data centres are estimated to consume more than 14 million litres of water a day. For farmers in North Devon, that statistic does not belong to some remote national ledger. It points directly to an anxiety they already understand in practical terms. Barbara Fryer, who runs a nearby cattle farm, has said the water table is already down and the grass has browned, with reduced livestock and lower food production to match. Her objection is not rhetorical. It arises from watching the land respond to stress in real time. When a farmer says there is not the water for a development of this size, that is not a cultural complaint dressed up as environmentalism. It is a judgement formed by repeated exposure to shortage.

The concern extends beyond individual holdings to the wider ecology of the area. Matthew Chatfield, another Devon farmer, has warned that the River Torridge could be placed under further pressure by a facility whose cooling demands would be immense. Xlinks insists the site’s cooling system would rely primarily on harvested rainwater, reducing the need to draw on the river or local supply. That assurance will no doubt be central to any eventual planning application, and it deserves scrutiny rather than dismissal. But here the company runs into a problem increasingly faced by developers of strategic infrastructure. Trust cannot be won simply by presenting a technical workaround. Rainwater harvesting on a major industrial site may sound prudent, yet sceptics will reasonably ask what happens in dry periods, how much storage is required, what supplementary sources would then be used, and which public systems might ultimately absorb the strain. In a country that still talks as if plentiful rain automatically means plentiful water, those are questions that arrive late and then all at once.

There is also the question of suitability, which is less easily quantified but no less real. Penny Mills of the Devon branch of the Campaign for the Protection of Rural England has described the scheme as an extraordinary proposal for the North Devon countryside and has said local feeling is strongly against it. That judgement is not simply nostalgic. It reflects an intuitive but important planning principle that Britain has often struggled to defend in practice: not every parcel of land should be treated as interchangeable merely because a spreadsheet can assign it a value. North Devon is not empty space waiting for higher economic use. It is a working landscape, environmentally sensitive, agriculturally productive and central to the identity of the communities that live within it. Once that kind of place is recast as the host for a giant industrial compound, the loss is not captured only in acreage or visual impact. It alters the logic of the place itself, and after that the argument about precedent becomes impossible to avoid.

It is significant that no land has yet been acquired and that planning permission has not been sought. In formal terms, the development remains an intention rather than an approved fact. Even so, the political and cultural battle has already begun. A petition opposing the scheme has attracted more than 2,000 signatures, evidence that the resistance is neither confined to a few neighbours nor likely to disappear under the promise of inward investment. Britain’s planning debates are often caricatured as a contest between modernisers and sentimental objectors. That framing is too lazy for what is happening here. Rural opposition to a project of this size is not automatically anti-growth. It may instead be a demand for seriousness from a political class that has become fond of invoking national missions while leaving localities to absorb the consequences. If ministers want an AI build-out of historic scale, they cannot be surprised when people ask whether the state has any coherent view of where such infrastructure should go, what it should displace, and which trade-offs are acceptable.

That challenge lands awkwardly for Labour, which has made clear that it wants the conditions for rapid investment in AI and data centres. The impulse is understandable. Every advanced economy now fears being left behind in the contest for computational power, semiconductor access and the industrial ecosystems clustered around them. Governments have started to regard data centre capacity in the way earlier generations regarded ports, roads or electricity grids: as enabling infrastructure without which national ambition becomes decorative. Yet Britain’s enthusiasm for this sector runs into a contradiction it has not yet resolved. A large-scale AI economy requires exceptional volumes of electricity and, in many cases, water. It also demands land, often in places where power connections are plausible and local opposition can be managed. That sits uneasily with stated commitments to net zero, environmental resilience and food security. None of those aims is impossible to reconcile with digital expansion, but the rhetoric of frictionless technological progress makes the collision harder to discuss honestly.

The Devon proposal is therefore not a local oddity. It is a sharp instance of a national pattern beginning to emerge. Across the country, data centres are being planned or built at pace, each one drawing on already stretched systems of energy distribution and water management. Thames Water has warned of rising pressure on supplies from data centres in its own region, and what is true in the South East will not become less true merely because a project is moved westward into a greener landscape. Indeed, one irony of the present moment is that rural areas may be judged suitable precisely because they appear spacious and wet, as if rainfall maps settled questions of environmental capacity. Janet Adams, a dairy farmer from Torrington who has offered support for the proposal, has observed that it is better to locate such facilities where there is rain. She is not wrong to identify geography as part of the equation. But rain is not an unlimited commons, and the fact that a region looks lush does not mean its aquifers, rivers and supply systems are there to underwrite whatever the digital economy demands next.

Xlinks and the Government will continue to frame developments of this kind in the language of jobs, growth and strategic necessity. A government spokesman has said ministers are determined to create the right conditions for investment in AI and data centre infrastructure while protecting communities and the environment. That is the formula every administration prefers: economic transformation without social or ecological injury. The difficulty is that such balancing acts are easy to describe before a scheme reaches a planning desk and much harder to prove when the details arrive. How many of the 650 jobs would go to local workers, and on what terms? How much of the economic value would remain in Devon rather than being extracted by contractors, operators and energy suppliers based elsewhere? What resilience standards would govern water use? What commitments would be enforceable rather than aspirational? The politics of AI infrastructure increasingly turns on that distinction between headline claims and binding obligations. Communities asked to live beside these projects are entitled to demand more than a prospectus.

There is a further question, less often stated plainly, about what kind of economy Britain is building when prime agricultural land becomes expendable in the service of machine intelligence. The country has spent years talking more seriously about food security after repeated shocks to supply chains, volatile weather and rising geopolitical instability. At the same time, ministers have encouraged the expansion of industries whose land and energy needs can compete directly with farming. In Devon, that contradiction is especially vivid. The proposed site is not derelict terrain or a spent industrial zone awaiting reinvention. It is productive countryside in a county whose rural economy depends not only on farming income but on the broader integrity of the landscape. There are circumstances in which the repurposing of land can be justified by overwhelming public benefit. But such decisions ought to be taken with candour about what is being surrendered. Too often Britain speaks of technological sovereignty as if it arrives free of territorial consequences.

What North Devon exposes, then, is the material reality that sits beneath the airy prose of the AI age. Intelligence may be artificial, but the infrastructure sustaining it is stubbornly physical. It needs transformers, cooling systems, roads, batteries, substations and water storage. It generates noise, heat, traffic and conflict over resources. It changes skylines and rearranges local priorities. Those facts do not amount to an argument against building data centres, still less against technological progress. They do, however, demand a more adult politics than Britain has so far managed. If the country wants to host the next generation of computation, it must decide whether it also has a framework for protecting landscapes, water systems and food-producing land from being treated as secondary considerations. In Great Torrington, before a single spade has gone into the ground, residents have already intuited the larger truth: the future may be written in code, but it will still be fought over on the land.