Nexus Europa

The AI boom is creating a new battle over power, water and land

Jian Oren
Posted September 2, 2026 · 8 views
Follow
Share
The AI boom is creating a new battle over power, water and land
Editorial Collage

The global expansion of AI is turning data centres into a growing political and environmental flashpoint. Communities are increasingly challenging projects over their heavy demand for electricity and water, while governments push ahead with construction in the name of technological competitiveness and digital sovereignty. The backlash is exposing a deeper question over who bears the costs of the AI boom — and how much local control should be sacrificed for it.

The rapid expansion of artificial intelligence infrastructure is fuelling growing resistance in communities around the world. Data centres, which until recently were seen as an unobtrusive part of the digital economy, are becoming a political issue because of their enormous demand for electricity, water and land.

For governments, they are critical infrastructure needed to develop AI and strengthen digital sovereignty. For local communities, they increasingly mean competition for resources, greater pressure on energy systems and decisions being made without sufficient public scrutiny.

The energy cost of artificial intelligence

The scale of future demand is changing the way energy infrastructure itself is planned. According to the IMF, by 2030 data centres could consume more electricity than all the countries in the world except the three largest emitters — China, India and the United States.

In the US, the problem is already emerging at the level of grid planning. According to Reuters, requests from large electricity consumers, primarily data centres, to connect to the grid in parts of the Midwest, Mid-Atlantic and South exceed 700 GW. That is more than ten times current estimates of data-centre consumption in the country. At the same time, a significant share of those requests could turn out to be so-called “ghost demand” — projects that will never actually be built.

Texas authorities have already paused new data-centre connections and begun checking which projects have genuine financing, customers and actual electricity requirements. In the state, applications to connect large consumers rose from around 48 GW in 2023 to more than 474 GW.

The problem is not simply the scale of consumption. If grid operators build new infrastructure based on projected demand that ultimately fails to materialise, the costs can be passed on to other consumers. In the PJM region, which covers 13 US states and parts of the Mid-Atlantic, rising actual and projected data-centre demand has already contributed to an increase in capacity costs of around $29.4 billion.

This is turning the AI boom from a technology issue into a question of energy policy.

Water is becoming as important as electricity

Data centres need more than electricity. Their servers generate enormous amounts of heat, while cooling systems can require significant quantities of water.

In the UK, the issue has already become a political dispute. The government recognised data centres as critical national infrastructure in 2024. At the same time, water companies have warned that the country may not have enough water to support the sector's planned expansion.

An equally important problem is the lack of transparency. The UK government has been unable to say who uses the country's largest data centres or for what purposes. Hundreds of new facilities are at the planning stage, while 315 are awaiting connection to the national electricity grid.

Governments are therefore trying to plan future electricity and water requirements without having a complete picture of what is being built and who will ultimately gain access to those resources.

From digital progress to political conflict

This is where the main source of the backlash lies.

Local residents are increasingly asking not whether the world needs artificial intelligence, but who is paying for its physical infrastructure.

In London, the government approved a data centre in Tower Hamlets despite opposition from local authorities and residents. The 5,200-square-metre facility is expected to consume around 5.2 MW of electricity. Local campaigners have also raised concerns about its use of water, noise and heat.

At the same time, data centres are increasingly being granted special status. In the UK, the government has allowed such projects to enter the Nationally Significant Infrastructure Projects regime, under which key decisions can be made at national level, bypassing the ordinary local planning process.

A similar trend can be seen elsewhere. The Netherlands has restricted the construction of large data centres, Amsterdam has extended an effective ban on new facilities and expansions until 2030, while US states are moving towards moratoriums, new environmental requirements and greater transparency.

The question is therefore no longer simply whether data centres should be built, but where, at whose expense and under what conditions.

Europe faces a choice between AI ambitions and resource constraints

For Europe, the problem is particularly difficult.

The EU and individual European governments are trying to close the technological gap with the US and China. AI infrastructure is increasingly viewed as part of economic competitiveness, digital security and strategic autonomy.

But a data centre does not, by itself, guarantee the emergence of a European AI champion. Much of the infrastructure is owned by or serves US technology companies. This raises a fundamental question: if Europe bears the environmental and energy costs while control over the technology, intellectual property and profits remains with companies based outside Europe, to what extent does this genuinely strengthen European sovereignty?

The contradiction is already becoming politically visible.

In the UK, the government argues that new data centres will bring jobs, investment and national-security benefits. Critics, however, point out that the state cannot always explain who will use the largest facilities or what public benefit they will generate.

The next battle will be over the rules

The global backlash against data centres is unlikely to stop the development of AI. The economic incentives are too strong, while demand for computing capacity continues to grow.

But it is already changing the conditions under which that expansion will take place.

US states are beginning to demand that companies disclose their actual electricity requirements, sources of financing, ownership structures, water use and tax incentives. Pennsylvania, Texas and other states are seeking to filter out projects that exist largely on paper.

This could become a new model for regulating AI infrastructure: companies receive access to resources and accelerated planning procedures, but in return must demonstrate the project's genuine economic value and take on part of the costs it creates.

The data-centre debate is therefore much broader than an environmental protest. It shows that the digital economy is no longer virtual. Behind every AI query are power plants, transmission lines, cooling systems, water, land and political decisions.

And as artificial intelligence becomes part of strategic competition between states, control over this physical infrastructure could become no less important than control over the algorithms themselves.

Sources: The Guardians, Reuters, UK Parliament

Share