The Data Centre Backlash: Why Social Licence Is Becoming The Real Constraint

For the last two years, AI infrastructure has been framed as a race for GPUs, then a race for power.

Both remain true, but the industry is now discovering a third constraint, public consent.

The data centre sector has moved from being welcomed as invisible digital infrastructure to being challenged as visible industrial infrastructure. Projects are no longer judged only on capacity, speed to market and tenant demand. They are judged on electricity use, water use, grid impact, land use, noise, tax treatment, local jobs and whether the surrounding community believes the project is being done with them or to them. And that shift really matters.

The old model was simple. Secure land. Secure power. Secure planning. Seek incentives. Build large. Point to tax revenue and digital progress.

The new model needs to be different. It needs to start with energy, water, planning credibility and community benefit, not with a press release.

Data centres used to be the prize

In the US, the political mood is changing quickly. The recent Politico and E&E News reporting captured the shift well: data centres were once the kind of investment that states competed to attract, but many are now reassessing the tax incentives that helped bring them in. According to the National Conference of State Legislatures, lawmakers introduced proposals in at least 28 of the 38 states with data centre incentives that would substantially amend existing tax breaks. At least nine states have considered bills to repeal those incentives entirely.

This is not a fringe issue, it is becoming mainstream politics because the scale has changed. The International Energy Agency estimates that data centres accounted for around 415 TWh of electricity consumption in 2024, about 1.5% of global electricity use. By 2030, that could more than double to around 945 TWh. The global figure is manageable in aggregate, but local concentration is the real problem. Data centres cluster around specific substations, water systems, fibre routes and planning authorities, which makes their local impact much more acute than their global share suggests.

That is why the backlash is not simply anti-AI, it is a reaction to local scarcity.

Communities are asking a fair question: if a campus uses the power of a town, the water of a neighbourhood, the land of an industrial estate, and the tax incentives of a public authority, what exactly does the community get back?

The resentment is rational

The industry often talks about public resistance as if it is a communications problem, it is not. It is an infrastructure trust problem.

Recent polling from Gallup found that seven in ten Americans oppose AI data centres in their local area, with almost half strongly opposed. The reasons are not abstract. Opponents cite excessive use of water and energy, pollution, noise, quality of life, higher utility bills, increased cost of living, traffic and the use of taxpayer funds. Supporters, by contrast, tend to focus on economic benefits, particularly jobs and tax revenue.

That tells the industry something important. People are not rejecting data centres because they do not understand them. They are rejecting projects where the risk feels local, but the benefit feels remote.

The most common objections are now familiar:

Power comes first. Communities worry that large loads will increase bills, consume grid capacity, delay housing and industrial connections, and force new transmission, substations or generation assets that local people ultimately pay for.

Water comes next. In water stressed regions, data centres are treated as competitors with homes, farms and local businesses. Even when developers argue that water use is limited, communities often do not trust the claim unless consumption is independently measured and publicly reported.

Noise matters more than the industry often accepts. Backup generators, testing regimes, construction traffic, fans, substations and operational hum can become planning issues, particularly near homes, schools and sensitive sites.

Land use matters. Communities object when large campuses appear to consume farmland, green belt, conservation land, culturally sensitive sites or land that might otherwise support housing, local enterprise or public infrastructure.

Transparency matters. Non-disclosure agreements, opaque tax packages and late consultation create the perception that decisions have already been made before communities are invited into the conversation.

Jobs matter, but the promise has to be credible. A hyperscale campus may be capital intensive, but it does not always create the level of permanent local employment that residents expect from the size and cost of the build. Where incentives are large and job numbers are modest, resentment grows.

Finally, fairness matters. If a project receives tax relief, discounted land, fast-tracked approvals or grid priority, the public will ask whether that support is proportionate to the local return.

Projects are now being stopped

This is no longer a theoretical risk. In the US, Heatmap reported that at least 25 data centre projects were cancelled in 2025 following local opposition, four times as many as in 2024. Those cancelled projects represented at least 4.7 GW of electricity demand, and at least 99 projects were being contested by local activists or residents.

In Western Australia, GreenSquare withdrew plans for a 120 MW data centre in Hazelmere, near Perth, after almost 1,900 public submissions and council concerns about diesel generator noise, proximity to a school and residences, and the site’s location near a river, conservation wetland and Aboriginal heritage sites.

In Chile, a court partially reversed Google’s permit for a planned Santiago data centre, requiring a reassessment of climate change impacts on the Central Santiago Aquifer. The project had drawn concern from residents and officials because Chile had been suffering from drought for more than a decade, and cooling water had become a politically sensitive issue.

In the Netherlands, Meta suspended plans for a large Zeewolde data centre after political opposition, including a Dutch Senate motion asking the government to temporarily block the project. The proposed site covered 166 hectares of farmland, and opponents raised concerns about power consumption, the use of limited renewable energy and whether residents had been properly heard.

Ireland shows what happens when national scale becomes local tension. The Irish Central Statistics Office reported that data centres used 22% of metered electricity consumption in 2024, up from 5% in 2015, and more than the 18% used by urban households.

The UK is moving into the same debate. Parliament’s Environmental Audit Committee is now examining how data centres affect energy use, water use, net zero goals and planning. A Hoare Lea review of 33 disputed UK applications found that data centre planning applications were being delayed by an average of 490 days, with objections linked to inadequate community engagement, unclear community benefits, design, infrastructure constraints and energy use.

The message is consistent across markets. The limiting factor is no longer just land, fibre or capital, it is credibility.

The industry needs to stop treating communities as a planning hurdle

The conventional playbook no longer works, in fact it has the opposite effect. It makes communities feel like they are an inconvenience to be addressed after the fact, which only increases resentment.

A community exhibition after the site has been selected is not engagement. A promise of “green jobs” without numbers, training pathways or local procurement is not economic development. A renewable power purchase agreement is not the same as proving that a project will not increase local grid constraints. A sustainability page is not a water strategy. A voluntary community fund is not the same as a binding community benefit agreement.

The industry needs to move from persuasion to proof. That begins with a simple principle: a data centre should not be a net burden on local infrastructure, it should bring benefits: Additional power. Additional resilience. Additional tax base. Additional jobs. Additional skills. Additional digital capability. Additional local enterprise.

If those benefits cannot be defined, measured and enforced, communities will increasingly say no.

A better model starts behind the meter

Behind the meter energy generation is not a silver bullet, but it changes the conversation when it is done properly.

If a campus can bring new generation, storage and flexibility, it can reduce pressure on the grid rather than simply joining an already constrained connection queue. The strongest model is not just “we will buy renewable certificates”, it is a physical energy strategy that shows where the power comes from, how it is firmed, how much is imported, how much is exported, how the site behaves at peak times, and how it supports rather than destabilises the local network.

The IEA has warned that around 20% of planned data centre projects could face delays unless grid risks are addressed, with transmission lines often taking four to eight years in advanced economies and critical grid components facing longer lead times. That is why energy first development matters.

A credible campus should be able to show that it is not just consuming power, but helping to finance and accelerate the energy infrastructure around it.

That could include onsite generation, co-located renewables, battery storage, grid services, heat reuse, demand flexibility and transparent reporting on grid imports. It should also mean paying a fair share of reinforcement costs, not pushing them invisibly into household bills.

The distinction is important. Behind the meter power that simply locks in fossil fuel generation and seeks subsidy may make public opposition worse. Behind the meter power that adds clean capacity, storage and flexibility can make a project more credible.

Water has to be designed out, not explained away

Water is becoming one of the strongest drivers of planning resistance.

The industry cannot ask water stressed communities to trust vague claims about efficiency. It needs to design for minimal potable water use, then publish the data.

This is where liquid cooling, closed loop cooling and new thermal architectures matter. Liquid cooling is often discussed as a density tool, but it is also a community tool. Used properly, it can reduce reliance on evaporative cooling, support higher density in a smaller footprint, improve heat recovery potential and materially reduce water consumption.

New building designs are emerging that take this further. ServerDomes, for example, claim up to 34% less energy consumption, around a 92% reduction in water use, and more sustainable construction compared with traditional designs. Those are vendor claims and should always be verified at project level, but the direction is the right one: lower overhead, lower water, smaller phased modules and measurable operating performance.

The industry should not hide behind averages. Every project should disclose expected water use, potable water use, cooling method, drought response, water source, discharge assumptions and ongoing reporting. In sensitive regions, the bar should be even higher.

No community wants to be told that a facility is “sustainable” while local residents face drought restrictions or rising water bills.

Community benefit must be binding

The most important change is not technical, it is contractual.

A binding community benefit agreement should become standard for serious projects. Not a voluntary donation. Not a one off sponsorship. Not a glossy social value statement. A binding agreement that defines what the community receives, when it receives it, who is accountable and how performance is reported.

For blocz, this is central to how we believe data centre infrastructure should be delivered.

A community benefit agreement should include commitments on local jobs, apprenticeships, technical training, local procurement, digital inclusion, startup funding, community energy benefit, transparent tax contribution and long term reporting. It should also include mechanisms for local oversight, so the community can see whether commitments are being met.

That is how trust is built. Not by asking communities to believe in future benefit, but by making the benefit enforceable from the beginning.

Tax revenue, not subsidy capture

The public is increasingly sceptical of subsidy led data centre development.

That scepticism is justified when tax breaks are large, local jobs are limited and the wider infrastructure cost is unclear.

A project that claims to create local value should be able to show its fiscal impact clearly. That means explaining the expected tax uplift, any incentives being sought, the net benefit to the local authority, the cost of grid or water reinforcement, and how public support is linked to public value.

At blocz, the aim is to increase tax revenue, not extract subsidy.

That distinction makes a huge difference. The industry needs to avoid the perception, and the reality, of subsidy capture. Public support should not be used to underwrite private infrastructure unless the community receives a clear, measurable and durable return.

What the industry needs to do differently

The next generation of data centre campuses needs a different planning doctrine.

First, energy strategy has to come before planning strategy. Developers need to prove that power is additional, resilient and fair to the local grid.

Second, water strategy has to be specific. Low water, closed loop and liquid cooled designs should be the default wherever possible, with transparent public reporting rather than vague assurances.

Third, site selection has to become more disciplined. Sensitive land, poor buffers, constrained water systems and weak grid locations create opposition for good reasons.

Fourth, communities need to be engaged before the scheme is effectively fixed. The public should not be asked to comment on a decision that already feels predetermined.

Fifth, benefits need to be binding. Jobs, training, startup funding, local procurement, digital inclusion and community investment should be written into enforceable agreements.

Sixth, tax and subsidy positions need to be transparent. If a project is asking for public support, it should show public value.

Finally, the industry needs to stop greenwashing. Communities are increasingly sophisticated. They can distinguish between renewable branding and real additional generation, between efficiency claims and measured performance, between community language and actual community power.

The future belongs to infrastructure that communities can believe in

Large data centre campuses are not going away.

AI, cloud, sovereign compute, scientific research, digital public services and industrial automation all need physical infrastructure. The question is not whether data centres will be built. The question is what kind of data centres will still be allowed to be built.

The answer will not be found in faster lobbying or better messaging, it will be found in better projects.

Projects that bring their own credible energy strategy. Projects that reduce grid strain rather than intensify it. Projects that minimise water consumption by design. Projects that use high density, efficient cooling rather than outdated thermal assumptions. Projects that create technology jobs, fund local startups, support skills and apprenticeships, and increase the local tax base. Projects that put community benefit into binding agreements rather than marketing language.

That is the model blocz is working towards. Not greenwashing. Not subsidy capture. Not asking communities to accept concentrated infrastructure costs in exchange for vague future benefit.

A different model is possible: energy first, water responsible, transparent, locally accountable and commercially serious.

At this scale, data centres will never be invisible, but they have to be credible. And credibility starts with building in a way that communities can support because they can see, feel, measure, and enforce the benefit.