TecEx | When the Perfect Data Center Site Isn’t the Perfect Site

The race to build the infrastructure required for artificial intelligence is accelerating. Billions are being committed to data centers, power infrastructure and the compute required to train and run increasingly sophisticated AI models.

At the same time, communities are questioning, opposing and, in some cases, attempting to stop the data centers being proposed in their towns and cities.

That presents an uncomfortable contradiction. We are demanding more AI, more compute and more digital services, while some of the communities being asked to host the physical infrastructure behind them are saying no.

Are communities opposing the data centers actually being built today, or the data centers they think are being built? And where their concerns are accurate, has the extraordinary scale of AI made some of those concerns more relevant than ever?

But perhaps the commercial question is not simply how developers engage communities once a site has been selected. It is whether community compatibility should help determine which sites are selected in the first place.

The data center has changed

Data centers are not new. But the facilities being built today can be very different from those built 20 years ago.

In one of my conversations for this article, Brian Galvin, Principal of PillarSite Strategies LLC, made a useful comparison: judging today’s data centers by the facilities of the past can be a little like comparing an older car with one built today.

There is some truth to that.

Data-center design, computing technology, cooling and energy efficiency have evolved considerably. Power Usage Effectiveness, or PUE, is one measure of facility energy efficiency. Uptime Institute’s annual survey shows average PUE falling from 2.50 in 2007 to 1.56 in 2024.

Cooling technology has evolved too. New approaches can improve efficiency, although the water implications of any facility remain highly dependent on its cooling system, climate and design.

Perhaps some perceptions of data centers have not kept pace with the technology.

But Galvin’s car comparison led me to another question.

What happens if the new car is considerably more efficient than the old one, but instead of one car, we now need ten?

The scale problem

AI has changed the equation.

As AI adoption spreads across businesses and everyday life, demand for the compute needed to train and run AI models is growing rapidly. Efficiency is improving, but more people and businesses are using AI and increasingly powerful applications require substantial computing resources.

The result is that greater efficiency does not necessarily mean lower overall electricity consumption.

The International Energy Agency estimates that global data-center electricity consumption was around 485 TWh in 2025 and projects it to reach approximately 950 TWh by 2030. Electricity consumption from AI-focused data centers is projected to triple over the same period.

Ireland provides a striking example of what concentration can mean. According to Ireland’s Central Statistics Office, data centers accounted for 23% of the country’s metered electricity consumption in 2025, up from 5% in 2015.

Those numbers complicate the argument that community concerns are simply based on an outdated understanding of data centers.

An individual facility may be considerably more efficient than its predecessor while the collective infrastructure requirement becomes much larger.

To return to the car analogy, we need more cars because there are more drivers.

The question is whether there is enough road for all of them.

What are communities actually worried about?

Community opposition is becoming a material development issue.

EPRI’s 2026 research into community-centered data-center development found that at least 25 US projects were cancelled because of local opposition in 2025, compared with six in 2024 and two in 2023. It estimated that $98 billion in US projects were blocked or delayed between March and June 2025.

The concerns behind that opposition vary significantly by location.

Power is an obvious issue, particularly where additional infrastructure is required and residents want to understand how the costs will be allocated.

Water is another, although the impact is highly dependent on facility design. Noise, land use, environmental impact and changes to the character of an area can also become important.

Then there is the economic question.

A data center can bring construction activity, tax revenue and other economic benefits. But those benefits vary by project, as does the number of permanent jobs that remain once construction is complete.

For communities, these are reasonable questions.

The difficulty is that they may be asked to assess highly technical proposals against compressed development timelines, confidential negotiations and incomplete information. Research from Brookings into rural US communities has highlighted how limited local expertise and opaque negotiations can contribute to distrust.

Perhaps some opposition is therefore not opposition to data centers themselves. It is opposition to uncertainty about what a particular development will mean for a particular community.

Development with communities, not simply in them

This question came through repeatedly in the expert perspectives I gathered while researching this article.

Nancy Coblenz, Co-Founder & President of Stellenium Corporation, drew an important distinction between obtaining the legal permission to develop and achieving genuine community acceptance. A project can satisfy the formal requirements for approval without necessarily addressing whether the people who will live alongside it understand or support what is being proposed.

Loring Tripp III, Owner/Consultant and Lead Planner at Loring Tripp Consulting, approaches the issue through a broader development and historical lens. His argument is that infrastructure development becomes more difficult when communities feel development is being done to them rather than with them.

There is a useful historical principle here.

Major infrastructure transitions have long created tension between broad societal benefits and concentrated local impacts. Railways transformed economies and the movement of people and goods, while individual railway projects could also cause disruption and encounter local opposition.

Data centers are not railways. But the underlying tension is recognisable.

AI can create benefits across countries and continents. The power infrastructure, buildings and other physical systems required to support it exist somewhere very specific.

AI may be global. Its infrastructure is always local.

What does the community get?

If a community is going to host a large industrial facility that occupies land and may place substantial demands on local infrastructure, it is reasonable to ask what that community receives in return.

There may be construction employment, permanent jobs and tax revenue. There may also be opportunities for infrastructure investment, workforce development, education, broadband or other benefits identified by the community itself.

But these cannot simply be generic promises.

EPRI’s research found that community-benefit claims can lose credibility when accountability is unclear, while projects achieving community approval tended to be associated with earlier engagement, greater transparency and visible benefits tailored to the community.

This changes the conversation.

Instead of simply explaining why data centers are important to the wider economy, developers need to understand what matters to the particular community being asked to host one.

If the benefits of AI infrastructure are global while many of its immediate impacts are local, what does the host community reasonably deserve in return?

When uncertainty becomes policy

The consequences of failing to answer these questions are extending beyond individual planning disputes.

By July 2026, lawmakers in 15 US states had considered proposals for bans or moratoriums on data-center development, although the proposals varied substantially and many had not become law.

A moratorium should not automatically be interpreted as an anti-technology position.

It can give policymakers and communities time to understand infrastructure requirements, collect information and develop appropriate rules. But stopping development temporarily does not resolve the underlying questions.

When information does not keep pace with development, uncertainty can become opposition, and opposition can become policy.

Across the Atlantic

The underlying issues are not uniquely American or European, but the regulatory structures differ.

The European Union has established a common reporting framework requiring operators of qualifying data centers with installed IT power demand of at least 500 kW to report information including energy and water consumption to a European database.

The US response is more fragmented, with states and local jurisdictions developing different approaches to data-center growth.

This raises an important commercial question.

If neighbouring jurisdictions take significantly different approaches to data-center development, could community attitudes and regulatory policy increasingly influence where investment goes, alongside traditional considerations such as power, land, water, fibre, permitting and cost?

A technically attractive location may not be commercially attractive if significant opposition emerges after substantial capital and time have already been committed.

Community due diligence

This is ultimately where my research and conversations have led me.

Before committing billions of dollars to a data-center site, developers undertake extensive due diligence. They assess power, land, water, fibre, permitting, tax, construction risk and economics.

Perhaps community risk now deserves a place on that list.

Before final commitment to a site, a developer should understand the attitude of the community that will host the project. Does it welcome data-center development? Does it oppose it? Or is it simply uncertain because residents do not yet understand what is being proposed?

That process cannot simply be about persuading a community to support a predetermined development.

Developers need to listen, understand the concerns and provide credible answers. Those discussions also need to involve people with enough authority for what they learn to influence the project.

Most importantly, the findings need to matter.

If a community has legitimate concerns about water, power, noise or another impact, the developer should be prepared to ask whether the design can change.

If the community will carry significant local impacts while much of the economic or technological benefit flows elsewhere, the developer should consider what tangible value can remain locally.

And if those concerns cannot reasonably be resolved, the developer may ultimately need to reconsider the site.

That may sound expensive. But discovering that a community and a project are fundamentally incompatible before final site commitment may be considerably less expensive than discovering it after significant capital, resources and time have already been committed.

Community engagement therefore should not simply be a communications exercise designed to win support for a predetermined project.

It should become part of commercial due diligence.

The next generation of data centers will still need the right power, land, water and fibre. But perhaps the definition of a data-center-ready location now needs to include something else: whether the project and the community that will host it can actually work together.

Because a technically perfect site that cannot ultimately be built may not have been the perfect site after all.

 

By Noa Sussman, Director of Global Expansion, AI & HPC, TECEX

 

 

 

Compliments of TecEx – a member of the EACCNY