As artificial intelligence drives demand for computing power, data centers are spreading rapidly across the United States. But communities are increasingly questioning the cost of this expansion, from rising electricity demand and water consumption to pollution and pressure on public infrastructure.
Technology critic Paris Marx has spent years examining these conflicts. In his new book, Hyperscale: The Ambition and Excess of Big Tech’s Data Empires, he argues that America can learn important lessons from countries that have already experienced the consequences of large-scale data center development.
Marx explains how resistance in Ireland, the Netherlands and elsewhere offers a preview of the political and environmental challenges now facing the United States.
Data center opposition didn’t begin in America
For many Americans, opposition to data centers may appear to be a recent development. But other countries have been confronting the consequences of rapid expansion for years.
Marx points to the Netherlands and Singapore, where moratoriums on data center development date back to 2019. In Ireland, concerns grew as facilities began placing increasing demands on the electricity grid.
Another important example emerged in The Dalles, Oregon, where Google faced scrutiny over its water consumption. According to Marx, the company sought to keep its water-use figures confidential, and the local community fought for nearly a year to obtain the information.
Residents eventually learned that Google’s facilities were using approximately 29 percent of the community’s water.
The episode highlighted a broader concern: large technology companies can become major consumers of essential local resources, while residents may struggle to obtain clear information about the costs.
When ChatGPT launched in late 2022, the expected growth of generative AI made the issue even more urgent. AI systems require substantial computing infrastructure, raising the prospect of more facilities being built in communities already experiencing resource pressures.
Ireland shows what happens when the electricity grid reaches its limits
Ireland became one of the clearest examples of the tensions surrounding data center expansion.
According to Marx, Dublin introduced restrictions on new data center connections after the electricity grid operator warned that the system lacked sufficient capacity to accommodate additional facilities.
Rather than abandoning their plans, operators began looking elsewhere in the country. Some also explored alternative ways to secure electricity, including gas generators and dedicated gas-fired power plants.
That shift raises an important question: if a data center cannot obtain enough power from the existing grid, should companies be allowed to build private energy infrastructure to support their operations?
Marx says data center companies also lobbied for changes that would permit private firms to build electricity cables, challenging Ireland’s traditional model in which public utilities control the construction of electricity lines.
Critics worry that these changes could reshape who controls essential infrastructure and who benefits from it.
The Irish experience is particularly relevant to the United States, where technology companies are increasingly exploring dedicated power plants to meet the enormous energy demands of AI.
The climate cost of powering AI
The move toward dedicated gas power plants has implications beyond electricity supply.
Marx notes that Irish data center operators obtained emissions permits allowing them to burn fossil fuels to power their facilities. Ireland’s environmental regulator warned that this additional fossil fuel use could make the country’s climate targets harder to achieve.
The same tension is emerging in the United States. Companies are racing to secure dependable electricity for AI infrastructure, sometimes turning to natural gas as they face delays or limitations in connecting to the grid.
The challenge is that the energy needed to support AI growth may conflict with commitments to reduce emissions.
The debate is therefore not simply about whether a country has enough electricity. It is also about how that electricity is generated, who pays for new infrastructure and whether the environmental consequences are justified by the benefits of the technology.
What happens if the AI boom collapses?
Marx believes the AI industry may be experiencing a bubble, although the timing and consequences of a potential downturn remain uncertain.
Companies are planning enormous investments in data centers, power plants and related infrastructure. Yet these projects depend on several factors, including continued demand for AI services, access to advanced chips and the availability of specialized equipment such as power-generation turbines.
If demand falls short of expectations, some planned projects could be delayed, scaled back or abandoned.
There is historical precedent. During the dot-com boom, companies invested heavily in fiber-optic networks and data centers. When the bubble burst, some businesses failed, and surviving companies acquired infrastructure that could later serve renewed demand.
But Marx argues that the current build-out may have a different weakness: AI hardware can become outdated quickly.
Even if a data center remains standing after an industry downturn, its equipment may no longer be competitive when demand returns. The facility’s long-term value could depend on whether it can accommodate new generations of chips and computing systems.
Power plants create another complication. If facilities are built specifically to supply data centers and those customers disappear, the plants will still exist. Their owners may seek new customers or other ways to justify the investment.
The result could be expensive infrastructure that communities must live alongside long after the original business case has weakened.
New Zealand offers a different vision for AI infrastructure
Marx’s book does not focus exclusively on opposition. It also considers alternatives to a model in which large technology companies build infrastructure primarily to serve their own commercial interests.
One example comes from New Zealand, where a community-led project worked to develop more accurate speech-recognition technology for the Māori language, using local cloud providers to host the data.
The project represents a different approach to AI development: using computing resources to support a specific community need and preserve a language, rather than building infrastructure solely to pursue the largest possible commercial market.
It also raises questions about who should control the data used to train AI systems, where that data should be stored and who should benefit from the resulting technology.
For Marx, examples like this demonstrate that the future of AI infrastructure does not have to follow a single model dominated by the largest technology companies.
Governments need to regulate both data centers and AI products
Marx argues that governments should impose stronger rules on data center development while also regulating the AI products that create demand for the infrastructure.
He believes policymakers have accepted ambitious promises about generative AI without adequately examining whether the technology is delivering sufficient public benefit to justify its costs.
His central questions are straightforward:
-
How much computing power does society actually need?
-
How much electricity, water and land should be devoted to AI infrastructure?
-
What measurable benefits are these systems delivering?
-
Are those benefits worth the financial, social and environmental costs?
These questions matter because the consequences of infrastructure decisions extend beyond the technology companies making the investments. Local residents may experience pressure on water supplies, electricity systems and air quality, while governments must weigh economic development against climate commitments and other public priorities.
Marx argues that a more deliberate approach could help communities distinguish genuinely useful technology from expansion driven primarily by commercial expectations.
Europe faces the same difficult trade-offs
The conflict between AI growth and environmental goals is not confined to Ireland or the United States.
As WIRED’s reporting notes, European countries face competing priorities: national governments may want to attract investment and expand AI capabilities, while local communities seek to protect resources and environmental standards.
Planning rules can slow construction, and climate commitments may limit the options available for supplying additional electricity.
These competing interests could make data centers an increasingly important political issue as European countries approach future elections.
The underlying dilemma is difficult to resolve. Governments want the economic opportunities associated with AI, but rapid infrastructure expansion can impose costs that are felt locally while the benefits may be distributed more widely.
The future should be about more than building bigger.
The international experience outlined in Marx’s book suggests that communities do not have to wait until data centers overwhelm local infrastructure before asking questions about their development.
Ireland’s electricity constraints, the controversy over Google’s water consumption in Oregon and New Zealand’s community-focused language project illustrate different parts of the same debate: who decides how computing resources are used, and whose interests should come first?
Marx’s argument is that governments need to assess the infrastructure and the technology together. Regulating data centers without examining the AI products they support could leave the underlying demand unchecked. Conversely, encouraging AI development without considering electricity, water and emissions risks overlooking the costs imposed on communities.
The bottom line: The global backlash against data centers offers an early warning for the United States. The challenge is not simply to build enough infrastructure to support AI, but to decide which projects are worth building, how they should be powered and whether the benefits justify the resources they consume.
As the AI boom accelerates, communities may increasingly demand answers before approving the next wave of construction.