Artificial intelligence (AI) fuels one of Asia’s fastest infrastructure expansions in decades – data centres, which are essential to achieve the promise of AI, such as advances in education, health care, longevity.
Governments are competing to attract them, investors deploying capital at scale, with utilities straining to keep pace.
Coal was once framed in similar terms as providing the necessary infrastructure to fuel growth and advance society. It powered development, created jobs, and underpinned economic expansion.
However, coal brought significant negative environmental impact over time. Opposition first emerged locally, as communities pushed back against mines that consumed land and water and disrupted livelihoods, before widening to the broader damage caused by burning fossil fuels with coal the single biggest contributor to anthropogenic climate change.
We now witness AI generating its own sources of backlash – from job displacement and rising inequality from the implementation of AI itself to a focus on data centres from their growing environmental pressure on air, land, and water. There has been far less focus on mitigating those harms versus just on the benefits AI will bring.
As a result, resistance takes shape, and its first target is not algorithms or models – it’s the physical infrastructure that anchors them: data centres.
They increasingly resemble heavy industry – not in what they produce, but in what they consume and displace. Vast energy and water use, permanent land commitments, and localised disruption are shifting environmental and social costs onto host communities.
If the responsible AI trilemma – environmental harm, increasing inequality and job loss – does not get addressed, the inevitable backlash to AI will come from the social disruption it will bring. Data centres will be where that backlash lands first.
AI demands a great deal of energy – who pays for it matters
Older generations of cloud computing could adjust to power grid capacity. AI cannot. Large models run continuously, draw power relentlessly, leave little room for flexibility and are scaling at industrial speed.
A single large AI cluster can consume as much electricity as an entire city such as Kuala Lumpur. Demand from AI-specific workloads within data centres could quadruple by 2030, with much of the additional power still expected to come from fossil fuels – ironically coal – as nuclear capacity will lag.
The build-out accelerates. By early 2025, nearly 1,200 hyperscale data centres – massive facilities built by major cloud and AI firms – were operating globally, with more than 500 more planned or under construction.
The pattern is familiar. In Asia’s coal era, demand outran infrastructure and regulation – and coal filled the gap. In Thailand, plans for a large coal-fired power plant in Krabi – justified as necessary to meet rising regional demand – triggered years of public opposition over land use, pollution, and local livelihoods, eventually derailing the project.
Data centres face the same fate: build too fast, ignore consent, and growth stalls.
Water: The part of the data centre story few people know (yet)
While energy dominates the headlines about data centres, water incorrectly stays in the footnotes.
Modern data centres built for AI consume vast amounts of water to keep servers cool – often millions of gallons a day, drawn directly from local supplies.
In Asia, those demands collide with reality. Data centres are expanding in cities already struggling with water stress – from India’s tech hubs, where groundwater depletion is chronic, to parts of China where authorities have begun restricting new facilities in water-scarce regions.
Each new centre competes with households, farmers, and natural ecosystems for the same finite resource.
While AI policy may be national, localities control water. Decisions about where to build are still made in town halls, deal by deal – a fast permit here, a tax break there - often without a credible plan for long-term water management in a hotter, drier future.
Coal mining was treated as an energy problem, not a water issue – until rivers were diverted, aquifers were strained, and communities pushed back.
Data centres risk repeating that history. As with coal, local communities – not global users of AI – will feel the costs first.
Ignore water, and the social licence for data centre growth will not last.
When growth hits geography
Technology may eventually solve for energy and water. It cannot solve for land.
Data centres carry physical weight. Even a single centre can require dozens of acres, while hyperscale facilities often sprawl far wider, locking large tracts into permanent industrial use. In dense, fast-growing cities, every parcel allocated to a data centre displaces housing, green space, or future public infrastructure.
Once land is committed, the consequences arrive quickly and close to home.
Round-the-clock cooling brings constant noise; construction and traffic reshape neighbourhoods – the character of entire districts shifts around an asset that never sleeps. Yet AI’s rewards – productivity gains, competitiveness, profits – accrue to those far beyond the places asked to host its infrastructure.
The imbalance mirrors coal’s legacy: local disruption, distant gains.
When infrastructure locks in space but exports value, consent erodes – and once lost, it rarely returns.
Heavy footprint, light employment
If land locks in disruption, jobs are where expectations fall apart. AI may destroy as many jobs as it creates – a wash at the macro level, but little comfort to those displaced by automation.
The backlash to globalisation offers a clear precedent: when promised jobs fail to appear, costs surface as resistance, protest, and political pressure.
Often promoted as job creators, data centres are labour-intensive to build but employ very few people once operational, relative to the land, energy, and water they consume. Unlike coal mines, large AI facilities occupy vast sites while sustaining only a few dozen to a few hundred permanent jobs.
The result is a brittle bargain: significant local disruption in exchange for limited employment. When that imbalance becomes visible, policy hardens quickly.
Time to get ahead of the coming AI backlash
The benefits AI promises will not materialise unless its negative impacts are recognised and addressed early. Those who say, ‘This is not my issue, it is the governments to solve’, miss the obvious.
Coal showed what happens when growth is prioritised while local costs are treated as secondary. As land, water, and environmental pressures mounted, impacts moved from balance sheets to backyards. Communities pushed back, trust eroded, and regulation followed.
Data centres approach that same inflection point. As their energy, water, land, and employment trade-offs become more visible, opposition forms.
In the United States, resistance has emerged over electricity bills, grid strain, noise, land use, and water pressure. In response, companies such as Microsoft have committed to absorbing full electricity costs, replenishing local water use, and paying full property taxes – explicitly to reassure communities they will not be left to “pick up the tab.”
Across Asia, governments are also drawing boundaries earlier: Malaysia now charges premiums for power and water use, while Singapore has capped growth through moratoria and strict efficiency rules.
These are not anti-technology moves. Local costs without consent inevitably trigger resistance and harder policy.
Data centres will drive AI, which will drive growth. But the infrastructure appears to be scaling faster than societies will tolerate, given the overall environmental impact, job loss, and increasing inequality AI will bring.
Grids strain. Water tightens. Land locks in. Where promised jobs fail to materialise, frustration hardens into opposition.
Investors need to realise AI’s concentrating benefits while localising its costs invites resistance. This holds for both the adoption of AI and the source of it. Mitigating those impacts cannot rest with governments alone. Nor can the burden be placed solely on the data centres. Businesses and investors must find ways to share benefits across society, reduce harm – both to the environment and job loss – and create real local value.
If a holistic approach to responsible AI fails to take place, data centres will become the collision point between business, communities, and government – and policy will harden accordingly.
Steven Okun is CEO of APAC Advisors, a Singapore-headquartered consultancy focused on geopolitics and responsible investing. Megan Willis is APAC Advisors’ senior advisor and Noemie Viterale is an associate.
