When people look back on this decade, they may remember not only the first dazzling AI tools they used, but also something far less glamorous: debates over where to put big, windowless data centers and who should pay their power and water bills.
Those debates may seem technical or even parochial. They are not. They represent an early rehearsal for the kind of AI-enabled society we will choose to build.
Research on AI and the future of work increasingly describes multiple plausible long-run scenarios. In some, automation significantly raises productivity and the gains are broadly shared through strong public institutions, cooperative ownership of digital infrastructure and robust systems such as universal basic income or universal basic services. In others, a small set of firms own most of the “artificial labor” in the form of data centers, models and robots, and the rest of society becomes dependent on these owners, often via the public sector, to fund basic services. These are not predictions, nor inevitabilities. They are possible paths, shaped by concrete governance choices.
And some of the earliest choices are unfolding not in theoretical discussions, but in zoning hearings, utility planning meetings and contract negotiations surrounding hyperscale data centers.
Hyperscalers as an Early Warning System
The major cloud and AI providers (Amazon Web Services, Microsoft Azure, Google Cloud, Meta, Tesla, OpenAI and a small number of others) are building the hyperscale data centers that power today’s AI boom. These facilities are capital-intensive, land-intensive and exceptionally demanding in electricity, water and grid capacity. They also anchor essential digital services for governments, hospitals, banks and schools.
The experience of Virginia, now home to roughly 13% of global data center capacity, provides a preview of both benefits and pressures. Hyperscalers have brought thousands of construction jobs, high-wage technical roles, and in some counties, up to one-third of local tax revenue. Yet unconstrained growth is projected to increase statewide electricity demand anywhere from 40% to over 100% in the coming decade, depending on the forecast model used; a range large enough to reshape grid planning, climate commitments and electricity pricing.
Water and land follow the same pattern. Some facilities rely heavily on evaporative cooling and can consume billions of gallons of water per year; others use air cooling or reclaimed water, depending on region and design. In arid or drought-stressed basins, even moderate volumes can affect aquifer resilience and regional planning. Residents near poorly sited campuses have reported years of construction traffic, low-frequency mechanical noise and nighttime lighting that alters neighborhood character.
These impacts vary widely by technology choice, local climate, energy mix and municipal rules. That variability is key. It means that governance, not technology, determines whether communities experience net benefits or cumulative burdens.
Hyperscalers condense into one visible package many of the questions AI raises more broadly:
- Who owns and governs the infrastructure that enables AI?
- Who captures the economic upside, and who bears the environmental and fiscal costs?
- Are communities genuine partners, or merely hosts?
- And most importantly: what norms of fairness do we establish now, before AI infrastructure becomes too entrenched to reshape?
If we treat hyperscalers as “just another business-attraction effort,” we risk normalizing a future in which AI infrastructure is privately owned, lightly conditioned and loosely connected to broad public benefit. If instead we insist on thoughtful, transparent, reciprocal arrangements now, hyperscalers can become prototypes for a more balanced, democratic and resilient AI economy.
Recommendations: How to Proceed Wisely With Hyperscalers
- Begin with independent analysis and radical transparency
Before incentives or approvals are offered, governments should commission independent studies that integrate:
- Economic modeling, not only for projected jobs and tax revenue but also for long-term opportunity costs and land-use alternatives.
- Long-term grid impacts, including how hyperscaler demand may accelerate the need for new transmission lines, substations or power plants.
- Water-use scenarios, taking into account cooling technology options, seasonal variations, drought cycles and competing municipal or agricultural uses.
- Lifecycle environmental effects, including emissions from construction, backup generators, supply chains and hardware turnover.
- Land-use trade-offs, such as lost potential for mixed-use developments, housing, conservation or other types of industrial activity.
- Environmental justice considerations, recognizing that large industrial facilities often disproportionately affect underrepresented or historically marginalized communities
These analyses should be shared in plain language. Residents deserve to understand not just projected jobs and taxes, but what proposed facilities mean for electricity rates, water resilience, truck traffic, noise and land value over the 10–20 years during which impacts unfold.
Early engagement with affected communities, including tribal nations and historically underrepresented neighborhoods, must be built into the front end of the process rather than bolted on at the end.
- Make communities co-owners of the upside
AI will create economic surplus. The critical question is whether that surplus is concentrated or shared.
Hyperscale negotiations are early opportunities to formalize broad-based benefit sharing:
- Local “AI dividends.”
A portion of data center tax revenues can be directed into transparent community funds supporting broadband, climate resilience, education, childcare and workforce development. Properly designed, these funds resemble modest social wealth mechanisms tied to AI infrastructure.
- Community Benefit Agreements (CBAs).
Negotiated CBAs can include commitments to green space, local business support, noise mitigation, affordable housing funds or school partnerships – with clear reporting and enforceable milestones.
- Workforce pipelines through universities and technical schools.
Internships, apprenticeships, scholarships and training programs ensure local residents gain access to high-skill roles rather than merely living adjacent to the economic value being created.
These mechanisms do not discourage investment; they ensure that growth deepens local capacity rather than extracting value without reciprocity.
- Ensure hyperscalers pay the full and fair cost of energy and water
Electricity demand from AI workloads may grow dramatically, but forecasts differ. What is clear is that utilities often build new generation and transmission capacity specifically to serve hyperscalers. If ordinary ratepayers shoulder the risk of those investments, public resentment and political backlash is predictable.
Key practices include:
- Cost-reflective tariffs.
Large customers should pay rates that reflect the actual cost of their demand, including contributions to new grid infrastructure and safeguards against stranded assets if load projections later fall.
- Flexible load and participation in demand-response programs.
Many data centers can temporarily ramp down during grid stress. This reduces reliance on peaker plants (expensive, high-emission generators used only during peak demand) and helps stabilize electricity prices.
- In-region clean energy procurement and storage.
Commitments tied to new renewable generation ensure that growth aligns with state climate goals rather than undermining them.
- Water stewardship frameworks.
Clear maximum-use thresholds, reclaimed-water preferences, drought contingency plans, and region-appropriate cooling strategies should be standard, especially in stressed basins.
These approaches do not punish hyperscalers. They simply align costs with impacts and prevent cross-subsidization from ordinary residents.
- Use incentives as levers not giveaways
Tax incentives can shape behavior, but only if they are conditional. Many states offer broad sales-and use-tax exemptions worth hundreds of millions of dollars annually. Instead of unconditional subsidies, incentives should be tied to:
- ISO 50001/14001 energy and environmental standards, which require continuous improvement in efficiency and environmental performance.
- Noise, lighting, and design requirements to minimize disturbances in nearby residential areas.
- Cleaner backup generation technologies, replacing high-emission diesel generators with alternatives such as natural gas microturbines or battery storage.
- Regular public reporting on jobs, wages, taxes, water use, emissions and compliance with community commitments.
Well-structured incentives reward high-performing companies and build public trust.
- Invest hyperscaler revenues in public-interest AI
The long-run question is not only how AI is used, but who gets to shape its development.
Cities and states can use a portion of hyperscaler-driven revenue to:
- create regional or municipal AI “public options” supporting small businesses, researchers and civic institutions;
- seed cooperative or community-owned AI ventures;
- fund education and reskilling programs so people can use AI creatively rather than merely be affected by it.
These modest steps diversify the AI ecosystem and ensure that communities possess capacity, not just consumption rights.
Looking Ahead With Cautious Optimism
AI opens the possibility of a society in which we automate much of the necessary labor and expand what people can do with their time: caregiving, creativity, learning, civic life. But that outcome is not automatic, and certainly not guaranteed. It depends on what we do now: in land-use hearings, in utility commission dockets, in incentive negotiations and in the expectations we establish around reciprocity and fairness.
Hyperscale data centers are our canary in the coal mine. They show, early and vividly, how easy it would be to centralize power and externalize costs; or, with care, how possible it is to build arrangements that share benefits, protect communities and strengthen democratic capacity.
If city, state, university and civic leaders approach these decisions with humility and foresight, insisting on fairness without hostility, on partnership without dependence, we can lay the groundwork for an AI-rich future that feels less like a corporate dystopia and more like a renewed social contract.
That future will not arrive fully formed. It will require patient institution-building, wise policymaking, and sustained civic imagination. But it is within reach. And our choices about hyperscalers today may be one of the earliest, clearest tests of whether we are ready for it.