AI & Environment · Part 4 of 5
In Prince William County, Virginia, developers have proposed data centres across roughly 2,400 acres of farmland the county's own planners had designated to stay undeveloped. That single project captures the question this chapter is built around: who actually benefits when AI infrastructure lands somewhere, and who's left holding the cost?
This isn't a call to stop building. It's a call to look honestly at both sides of the ledger — the communities absorbing land loss, dust and rising bills, and the genuine environmental gains AI is already delivering elsewhere, from wildfire detection to precision irrigation.
A data centre campus rising out of what used to be open farmland — the land side of AI's physical footprint. Source credit to be added on upload.
Part 4 of the AI Civilization Balance Sheet series. Here's where this chapter takes you:
500 to 800 Acres at a Time
A modern AI-scale data centre typically needs 500 to 800 acres of land, and the biggest campuses now being built dwarf that figure entirely. Meta's Hyperion campus in rural Louisiana was originally planned for 2,250 acres and has since expanded to 3,650 — roughly the footprint of 2,765 football fields, and about twice the size of the state's international airport.
Because land that's cheap, flat and close to existing power infrastructure is exactly what rural counties tend to have, developers are increasingly siting these campuses on farmland. Around 39% of currently planned data centres are going into counties that have never hosted one before, and the total US count is on track to jump from roughly 4,100 active facilities to nearly 7,000 as thousands more are completed or begin construction.
This isn't happening at the margins of the AI story — it's the physical expression of everything covered so far in this series. Every gigawatt of electricity from Part 2 and every gallon of water from Part 3 has to land somewhere, on land that used to be something else.
↑ Return to NavigatorWho Pays When the Data Centre Moves In
A 2026 survey by researchers at the University of Illinois and Purdue University found rural Americans consistently more worried about data-centre impacts than urban or suburban respondents — and their top concern, ahead of even water use, was rising electricity bills. That worry has evidence behind it: one Virginia legislative analysis projected typical residential electricity bills could climb by $14 to $37 a month by 2040, a 9% to 25% increase, purely from grid strain tied to data-centre growth.
The bill isn't hypothetical for everyone. One Manassas, Virginia resident who'd lived in his home for nearly forty years described opening a January electricity bill for $281, more than double what he'd paid the previous month, with no clear explanation beyond the AI-driven demand reshaping his local grid. Multiply that story across a county, and it becomes a housing-affordability problem, not just an inconvenience.
Land loss compounds the financial pressure. In places like Illinois, Indiana, Michigan, Minnesota and Wisconsin, data-centre energy demand is projected to more than double or even triple by 2030, and much of the land absorbing that growth is active farmland — the same land that historically anchored these communities' economies and identities. Add in the truck traffic and diesel exhaust that construction phases bring, and the health and quality-of-life costs start to look as real as the financial ones, particularly for children and older residents living nearby.
↑ Return to NavigatorThe Environmental Justice Question
Strip away the technology and this is a familiar pattern in industrial development: the economic upside — jobs, tax revenue, cloud services used by people thousands of miles away — flows broadly, often nationally or globally, while the physical cost concentrates hyper-locally, in one county, on one grid, on one aquifer.
Data centres do bring real local benefits. They're capital-intensive enough to generate significant property tax revenue, and developers frequently secure zoning changes and tax incentives specifically because local governments want that revenue. But those same incentives mean the facility often pays a smaller effective share of the infrastructure burden it creates, leaving other ratepayers and residents to absorb more of the grid, water and road costs than the deal on paper suggests.
Worth naming directly: this pattern doesn't fall evenly by geography or income. Rural and lower-income communities, with less bargaining leverage and often less legal capacity to negotiate development agreements, tend to be the ones offering the cheap land and cheap power in the first place — which is exactly why the environmental-justice framing matters here, not just as an ethical add-on but as a predictor of where future conflict is most likely.
↑ Return to NavigatorWhere AI Is Genuinely Helping
None of this should tip into treating AI as a purely extractive technology, because that would be just as dishonest as ignoring its costs. Set against the land and community costs in this chapter is a real, growing list of environmental gains AI is delivering right now.
On electricity grids, AI-driven forecasting is improving renewable integration, demand prediction and fault detection, helping grid operators absorb more solar and wind without destabilising supply. In buildings, AI-optimised heating, cooling and lighting is cutting energy waste at scale. In agriculture, precision irrigation and AI-based pest and yield monitoring are reducing water and fertiliser use on the very farmland that data centres are competing for. In climate science, AI is measurably improving weather prediction, wildfire detection and flood forecasting — capabilities that save both lives and property as extreme weather becomes more frequent.
These aren't marketing claims; they're the same category of measurable, sector-specific gain this series flagged back in Part 1 as AI's real value proposition. The honest accounting isn't AI's environmental footprint minus zero — it's AI's footprint minus the emissions and resource use it's demonstrably helping other sectors avoid. That subtraction is harder to do precisely than either side alone, but it's the only version of the number that means anything.
↑ Return to NavigatorWeighing Both Sides Honestly
Land is where every earlier chapter in this series becomes visible and personal. Electricity, water, and minerals can feel abstract until they show up as a fenced 800-acre campus next to a farm that's been in a family for generations, or a $281 electricity bill that didn't used to exist. That's what makes the land chapter different from the ones before it — the costs stop being statistics and start being someone's specific street.
At the same time, dismissing AI's environmental contributions because its infrastructure has real local costs would throw away a genuinely useful part of the picture. Grid optimisation, precision agriculture and better climate forecasting are not hypothetical; they're operating today, at meaningful scale. The fair version of this story keeps both facts in view rather than picking whichever one makes a better headline.
Part 5 closes this series by putting that tension to its hardest test — looking at two of AI's strongest possible cases, drug discovery and space science, before pulling every thread from this series together into a country-by-country comparison and a final verdict on where the AI Civilization Balance Sheet actually comes out positive.
K M Manohar’s Insight
AI should not be judged only by what happens inside the data centre.
The real test is what happens around it — to the land, the water, the electricity grid, and the people who live there.
If a community provides the land and resources, it should not receive only the disruption while the economic value travels elsewhere.
But the opposite extreme is equally dangerous: rejecting AI’s genuine environmental gains simply because its infrastructure has costs.
My view is simple: AI must create a better balance, not merely a bigger footprint.
The communities hosting this infrastructure deserve transparency, a meaningful share of the benefits, and a voice in decisions that permanently reshape their future.
And AI companies must increasingly prove not just that they can build faster, but that they can build fairer, cleaner, and more responsibly.
The future should not be AI versus communities.
It should be AI working with communities — while making the gains measurable and the costs accountable.
That, to me, is the real meaning of an AI Civilization Balance Sheet.
↑ Return to Navigator
Comments
Post a Comment