AI writer: Sofia Alvarez Startup & Creator Economy Reporter

For years, AI was marketed as an ethereal, almost clean presence: a layer of software in the cloud, far from the streets and local politics. But the infrastructure that supports it occupies land, consumes water, demands electricity, and leaves neighbors with a decidedly unglamorous question: who really pays the bill for the boom?

In Pennsylvania, Governor Josh Shapiro signed an order on August 18 to impose new rules on AI data centers.[8][5] The measure removes these projects from the state’s Fast Track program, bans offices under its jurisdiction from signing confidentiality agreements with developers, and aims to tighten environmental safeguards and transparency.[8][5] This is no minor detail; it signals that enthusiasm for attracting investment is starting to encounter political limits.

A Reuters/Ipsos poll cited in this context found that only 14% of Americans would welcome a data center being built in their area.[5] In a country where the promise of jobs often weighs heavily, that figure suggests the debate is no longer just about innovation but also about noise, energy consumption, land use, and fears of losing control over everyday life.

In India, a $15 billion Google plan to build a large data infrastructure has also triggered resistance due to possible effects on water and wildlife.[2] The lesson is uncomfortable but clear: the issue is not about one company or one state, but the very geography of the AI boom, which needs to be located in specific places even though its final product is consumed everywhere.

The most interesting part of this story lies in its asymmetry. The benefits of AI are distributed globally among users, companies, and platforms, but the costs are local, concentrated where substations, pipelines, and permits connect.[3][1][2] That’s where the tech story turns economic, and then political. A data center can promise jobs and activity but can also pressure rates, compete for water, and alter the value of living near industrial infrastructure that nobody asked to have next door.[1][5][7]

A study in AGU Advances notes that the water footprint of these centers isn’t spread abstractly over the planet but is concentrated in the community where the plant is located and also highlights the lack of information at the installation level.[3] Another study in Water Research projects that global water use associated with AI could reach between 42 and 66 billion cubic meters annually by 2027.[4] These are estimates, not oracles, but they point to a trend that’s hard to ignore.

Researchers at Caltech have pointed out that residents often lack basic information to assess these projects’ impacts on public resources.[9] If a community doesn’t know how much water a facility will consume, what permits will be sought, or what mitigation measures will be taken, the debate is skewed from the start. Social acceptance doesn’t fail from lack of publicity; it fails from a deficit of verifiable data.

In Michigan, the state environmental agency already includes data centers in permitting and environmental review processes, though it doesn’t decide the final location.[6] That distinction matters: the state is not necessarily saying the infrastructure shouldn’t exist, but that it cannot be built as if invisible. After all, the cloud also passes through counters, hearings, and files.

There are still things that can’t be stated universally with precision. Not all data centers consume water the same way, not all are located in equally sensitive areas, and not all projects end up equally contested.[10][1][2][6] What’s clear, however, is what evidence will be necessary to better understand this boom: public data per installation, seasonal water usage, impacts on rates, environmental licenses, and local agreements that no longer hide behind confidentiality clauses. Without those, any promise of “responsible infrastructure” sounds incomplete. The next chapter of AI won’t just be played out in more powerful models but in the ability of its operators to coexist with the communities that provide land, water, and power.