“AI and the environment” is seven arguments wearing one coat.
Some of them are global. Most of them are not. Which one you’re actually asking about determines what the evidence says, who has the answer, and what you can do about it.
Built on federal energy reporting, peer-reviewed water and health research, open hydrological data, state statute, and live utility dockets. Each guide names where the evidence is strong, where it is contested, and where we haven’t done the work.
Seven questions, seven different scales.
Arguments about AI and the environment go wrong when a claim that's true at one scale gets asserted at another. Both of the usual mistakes sound like conviction.
| # | The question you’re actually asking | True at the scale of | Where the evidence stands | |
|---|---|---|---|---|
| 1 | How much electricity does this take?The one most people start with, and the best measured. | National totalA country-wide aggregate. Real, and it says nothing about where the load lands. | Well establishedLBNL puts US data centers at 4.4% of national electricity in 2023, headed for 6.7–12% by 2028. Federal, and it says nothing about where the load lands. | Read the guide → |
| 2 | What does that electricity actually emit?Not the same question as the first one, and the answer isn't in the first number. | GlobalThe atmosphere is genuinely shared. A ton emitted anywhere counts everywhere. | UnresolvedDepends entirely on grid mix and on what the marginal plant is. The marginal-emissions question is unresolved and we say so. | Read the guide → |
| 3 | How much water, and whose?Where the famous statistic lives — and where it was retired. | WatershedA river basin or aquifer. Water does not move between them. | Well establishedReal as a peak and local problem. The per-query figure everyone quotes was withdrawn by the researchers who produced it. | Read the guide → |
| 4 | What does it do to the air people breathe?The least discussed and among the better substantiated. | AirshedDownwind counties. Pollution travels, but not evenly and not forever. | Well establishedSubstantiated and scenario-dependent, and almost always misreported — the burden is dominated by generation, not by backup diesel. | Read the guide → |
| 5 | Who pays for the buildout?Being decided right now, in rooms most people don't know exist. | Utility territoryThe defined set of customers on one utility's bill. | ContestedActively contested in live rate cases and dockets. Attribution of specific retail increases to data centers is disputed by serious people. | Read the guide → |
| 6 | Who decided this gets built here?The most local question, and where most actual decisions get made. | County / parcelZoning, permits, and who got asked. | Well establishedThe accountability gap is documented: tax determinations that require nothing in return, and use figures that are contractual trade secrets. | Read the guide → |
| 7 | What did it take to make the hardware?Mining, materials, and e-waste — global in flow, local in harm, mostly displaced abroad. | Supply chainGlobal in flow, concentrated in harm, and usually somewhere else. | We haven't researched thisCPAI has not done this research, so there is no guide. Building a thin page on material we haven't verified would break the rule this whole section is built on. | Tell us if you need this → |
The one most people start with, and the best measured.
LBNL puts US data centers at 4.4% of national electricity in 2023, headed for 6.7–12% by 2028. Federal, and it says nothing about where the load lands.
Read the guide →Not the same question as the first one, and the answer isn't in the first number.
Depends entirely on grid mix and on what the marginal plant is. The marginal-emissions question is unresolved and we say so.
Read the guide →Where the famous statistic lives — and where it was retired.
Real as a peak and local problem. The per-query figure everyone quotes was withdrawn by the researchers who produced it.
Read the guide →The least discussed and among the better substantiated.
Substantiated and scenario-dependent, and almost always misreported — the burden is dominated by generation, not by backup diesel.
Read the guide →Being decided right now, in rooms most people don't know exist.
Actively contested in live rate cases and dockets. Attribution of specific retail increases to data centers is disputed by serious people.
Read the guide →The most local question, and where most actual decisions get made.
The accountability gap is documented: tax determinations that require nothing in return, and use figures that are contractual trade secrets.
Read the guide →Mining, materials, and e-waste — global in flow, local in harm, mostly displaced abroad.
CPAI has not done this research, so there is no guide. Building a thin page on material we haven't verified would break the rule this whole section is built on.
Tell us if you need this →A global problem hands you guilt. There is nothing at the end of it you can actually do on a Tuesday. A watershed problem hands you something else entirely: a named municipal water system, a drought plan, a withdrawal permit, and a disclosure requirement that either exists or doesn’t. One of those is a feeling. The other is a phone call.
Getting the scale right is not a way of caring less. It is the only way to do anything.
Two of these move in opposite directions.
If you need proof these are separate questions rather than one issue with several faces, it's this — and it is the least-quoted finding in the literature.
“Water use by data centres can be negatively coupled with CO2-equivalent emissions, with methods of reducing water consumption increasing carbon emissions in some cases.”
Chien, Gupta, Ren, Sriraman & Tomlinson (2026), Nature Reviews Clean Technology
Closed-loop “zero water” cooling cuts water use and raises electricity demand — which raises emissions. The water question is answered in a basin; the carbon question is answered in the atmosphere. Anyone demanding both zero water and zero carbon is asking for something the engineering does not currently offer. Two impacts that can move in opposite directions were never one issue.
Five guides, one question at a time.
Each guide names the overstated version and the dismissive version of its claim before saying what the evidence supports. That is the whole method.
What Are We Arguing About?
The seven separate questions hiding inside “AI and the environment” — and the geographic scale at which each is true. Start here.
Read the guide →Electricity and Emissions
US data centers used 4.4% of national electricity in 2023, headed for 6.7–12% by 2028. What that does to emissions depends on something the number doesn’t contain.
Read the guide →The Water Question
The most-quoted statistic in this debate has been retired by its own authors. What replaced it is more useful — and it points at your watershed, not the planet.
Read the guide →Who Pays
The health and cost burden of the AI buildout lands on specific counties and specific ratepayers — and the algorithms optimizing for aggregate efficiency make that worse, not better.
Read the guide →The North Carolina Case
A repealed carbon target, 37 unpublished tax determinations, a large-load queue the state’s own Commissioner calls 70% data centers, and an Attorney General who wouldn’t sign.
Read the guide →What did it take to make the hardware?
Mining, materials, and e-waste is a real question and the one most likely to come up in a room. CPAI hasn’t researched it, so there is no guide here. A thin page assembled from unsourced material would break the rule the rest of this section is built on — and it’s the rule we’d be most embarrassed to break here.
Tell us if you need this one →Both of the usual answers are wrong in the same way.
“AI is bad for the environment.”
The version that goes too far
Treats seven claims of wildly different evidence quality as one settled verdict, and leans on figures their own authors have since retired. The strongest version of this argument is weakened, not helped, by the numbers most often used to make it.
The version that waves it away
Answers a county-level rate increase with a global denominator — “it's only about 3% of world electricity” — which is true, and beside the point for the person holding the bill. A global fraction is not a response to a local cost.
What the evidence supports
Electricity demand is real and well measured. The carbon consequence depends on grid mix and on the marginal plant, which is unresolved. Water and health burden are real, local, and concentrated in specific basins and downwind counties. And the cost and consent questions are being decided right now, in dockets and county meetings most people don't know exist.
Sources for this split: lbnl2024 · healthComputing · smallBottle · ncCommerceMemo — full citations below.
Where this leads
Knowing which question you're asking is a skill.
This series applies it to one contested topic. The Applied AI Certification builds the underlying capability — telling a claim's scale from its rhetoric, and knowing what evidence would change your mind — across every domain where AI shows up, not just this one.
What this series is built on.
Every empirical claim across the five guides traces to one of these. Each carries three things: where it was published, how the number was produced, and the geographic scale at which its claims hold. A federal report and a preprint are not the same kind of thing, and a modeled estimate is not a measurement.
Want CPAI to teach this in your community or classroom?
We deliver this material as workshops and sessions for schools, libraries, local government, and community organizations — including the version built for county-level decisions.