Frequently Asked Questions
Is This Written by AI?
Yes, in part, and it is disclosed rather than hidden.
AI is used openly and heavily on this project as a drafting and analysis tool. It reads documents in volume, reconciles figures across filings, transcribes audio, and writes first versions of explainers, site pages, and data tooling.
What it does not do is decide what is true, what matters, or what publishes. Every investigative question originated with Brandon, every editorial judgment is his, and nothing publishes that he cannot explain himself, unaided, from the primary sources.
The reason that rule exists is this: an AI states a wrong thing in the same confident register as a right one. Or, in layman terms for people like me, it lies. And it lies confidently. There is no tell. So the check cannot be “does it sound right,” it has to be “does the document say it,” and every document is published here in full at Sources so you can run that check yourself.
The whole policy, not a summary of it
The division of labor, what AI drafts freely, where it never holds the pen, and how verification actually works: How This Investigation Uses AI. The AI’s own account of what it is useful for and where it fails is on the About page.
What Is a Data Center?
A data center is a warehouse for computers.
That sentence is not an oversimplification, it is the most honest description of what these buildings actually are, and it is deliberately obscured by the industry that profits from them.
Specifically: a data center is a dedicated building housing servers (computers that store and process data), storage arrays, and networking equipment, plus the power systems and cooling infrastructure to keep all of it running 24 hours a day, 7 days a week, 365 days per year. Roughly half of a data center’s electricity goes directly to the IT equipment. Cooling typically consumes another 30 to 40% of total power because computers generate heat, and heat destroys computers.
Why We Need Them
Nearly every digital service a person uses runs on a data center somewhere: cloud storage, streaming, banking, email, government records, medical systems. This part is real. The question this project asks is not whether data centers should exist. We all know they must. The question here is who builds them, where are they being built, at what speeds, at whose expense, and with whose consent.
What changed in the last 3 to 5 years is generative AI. Training and operating models like ChatGPT, Grok, and others requires GPU clusters (a specialized type of processor that makes computer go brrr..) that draw 5 to 10 times the power per rack compared to traditional servers. This is the actual driver of the current buildout boom. Not your Instagram habit.
| Figure | What It Measures |
|---|---|
| ~50% | of data center electricity goes to the IT equipment itself |
| 30 to 40% | of electricity consumed by cooling systems |
The Consumer Blame Frame: What It Is and Why It Matters
There is a documented rhetorical shift toward framing the data center boom as a product of individual consumer demand, which is analogous to how the plastics industry promoted “personal carbon footprints” to deflect from corporate responsibility.
The International Energy Agency itself frames growth as driven by “rising interest in AI from individual consumers and businesses.”
The counter-evidence: the buildout is corporate capital allocation. Four companies, Amazon, Google, Meta, and Microsoft, have increased data center spending 76% in 2025. Amazon alone spent $131 billion.
Scale Reference Card
Use this whenever a megawatt figure appears in a story. These are the human anchors that make abstract numbers real.
| Unit | Human Equivalent |
|---|---|
1 MW | ~750 to 800 average U.S. homes |
30 MW (Fisk proposal) | ~22,000 to 24,000 homes; NES calls this “large” |
100 MW | ~75,000 to 80,000 homes; ~2 million gallons water/day |
300 MW (Meta Gallatin actual draw) | ~225,000 homes, every household in Nashville |
500 MW (Meta “headline”) | ~375,000 to 400,000 homes, a mid-sized American city |
3,700 MW (TVA new construction) | ~2.8 to 3.0 million homes, essentially the entire state of Tennessee |
1 GW (1,000 MW) | ~750,000 to 800,000 homes |
176 TWh (U.S. data centers 2023) | Power for ~16 million homes for a full year |
Important Caveats
Who is suing over xAI in Memphis. The lawsuit was brought by the NAACP with the Southern Environmental Law Center and Earthjustice, not the ACLU. Some coverage has attributed it incorrectly.
Megawatt figures are contested. Meta Gallatin is “500 MW” (solar capacity) and “~300 MW” (actual draw). Headlines conflate them. Always ask which number.
Hendersonville contract terms. The contract exists via primary source. The specific dollar amount and three-year term are reported-but-unverified pending a public-records request.
Gallatin Steam Plant connection. No reporting establishes a causal link between TVA Gallatin infrastructure changes and Meta’s facility. The timeline overlaps. The connection is circumstantial.
Water comparisons. National data center water use is modest vs. agriculture. But local concentration in water-stressed areas is the real issue. Aggregate vs. local is not the same argument.
Forward projections. Growth figures (doubling by 2028, etc.) come from market research firms with an interest in projecting growth. Treat as estimates, not established facts.
$4.1M TN tax figure. The Tennessee annual foregone revenue figure from data center exemptions is almost certainly too low given post-2020 buildout scale. Flagged for independent verification.
Self-generation gap. Facilities generating their own power (like xAI’s gas turbines) may not appear in utility consumption data, meaning the 4% national figure may understate true energy use.