Eric Keller
Founder & investigative data journalist, DataCentersExposed
I build independent transparency databases that turn scattered public records into something the people most affected can actually use. DataCentersExposed is the one focused on the AI data-center buildout: who really owns each facility behind its shell-LLC codename, what tax breaks the public handed over, the environmental record, and the cost that lands on ordinary utility bills.
The work is reporting, not aggregation. Every figure on this site traces to a primary source — an SEC filing, a state incorporation record, an EPA enforcement docket, a county abatement agreement, a water-withdrawal permit, a capacity-auction result — and carries a visible confidence grade. I do the joins that trackers stop short of: shell companies to real corporate parents, subsidy disclosures to specific buildings, violations to the substation that powers them.
How I work
- Primary sources only.Claims that can't be traced to a public record don't ship. Known gaps are published on the coverage scorecard rather than papered over.
- Corrections in public. Every change to a published figure is logged with the date, the before/after, and the source in the corrections log. No silent edits.
- Independence by design. Reader-funded, no advertising, no industry money, no positions in any covered company — the full ethics policy is public.
- AI is a tool, not a source. AI assists with entity extraction and document processing against sourced records; a human is responsible for everything published.
Other work & contact
DataCentersExposed is part of the Exposed Network — a family of independent watchdog databases I build and run solo, applying the same primary-source method to different concentrations of money and power: EpsteinExposed (the Epstein case files), CapitolExposed (congressional stock trades), and ScammingExposed (a worldwide scam database, in pre-launch).
WIRED profiled that work in March 2026 — “He Built the Definitive Epstein Database” — after the launch thread drew more than 10 million views on Reddit. Same method, same standard of proof, applied here to the data-center buildout.
Reach me with a tip, a correction, or a press question through the contact page. If this work helped you, your town, or your reporting, you can support it directly — it's how the site stays free and independent.