Limitations

What this tool does and doesn't capture. Read this before relying on it for anything consequential.

Coverage is not exhaustive

  • State coverage is selective. Official-source and legislative-provider searches use defined terms and source boundaries. Ohio also inspects selected cross-cutting infrastructure titles, but this wider discovery is not nationwide. Relevant provisions inside broader tax, water, utility or permitting bills can still be missed. A successful source check does not establish complete coverage.
  • Federal Register discovery is keyword- and agency scoped. The feed covers DOE, EPA, FERC, Commerce, IRS, CEQ, and Army Corps, plus a separate keyword gated presidential document pass. Relevant actions using unexpected language or issued only as agency guidance can still be missed.
  • No local or county data. Much data center siting is fought at the municipal and county level, which no national dataset covers well. None of that is here.
  • FERC docket filings are not ingested directly. FERC eLibrary has no usable public API, so individual docket filings aren't pulled. FERC's major actions still appear via the Federal Register, and a small curated watchlist can point to exceptionally important official dockets.
  • Governor monitoring is source by source. The registry checks official governor newsrooms for all 50 states and D.C., but government sites can block automated requests, move URLs, paginate older announcements, or publish an action only on an agency page. The Executive Actions page reports each source's latest check status; an uncheckable source is not represented as live coverage.
  • No comprehensive ISO/RTO or PUC/PSC proceedings. Those systems are fragmented and primarily concern regional grid operations, retail rates, and project specific service. They are intentionally outside this tracker's federal first policy scope.
  • Federal analysis is contextual. GAO and GovInfo reports help explain policy but do not themselves change the law. The official CRS site does not expose a supported automated API, so CRS links are curated rather than described as a live feed.
  • The Facility Atlas is an evidence viewer, not a facility census. It shows current source specific observations from Compute Atlas, PNNL IM3/OpenStreetMap, and the geocodable U.S. portion of Epoch AI's selective AI data center dataset. These sources have different definitions and are not silently merged. IM3 is OSM derived, Epoch power values are estimates, and Census geocoded addresses are approximate context; no marker should be interpreted as a parcel boundary, entrance, or verified operating location.
  • Overlays are context, not project findings. State policy counts do not prove applicability. EIA values do not prove utility service or capacity. The USGS water layer is a historical surface water model aggregated to state context and excludes groundwater, transfers, ecological flow needs, and facility specific supply.
  • Entity matching remains incomplete. Cross source candidates are deliberately conservative and can miss the same facility or propose nearby but different campuses. Only reviewed relationships are labelled confirmed, and confirmation does not erase source specific observations.
  • Documented positions are actions, not ideology scores. The positions ledger derives a neutral category from a public official's source linked action on one recorded measure. It does not infer the person's private views, overall sentiment, or position on every data center proposal. Absence from the ledger is not evidence of no position.

The relevance score is an AI estimate

Each item is scored 0–100% for how relevant it is to data center policy by a language model, and items below 40% are hidden by default (toggle "Include low relevance" on the Bills page to see them). These scores are useful for ranking and filtering, but they are estimates and can be wrong in both directions, a genuinely important item can be under scored, and an incidental mention can slip through. Topic tags carry the same caveat.

Other things to know

  • "State" on the map means different things. For state bills it is the jurisdiction; for federal bills it is the sponsor's home state. The map blends both.
  • Comment counts are docket level. Regulations.gov comment totals count all comments on a rule's docket, matched to the Federal Register document, a good engagement signal, not a per document count.
  • Refresh cadence varies by source. Core policy feeds run hourly; governor sources run every six hours and federal research runs daily. Newly published items can lag the official record.
  • This is not legal advice. Always verify against the primary source (linked on every item) before acting.

← Back to the overview