The map can scale fast if we split the work into evidence tiers: bulk-load baseline facility records, auto-promote low-risk existence claims, and keep NDA, public money, water, and power claims behind a manual evidence gate.
1,382raw IM3 centroids
9,408raw FracTracker/ATLAS/PeeringDB rows
2,934likely duplicates held
4,360baseline records promoted
Auto-promote baseline facility
Auto action
Row has operator or facility name, city/state, coordinates, source URL, and no duplicate match above 0.86.
Hold if
Only metro/region is known, operator is unknown, coordinates are missing, or the name collides with an existing campus.
Output
Public map pin with No NDA reported, Watch concern, and all funding/water fields marked needs records.
Auto-upgrade confidence
Auto action
Two independent sources agree on operator, locality, and status, or one operator/local government source confirms the campus.
Hold if
Capacity/status conflict across sources, or source is only a commercial directory snippet.
Output
Candidate confidence high/medium plus a source badge on the import queue.
Never auto-claim secrecy or public money
Auto action
Never. These fields can be linked automatically only after a human-selected public source is attached.
Hold if
A row mentions NDA, shell company, abatement, utility cost, water volume, or tax exemption without a direct source.
Output
Manual review task with evidence URL, quote/context note, and affected jurisdiction.
Batch dedupe
Auto action
Normalized operator + facility/campus + county/state is unique, or distance to nearest same-operator row is greater than 15 miles.
Hold if
Same operator and county already exist, or a project phase name looks like B1/MKE03/PH03/I-VII.
Output
Merge suggestion instead of a second public pin.
Bulk source stack
These sources are ordered by what they are allowed to prove. Baseline datasets can establish where facilities are; local records establish who got public help and who signed secrecy agreements.
coordinate-backed global facilities with U.S., Canada, and Mexico coverage
Automation
Import North America coordinate-backed rows as baseline location leads after dedupe against curated, IM3, and FracTracker records.
Manual gate
Attribution is required, and the license warns coordinate precision varies; do not treat ATLAS rows as exact facility proof without local/operator confirmation.
North America interconnection and colocation facilities with coordinates
Automation
Import U.S., Canada, and Mexico facilities with valid coordinates after dedupe against curated, IM3, FracTracker, and ATLAS records.
Manual gate
PeeringDB facilities are interconnection locations, often data centers; treat them as colocation facility leads until operator/local records confirm site type and impacts.
Import the downloadable IM3/PNNL or FracTracker records as candidate rows.
Hundreds to low-thousands of candidate records in one pass.
02
Operator confirmation sweep
Run operator pages against candidate rows for Google, Meta, Microsoft, Equinix, Digital Realty, QTS, Vantage, CyrusOne, Stream, Stack, and AWS.
Rapid confidence upgrades and fewer unknown operators.
03
Local money trail pass
For promoted rows, auto-generate state/county search tasks for abatements, utility filings, water permits, and NDA stories.
Turns the map from inventory into accountability reporting without slowing baseline coverage.
Faster does not mean looser: facility existence can be automated, but accusations about NDAs, subsidies, ratepayer risk, and water burdens stay source-linked and review-gated.