Live data sources

Underfollowed public data ingested on a schedule
—sources —fresh

This map ingests publicly available data that mainstream equity research analysts don't usually read. Each source is fetched on its own TTL, cached server-side, and used to enrich the per-facility signals panel and the map's live snapshot. The goal: high-signal-to-noise leading indicators on AI capex.

Tier 1 = direct commitments (IRPs, queues, PPAs). Tier 2 = permits & land. Tier 5 = OSINT & remote sensing. Refresh cadence and last-fetched timestamp shown per source.

Sources index

loading…

Snapshots

Latest payload from each live source (truncated).

loading…

Not ingested yet — manual data pulls

Tier 2 County recorder deed transfers — Loudoun, Williamson, Maricopa, Douglas (NE), Mecklenburg. Track shell-LLC grantees (Hibiscus, Vadata, Lapis, Cottonwood, Sweet Sky Bend). Each county exposes a different web form — needs per-county scraper.
Tier 2 FAA OE/AAA Form 7460 filings — public REST API exists (documented). Direct query by sponsor/location. Would catch DC cooling-tower / genset stack notifications 12+ months pre-commissioning.
Tier 2 EPA AIRS air permits — large emergency-generator banks file ~12 months before commissioning. Reveals MW class.
Tier 2 State PILOT / MEGA filings — Iowa EDA, Virginia VEDP, Georgia, Texas Enterprise Fund. Filed before public site announcement.
Tier 2 UCC-1 filings at state Secretary of State — Schneider/Vertiv/Caterpillar liens secured by equipment at named DC addresses.
Tier 5 OpenStreetMap building diff — Overpass API query over DC-county polygons, week-over-week. New buildings >50k sqft = likely DC.
Tier 5 Microsoft Building Footprints annual diff — ML-derived from satellite. Detect new construction year-over-year in DC counties.
Tier 5 Sentinel-2 NDBI change detection — scripted scene download + spectral index diff at known under-construction sites (Hyperion, Stargate, Project Rainier, Colossus 2).
Tier 5 LinkedIn job postings — geographic search by hyperscaler company ID. Volume spikes by metro = pre-commissioning. Currently no public API; would need scraping.

Why these and not others

The biggest sources of value here are the ones almost nobody reads: public-power IRPs (TVA, BPA, SRP, OPPD) covering huge AI-build geographies but ignored vs. IOU IRPs; sub-zone PJM transmission service requests; FAA Form 7460 filings; and county recorder deed transfers to shell LLCs. These have lead times of 6–24 months ahead of any equity-research-quality announcement.

We deliberately skip what's already heavily covered: hyperscaler 10-Qs (every analyst reads them — capex guidance is in the aggregate-tracker block but isn't where you'll find edge), Nvidia revenue concentration data, etc. Tier 4 stuff is included for completeness on the capex tracker but isn't the alpha.