How we build the intelligence
We aggregate public, primary-source data on the AI data-center buildout, label every figure with its source and confidence, and refresh it on a schedule. Nothing here is invented — where data isn't yet available, we say so.
Free
- Full U.S. data-center map — all mapped facilities
- Live power dashboards — grid demand, ISO mix, generation
- Policy tracker — current bills & recent actions
- Intelligence feed, daily brief, deals & filings
- Company directory with current equity snapshots
- Search across current intelligence
Paid Pro
- Premium analyst research reports
- Deep facility & bill profiles (capacity, developer, sponsor, risk)
- Power demand & mix history with CSV export
- Screening & advanced filters across projects, companies, deals (coming soon)
- Alerts — bills advancing, status changes, new deals/filings (coming soon)
- Data API & MCP server (docs)
Source registry
| Source | Data | License | Refresh | Powers |
|---|---|---|---|---|
| OpenStreetMap | Data-center locations (telecom=data_center) | ODbL | Weekly | Map · Projects |
| PNNL / IM3 Data Center Atlas | Facility operator, county, footprint (sqft); locations from OSM | ODbL | Quarterly | Map · API |
| Epoch AI | Frontier data centers, AI supercomputers, AI companies, model compute | CC BY 4.0 | Daily | Map · Power · Markets |
| U.S. EIA | Grid demand (national + ISO), generation mix | Public domain | Hourly | Power |
| CAISO OASIS | California ISO intraday demand | Public | Hourly | Power |
| LegiScan | State & federal legislation | CC BY 4.0 | Hourly | Policy |
| Massive Market Data | News, equities, deals | Licensed API | Hourly | Feed · Deals · Companies |
| SEC EDGAR | Company filings | Public domain | Hourly | Deals · Companies |
Dashboards and the API read these source feeds directly, so they reflect the cadence above. The semantic search & assistant index (which powers chat retrieval and /v1/search?mode=semantic) is rebuilt on a periodic basis and can briefly lag the live feeds; for the freshest figures, the dashboards and structured /v1 endpoints are authoritative.
Policy posture classification
The policy map’s “posture” view is a derived signal, not a legal determination. LegiScan’s “data center” search returns loose matches, so we apply two transparent, deterministic steps:
- Relevance gate. We keep only bills whose title plainly concerns data centers (e.g. “data center,” “hyperscale,” “colocation,” “server farm”), dropping unrelated bills that merely surfaced on the search. We deliberately under-include rather than guess.
- Posture lean. Each relevant bill is tagged by keyword into categories — moratoriums/limits, zoning & siting, ratepayer protection, water use, environmental, and new taxes/fees lean restrictive; tax breaks and economic-development incentives lean incentivizing; studies & disclosure are neutral. A state’s posture is the net balance of its relevant bills.
Classification runs on the bill title and latest action only, so it is directional, not a substitute for reading the bill. Facts and bill text link through to the primary source on LegiScan.
Confidence model
High
Directly from a primary or official source (government data, regulatory filings, mapped coordinates).
Medium
From a credible secondary source (established trade press, company releases) and consistent with other signals.
Developing
Early or single-source signal — surfaced for awareness, not yet corroborated.
Attribution: facility map data © OpenStreetMap contributors (ODbL) and the PNNL/IM3 Open-Source Data Center Atlas (ODbL); frontier data centers, AI supercomputers, AI companies & model-compute data from Epoch AI (CC BY 4.0); policy data via LegiScan (CC BY 4.0); grid & energy data from the U.S. EIA and CAISO; market data via Massive Market Data and SEC EDGAR.
