- Home
- Sources
Sources
Each fact on the site links to its own source. This page lists the kinds of source we rely on, the open datasets we credit, and what we leave out.
Primary sources
| Source | Used for |
|---|---|
| SEC EDGAR: Form D, 8-K, 10-Q, 10-K, 6-K, S-1 | Amounts raised, contracts, debt, capacity and backlog of US-listed and US-filing companies |
| UK Companies House, exchange filings (HKEX and others) | Directors, share issues, listed-company reports |
| Company press releases, official blogs and posts, investor relations pages | Rounds, investors, launches, partnerships, capacity announcements |
| Earnings call transcripts | Capacity, backlog and customer statements by management |
| Public job boards of the companies | Open roles as a hiring signal, office locations |
| arXiv, Hugging Face, GitHub | Papers, released models and code |
| Company websites, X, LinkedIn, Google Scholar | Addresses of public profiles of companies and people; profile content is not copied |
Media
Reporting by established outlets is used for transactions that companies do not announce themselves, and is the usual source of Reported status. The outlet is named on the record. When the original is behind a paywall we link to an open account of the same report and name the original outlet.
Open datasets
Some records were first identified through open lists and datasets. Where a dataset's licence requires attribution, it is credited here and on the facts that rely on it.
- Juniper Ventures, AI Neolab Tracker (CC BY 4.0): used as a starting list of companies. Figures on EveryNeo are taken from the primary and media sources cited on each record.
- Epoch AI, data on AI models and data centres (CC BY 4.0): used for cross-checking capacity figures where cited.
What we do not use
- Commercial funding databases whose licences forbid republication.
- Data from professional networks and other sites that require a login.
- Anonymous posts as a sole source for a figure.
How sources are ranked and how status is assigned: methodology.