User guide

What is free to read, and what is supplied on request

• Free, no registration: every country and city page with its facility counts and development stages, the ten leading countries and ten leading companies by aggregated power, the development-stage split, ten fixed source-linked sample facilities, and the full methods, coverage, release notes and data dictionary pages.
• On request: the facility-level CSV — every facility and every released factual attribute in a sortable file, with at most one primary public source and its reading date per facility — for the whole catalogue or for the countries, cities or operators you need. Write to contact@provenancedatahub.com; the get-the-data page sets out what each row holds and how a request works.

How the dataset updates

The catalogue is rebuilt monthly. Each cycle re-reads public sources for a set of facilities, adds newly evidenced facilities, advances development statuses as projects move from planning through construction into service, applies corrections against new public evidence, and extends attribute coverage. Every published rebuild passes automated release gates and is read by a person, then ships as a new release with its own ID; what changed is counted on the release notes page, release by release.

• Facility IDs are permanent. A facility keeps its ID across releases, so a reference saved from one file stays valid in the next.
• A delivered CSV never changes. It is built from one release and carries that release ID; a later release is a new file, never an edit to one already delivered.
• Months without material change publish no release. A new release ID always means real changes, counted in the release notes.

Reading facility power correctly

• Facility power is the facility or campus power under the source’s definition, in MW — it is not annual electricity consumption.
• Operating, construction, planning, and status-not-specified values are always kept separate; planned capacity is not operating capacity.
• A blank value means the public record does not state one. It never means zero.
• Country and company totals cover facilities that state a power value — read them as evidence-backed floors, not installed-capacity estimates.
• Announced full build-out (“up to X MW”) lives in its own ultimate-capacity field and is never summed into any stage total.
• The platform publishes no calculated electricity or emissions results; that layer is reserved — see methods.

Working with sources

Each facility carries one primary source URL and its reading date, the public page that best identifies it. The link is an orientation reference and may not support every individual attribute; some facilities have no eligible customer-facing link. Source pages can change or move after a release is generated, so the primary reference should be checked before use. The reading behind every value — pages screened, pages read, field-level citations — is counted on the methods page.

Typical workflows

• Market sizing and site benchmarking: start from a country or city page, compare stage-separated power, and use the ultimate-capacity field in the CSV to see announced headroom.
• Power and grid planning: work from the CSV and sum capacity by region and stage. A campus and the buildings inside it are counted once, so totals do not double-count.
• Supply-chain and vendor analysis: query cooling approach, liquid-cooling flags, rack density, and PUE where recorded — the coverage page states how many facilities state each attribute family.
• Due diligence: open the primary source, and check current operator or official disclosures before relying on a single figure.

Frequently asked questions

• Why do some facilities show no power value? Because no public source states one. The facility still carries its identity, location, status and its source; the value stays blank rather than being estimated.
• Why does attribute coverage differ between countries? Disclosure practice differs by market and operator. The coverage page states availability per attribute family and by development stage so you can judge fit before requesting the file.
• Can I use planning-stage facilities as a project pipeline? Yes — that is why stages are never blended. Treat planning values as announced intent, construction as committed, and check the source date for staleness. The release notes show how many projects advanced a stage in each release.
• How do I get the facility-level CSV? Write to contact@provenancedatahub.com with your organisation, the intended use and the countries or cities you need; a quote follows within two business days. What each row holds is set out on the get-the-data page.
• Is there an API or bulk feed? Not at present. Scheduled monthly delivery, organisation-wide use and client-use rights are arranged through a commercial enquiry.
• Can I train an AI model on the data? No — AI/ML training, evaluation and retrieval-grounding are excluded unless a signed licence says otherwise. See the data license.
• Does a delivered CSV change when a new release is published? Never. A delivered file is built from one release and carries its release ID; later releases are new files. The release notes state what changed between them.

Corrections and support

Found a value the public record contradicts? Send the facility ID, field, proposed correction, and public source URL through Contact & corrections. Substantiated corrections ship in a later release, counted in the release notes. Support requests receive a first reply within two business days.