Data methodology
H2Prism organizes what public sources report. Our responsibility is to transfer information accurately, retain its provenance, explain how values are selected and correct known errors. A value's presence on the site is not a guarantee that the source's underlying claim is true.
Last updated 2026-11-16
Evidence and publication checks
For a reported value, our standard requires an identifiable publisher and document, supporting original evidence and a second AI comparison of the subject, scope, value and unit. Values that cannot be compared with the evidence, or remain mismatched, should not be published. A document's date and whether it calls a value an estimate are recorded when available; a missing date does not turn the date we collected it into the source date.
These checks address transcription, attribution and meaning. They do not independently establish that a project will proceed, a forecast will be met or an announced performance will be achieved.
Source types, not accuracy grades
We distinguish material from the parties involved, government and public bodies, news publishers and compiled datasets. We do not assign A, B or C source grades or a score purporting to measure whether each value is true. An official source can be incomplete or outdated, and a later report can document a real change.
We record usage conditions when known and follow confirmed attribution and reuse requirements. When confirmed terms explicitly prohibit the reuse of reported facts or figures in our intended use, our publication standard excludes those values and source references. An item-specific restriction on reproducing text or images is not treated as a blanket prohibition on every reported fact; the material and intended reuse still require review. Unconfirmed usage terms are not a grant of permission to reuse the original work. Full articles and reports are not republished.
Selecting among conflicting values
We first ask whether sources describe the same subject, phase, boundary and quantity. Different phases or measurements belong in different fields. Where values conflict, a documented later development, such as downsizing, delay, cancellation or restart, takes priority. A still later statement by the responsible party confirming that the earlier position remains unchanged takes priority over that report. A newer publication that merely repeats an old announcement is not a later development.
If sources disagree without describing a change, we prefer the responsible party's original material, then more recent material, then greater specificity. If the sequence or value still cannot be resolved, the field remains blank. Our data standard requires us to retain prior values and selection reasons when a current value changes. We do not select a different number simply because it would be easier to publish.
Conversions, calculations and estimates
A deterministic unit conversion, such as GW to MW, remains a reported value and does not by itself become an estimate. We retain the conversion rule. Calculated results require eligible input values, a reproducible formula and stated assumptions from the relevant method. We do not invent a project-specific assumption to fill a missing fact.
The label Est. is used for a source's stated estimate or a calculation that relies on assumptions. Plans and targets are identified through the field name. H2Prism interpretation belongs in explanatory text rather than being presented as a reported value. Browser calculators also let you test your own assumptions; their results describe that scenario.
Reading values and source information
List views emphasize values and units. Detail views group source information with related facts. Source details can identify the document, publisher and relevant dates; not every source has an accessible link. A source date describes the material's date, an effective date describes when a policy applies, and a recheck date means the original material was actually reopened. These dates are not interchangeable.
A dash (—) represents a missing displayed value, never zero. Other empty fields may be omitted. Displayed values may be rounded to a readable precision without adding precision the source did not provide. Check the original value and unit in the source information when making a precise comparison.
A record of more complete supporting documentation does not amount to an accuracy or investment-quality rating.
Project counts and stage definitions
Project counts, stage groups and shares follow the definitions displayed with each view. Counts describe the displayed collection, not all projects in the market. Separately recorded phases can count separately.
Our counting standard defines active projects as excluding cancelled, dormant and decommissioned projects. On-hold projects remain distinct from operational projects and can be included in an active total. Read the view's population and filters before comparing counts.
Under our stage definitions, operational means the normalized operational stage. FID / Construction preserves a combined source category and does not prove a separately dated investment decision. FID+ groups the combined category, separately identified FID or construction, and operational projects. Demonstration and unknown stages do not establish operational or FID+ status. These definitions set the comparison standard; the labels and coverage displayed with a particular view explain its actual calculation.
Capacity, shares and dates in comparisons
Our comparison standard keeps electrical capacity separate from hydrogen output. Compatible power units can be converted to MW and compatible annual hydrogen mass units to an annual mass basis. A volume flow requires a stated method and operating-hours assumption before it can be expressed as annual production. Target startup and FID years remain separate from actual operating years and FID dates.
A capacity total must identify the included population, the measure and how many projects have a recorded value. Missing capacity is not zero. Project and phase boundaries may mix total and incremental capacity, so a sum of recorded capacities is not a market-size estimate.
A regional share must identify its region definition and denominator and explain how unknown or multi-region projects are treated. Our regional comparison standard uses active projects with a single known region and keeps unknown and multi-region counts separate. Projects using several technologies can count under more than one technology, so project shares can sum above 100%. MW shares require an unambiguous allocation, with mixed and unspecified capacity identified separately. Check the actual measure, filters and denominator shown with the chart; these determine what the share means.
What this release collects
This release collects four kinds of public records: hydrogen projects, companies, policies and countries. A record is included when a publisher and a document can be named and the value can be checked against the original evidence.
How a source is attached
Each published value keeps the document name, the publisher and the source date when the source gives one. Detail pages show the publisher names for a group of values. Open that group to read the document name, the publisher and the date. List pages show the value and the unit, not a source on every number.
The publishers cited most often in the current release are listed on Sources.
Second comparison before publication
A transferred value is published only when a second AI comparison agrees on the subject, the scope, the value and the unit. A value that cannot be compared, or that still disagrees, is not published. A unit conversion with one determined result stays a transferred value, and the conversion rule is kept. A calculated value needs eligible inputs, a formula that can be repeated and assumptions taken from the stated method. An assumption invented to fill one project's gap is not used.
Usage terms
When a source states terms that allow reuse, those terms are followed, including any required attribution. When the terms are not confirmed, the value may still be published, and the original work is not treated as cleared for reuse. When the terms clearly prohibit reuse of the reported facts or figures, neither the value nor the source name is published. A general copyright notice is not that prohibition.
Capacity units
Electrical capacity stays separate from hydrogen output. Compatible power figures are shown in MW. Compatible annual hydrogen mass figures stay on an annual mass basis, such as tpa or ktpa. A volume flow is not converted to annual production unless the method and the operating-hours assumption are stated. A capacity total names how many projects have a recorded value. A missing capacity is not zero.
How often records change
A record changes when supported new evidence is incorporated. There is no fixed refresh interval for every record, and the site is not a real-time feed. A site build does not mean every original document was opened again. The update record lists what changed in a published release.
Updates and limitations
We update information when supported new evidence is incorporated. We do not promise a fixed refresh interval for every record, and the site is not a real-time feed. Priority depends on the significance of a change, available evidence and correction reports. A site build or page revision is not a recheck of every original document.
Coverage and source availability vary across countries, technologies and project stages. Missing values and duplicate or separate project phases can affect comparisons. Use the collection to locate evidence and frame questions, then verify the source and boundaries before relying on a result.
Suggest a correction
Email [email protected] with the page, field, proposed correction and supporting original source. Include the unit, date and project phase where relevant. Our editorial policy explains corrections and later developments.