How to read these pages

Every data page carries a line above the fold saying when the data was captured and naming the file it came from, with a link. A live page shows the operator’s own timestamp; a directory page shows the survey period. If a capture is older than the dataset’s own staleness threshold, the page says stale.

An em dash in a table means the source published nothing for that field. Zero means the source published zero. The two are different facts and the site never collapses them — a plant with no reported nameplate capacity shows a dash, not 0 MW, and it is left out of capacity sums rather than counted as nothing.

Power plants: /power-plants/{fuel}/{state}, with us in the state position for the national page, and /power-plants/{fuel}/{state}/{eia-id}/{name} for one plant. The EIA plant ID is the key; the name in the URL is decoration and a stale one redirects.

Planned projects live under /power-plants/planned/{fuel}/{state}, separately, so an operating count is never inflated by something that has not been built. Data centres are /data-centers/{state} and /data-centers/{state}/{ref}/{name}.

A plant is one EIA Plant ID at one site. A generator is one unit inside it — a reactor, a turbine, a battery rack. Capacity is nameplate, summed across every unit EIA lists at the site, which is why a single plant can carry more than one technology. A plant is filed under the fuel family holding the most capacity, and its page names the split when there is one.

Take the source link from the as-of line, open the file, and look for the number. That is the whole verification path, and it is meant to be short. If the number is not there, the page is wrong.