Construction begins on a fusion plant built to sell electricity to a customer

The step before net electricity: someone breaks ground on a plant whose purpose is to sell power, with a customer signed. It has happened. A company backed by a purchase agreement with a large technology firm began site work on its first plant in Washington state in July 2025, aiming to deliver electricity in 2028, with further permitting still to come at the time; the county's conditional-use decision followed in October 2025. The node stays in, resolved, as the first rung on the fusion ladder. It was added in September 2026, after it had already happened, so no forecast was ever on the record for it and the calibration score leaves it out.

The test: what counts as it having happened

A company begins site construction of a fusion power plant that has a signed agreement to sell electricity to a named customer, confirmed by the company's announcement and local permitting records.

Settled by: The company's announcement and the county's permitting record

The odds, by how long I live

No number yet for this one.

What the ledger says about this step

• Forty-one steps had their criteria tightened so that two readers would agree when each has happened, with no number moved for that alone 2026-09-20 The tree itself changed
• A company broke ground on a fusion power plant that already has a signed customer for its electricity 2026-09-19 Settled

Revisions

Date	Field	From	To	Why
2026-09-19	resolved_links		https://www.helionenergy.com/newsroom/helion-secures-land-and-begins-building-site-of-worlds-first-fusion-power-plant; https://www.world-nuclear-news.org/articles/helion-begins-work-on-fusion-power-plant	Every event marked as having happened now links to the record that settled it, so a reader can check the claim (site rule added September 19th, 2026).
2026-09-19	resolved_links	the announcement and trade press only	the Chelan County hearing-examiner index for 2025 added, for the conditional-use decision of 2025-10-14	Step 27 outside-model review, item 63 (docs/reviews/2026-09-19-rulings.md): the criterion's confirmation is two-part (announcement and permitting record) and only the announcement was linked; resolved_on stays July 30th, 2025 on the announcement, the permit date is recorded and marked verify until the county link is opened. Decided September 19th, 2026.
2026-09-19	description	'began building its first plant'	'began site work', with the later permit noted	Step 27 outside-model review, item 63 (docs/reviews/2026-09-19-rulings.md): the announcement establishes site work with further permitting still to come. Decided September 19th, 2026. The description's sentence about the calibration score is item 102, deferred to the methodology pass, and is left as it was.
2026-09-25	description	'The node stays in, resolved, as the first rung on the fusion ladder and a resolved leaf the calibration score can count.'	'The node stays in, resolved, as the first rung on the fusion ladder. It was added in September 2026, after it had already happened, so no forecast was ever on the record for it and the calibration score leaves it out.'	Step 27 outside-model review, item 102 (docs/reviews/2026-09-19-rulings.md), landed at the methodology pass (docs/methodology-pass-2026-09-25.md) as directed in its session brief (docs/handoff-2026-09-25.md, Session B): the old sentence contradicted the scoring rule in docs/methodology.md, which leaves out any node that resolved before a probability was recorded for it. Text only; no number, status or criterion changed.

Notes

Added at step 7 (September 8th, 2026); the date and facts checked the same day against the company's newsroom and trade press.

---

Think this estimate is wrong? Here's how to tell me, and what happens next. The record itself is data/E-energy.toml in the public repository.