Pieza independiente
The ones that went critical in 2026
Movement 2 · The Metabolism — September 2026
I The only binary milestone on the board
In two years of announcements there is one class of event that admits no shading. A reactor is critical or it is not.
There is no eighty per cent of criticality. No criticality in beta, no phased criticality subject to availability, no criticality expected in the second half. The threshold is physical, the reaction either holds itself up or it dies, and crossing it leaves a date that somebody has to sign.
In a sector that measures its progress in adjectives, that is a luxury.
“Anything that can be announced by halves was never measuring anything.”
II What crossed the threshold
| Date | What | What it proves |
|---|---|---|
| Jun 2025 | the federal pilot reactor programme starts | intent with a budget |
| Oct 2025 | conditional permit for a fusion plant | a permit, not power |
| Feb 2026 | measurable deuterium-tritium fusion, per the company itself | a figure from the interested party |
| Jun 2026 | two reactors in the programme go critical | threshold crossed |
| 3 Sep 2026 | a microreactor goes critical; announced on the 16th | threshold crossed |
| Sep 2026 | 396 MW geothermal contract for data centres | an obligation to deliver |
The last two rows are a different kind of thing from the first four. One records physics; the other records a penalty for non-delivery.
Two independent ledgers of these same twenty-seven months disagree on exactly one branch: fusion. One logs four events [construction, permit, demonstration, funding] and the other logs none. Where the density of announcement is highest, the physical work is thinnest.
III Thirteen days
Between the 3rd and the 16th of September there are thirteen days between the thing and the telling of it. That gap deserves respect. It means the thing was done first and reported afterwards, which is the reverse of almost everything else that happened this year.
A signed record has one virtue no press release has: it is signed by someone who answers for it. A press release is signed by a department, which is to say by nobody, spread across several people.
IV Risks of taking criticality as the test
I got this wrong for a month, so I am leaving it written where the error was. I took criticality as proof of real acceleration, and criticality is a laboratory threshold.
The first chain reaction in history, in December 1942, produced half a watt. It lit nothing. The first time a reactor lit anything it was four light bulbs in Idaho [December 1951, an experimental reactor in the middle of the desert]. Nine years between the two, in the most accelerated and best funded period the nuclear industry has ever had.
Criticality is not electricity. A critical reactor in a test facility proves the design works and proves nothing about anyone paying for its kilowatt-hours. Between the two there is an operating licence, a grid connection, a tariff and a customer [four procedures, none of them technical].
V The test, hardened
Three steps, and each one disqualifies the one before it as sufficient proof:
- Announced capacity. Good for knowing the intent and good for nothing else.
- Criticality with a date and a signed record. Proof that the physics answers.
- A supply contract with a penalty clause. Proof that there are electrons and that somebody has bought them.
The third one carries weight. A contract for three hundred and ninety-six megawatts obliges you to deliver; an announcement of ten gigawatts obliges you to repeat it next quarter.
VI Coda
The reactors that crossed the threshold this year did the same work the Chicago pile did, and it is not the work it looks like. They are permission instruments. They exist so that a committee can authorise the next thing.
The plant that will matter in ten years is none of these. It is the one that turns up on somebody's bill.