How this was made. The version number counts drafts of the text. It does not measure the inquiry behind it, which has run over days and across several AI systems, with argument between those systems and within them, directed, refused and repeatedly redirected by the author. The source material was AI-generated, and then adversarially and iteratively refined across a range of tools — systems built by different companies in different jurisdictions, set against each other and against the author. No one of them produced this text, and no one of them reviewed it alone. The plurality is deliberate rather than incidental. A single model carries a single set of priors about which sources are authoritative, and this series argues that an evidence base narrowed in exactly that way is how a contested question comes to look settled. Using one model to investigate that claim would have been the claim refuting itself. To name a single model on it would credit that model with work that was neither its own nor done in a single pass. The plurality was also necessary, and the record should say why. In drafting, the assisting model repeatedly led with United States institutional sources — a national laboratory, an industry association, a market study nineteen years old — and presented conclusions drawn from them as the state of knowledge. On one occasion European measured data contradicting those conclusions was present in the same research return and was placed below them. Framings were proposed that would have argued against this series’ own position using that evidence base, and offered as rigour. Each was refused by the author and the material rebuilt. That is the mechanism these documents describe, occurring in their own making, and it is recorded because a series arguing that evidence bases narrow without anyone deciding to narrow them cannot credibly claim its own production was exempt. The framing, the corrections and the judgements are the author’s, and so are the errors. How this site is written sets out what is declared on every piece, who checks it, and where the per-piece record lives.
Because the numbers are not known, and a placeholder does not stay a placeholder. An illustrative figure published for the shape of an arithmetic gets quoted back as an estimate, usually without the label that qualified it, and usually by someone who was not there when it was written.
That is the reason there is no such table here.
There is published evidence, and the difficulty is what it can settle. The European Commission’s first mandatory reporting round, published in July 2025, covers 770 facilities — mandatory reporting, which is a considerable advance on self-selected survey data.
Set against it, the Lawrence Berkeley National Laboratory report of 2024 describes a simulated US estate, and its own authors note that it assumes systems “operate as designed. This is rarely the case.” Both are good work. Neither answers what it costs a particular institution to hold authority over the software acting for it, which is the question here.
Three. Each is set out on this piece’s sources page, with the reason the figure did not support what it had been cited for, rather than quietly dropped.
A claim that most servers sit in small rooms traced back to a market study from before the cloud era. A figure about water use travelled through two intermediaries before reaching this series, and did not say what the intermediaries said it said. A manufacturing-carbon share was correct for handsets and laptops and was attributed to servers, where the proportion is substantially different.
None of them was fabricated. Each was a real measurement of something other than the thing it was used to support, which is the more common failure and the harder one to notice.
That is one of four open questions rather than a prediction, and it is asked because the answer changes what an institution should do now. If the present build-out is financed on returns that do not arrive, the pricing available today is not the pricing available later, and an institution that has built on the assumption of cheap inference has a different problem from one that has not.
The piece does not say the cycle will break. It says an arrangement worth having should survive either outcome, and that is a checkable property of an arrangement rather than a forecast.
Because the four things it asks about are known to people running systems and are not written down anywhere. What an organisation actually spends to hold its own audit records. How long revocation really takes, measured. What a supplier said when asked the five questions. What had to be given up.
None of that is in the literature. It is in the experience of people who have done it, and the piece provides somewhere to put it.
A settled answer arriving from somewhere else — a body with standing publishing measured figures on what this costs institutions of different sizes, in terms an organisation could apply to itself.
If that happens, these questions get replaced by that work, and the right response is to say so rather than to keep asking questions that have been answered.
What this publication does not claim, and what is outstanding against it in the register.
Asserted, not yet measured: this publication.
A question this rests on is parked: What secession premium will members bear?
We publish no pricing and no affordability claim, and we do not assert that members value the secession guarantee.
Alongside: the publication · sources and provenance