Skip to main content
Sociable Systems
Search by title, term, episode, or route...
Subscribe
Episode 246 · The Ground Address · 2026-09-03

The Same Account

Two draws arrive at the same municipal system, and one of them is contractually guaranteed while the other queued in living memory. Site-boundary assessment cannot see any of it, because the boundary is the instruction.

Cover art for episode 246: The Same Account
Ground Address ArcSocial Impact AssessmentDistribution
Episode 246: The Same Account

Two draws, one pipe. One is contractually certain. One queued, in living memory, and the queue was not a metaphor.

The record has had three episodes. Distribution has had none, which is roughly the ratio the files themselves run at. Not consumption in the abstract: claims on shared systems, and who stands where in the queue they create.

The operative fact is not that a facility uses water and electricity. Everything uses water and electricity. The operative fact is that two draws arrive at the same municipal account, the same pipes and pressure and grid section and emergency load-shedding arithmetic, and one of the two draws is contractually guaranteed while the other one queued. Not hypothetically queued. In cities in this country, in living memory, with buckets, in summers everyone older than thirty can place. When a new industrial-scale draw connects to a system with that history, the distributional question is the oldest question the system knows, asked by a new kind of customer.


There is a reason the guarantee sits where it does, and the industry states it more plainly than any objector would. Electricity is a small line in this business. A gigawatt-scale facility carries capital costs in the tens of billions, most of it silicon, against an annual power bill well under a billion, and investors in the sector say without embarrassment that an operator offered power tomorrow at twice the price would take it. Willingness to pay is close to unbounded at the top of this market, which makes price a poor instrument for deciding who gets the water in a dry summer. The household on the same system has a different relationship to price entirely, and price is the only argument it is given.

What a social baseline for this class of plant holds

Site-boundary assessment cannot see any of this, and the practitioners are not at fault: the boundary is the instruction. Assessed as an ordinary commercial building, the affected area is the site and its noise contour, and the receptors are the neighbours in the acoustic model and nobody else. Every distributional question the field exists to ask is out of scope before it is asked, because the classification placed it there before anyone looked.

A social baseline written for what the plant actually is would look different, and none of it is exotic. Receptors beyond the noise contour: everyone on the shared system, mapped by claim rather than by proximity. Cumulative municipal draw across facilities, plural, because no process anywhere assesses the campuses together, and the aggregate is the number the city actually has to plan around. Employment intensity per unit of allocated resource: fifty permanent positions against a town-scale draw is a ratio, and ratios can be compared, which is what made it assessable elsewhere. And the standby-generation reality behind the uptime promise, disclosed as a public-health variable rather than an engineering appendix, because the hours when the generators run are precisely the hours when the grid fails the street, and both events share a cause.

The cooling question reads the same way. Closed-loop, air-cooled, recycled: each is a real engineering choice, described in good faith in brochures. None of it binds until it is a condition of consent rather than a page in a marketing document. The test is which technology the applicant is required to use, and what happens to the answer if the water situation changes after commissioning.

The energy claim behaves identically and is further along the same road. A facility can be described as renewable-powered on the strength of certificates bought from generation that already exists, which adds no new megawatt to the system while adding a new load to it. Australia's energy market commission named exactly that as a loophole in July and recommended closing it, by tying the certificates to new generation and requiring that generation to be firmed. Without a condition of that kind the claim is true as written and says nothing about what the grid actually burns, and what it burns to serve this class of facility worldwide is still more fossil generation than clean. A real instrument, described accurately, binding nothing.

The sector's own instrumentation says as much. Uptime Institute, whose tier standard defines the uptime promise in the first place, surveys operators worldwide every year, and fewer than half of them collect water-consumption data at all. Its own explanation is that large expansion projects carry high resource consumption, which makes sustainability targets harder to reach and reduces operators' appetite for collecting and reporting the figures. That is the industry describing from the inside the mechanism this week has been describing from the outside, and it arrives without any of the framing an objector would bring to it.

The instruments exist

The Liability Sponge named the general pattern: someone always absorbs the strain, and it is rarely the party with power to alter it. The field's answer to that pattern was built over decades and it works. The mitigation hierarchy, the social baseline, the affected-area definition, the grievance mechanism. These instruments have absorbed smelters, highways, resettlements. They are mature, they are staffed by capable practitioners, and they are not being asked.

That last clause is where the Mozambique work is relevant, as method and nothing more: evidence chains, cumulative registers, the discipline of asking who pays when a claim proves wrong. An industrial erf is not a resettlement village, and the comparison would flatter neither. But the method for making distribution visible was built for exactly this class of question and transfers whole.

The mitigation hierarchy orders its answers avoid, minimise, restore, offset, by decreasing reversibility. It is the right hierarchy for impacts, and it is silent on allocation, because allocation is a decision about who shares a system rather than an impact of a project. That gap comes back in a later arc.

The framework exists. The practitioners exist. What is missing is the expectation that this class of facility earns its consent the way every other heavy-infrastructure class here has had to. Expectations are the cheapest instrument in the entire toolkit. They cost a sentence, and they change what everyone downstream does.


Companions

  • The classification that drew the boundary: Assessed as a Shed.
  • Who absorbs the strain: The Liability Sponge.
  • The submission that asks the distributional questions in public: the civil-society filing to the South African Human Rights Commission, August 2026.
  • The loophole in renewable-supply claims, and a regulator's recommendation to close it: the AEMC's July 2026 advice to energy ministers.
  • What the grid actually supplies this class of facility: the IEA's Energy and AI analysis.
  • What the power actually costs against what the building costs: Epoch AI's cost breakdown for a one-gigawatt facility.
  • The sector's own account of why the figures are thin: Uptime Institute's Global Data Center Survey 2025, which puts worldwide collection of water-consumption data at 47 percent.
  • The instrument, unmodified and ready: the environmental and social assessment framework this country already runs.
  • The asymmetry, in figures: What It Draws sets a restriction scenario against a facility's unchanged draw. Under fifty litres a person a day a household is held to about sixty kilolitres a year, and the uptime guarantee does not move, because one of those allocations is contractual and the other is not.

These notes come out of Sociable Systems, a practice that reads AI-shaped documents the way a hostile reviewer will, before a lender or a court finds the gap. The argument has an operational form: the Interim Protocol sets out four rules for AI use in environmental and social deliverables, covering disclosure at touch-point grain, evidence custody, the phrases no automated screening may settle, and a hostile read before anything ships. Free, and written to be cited or retired once institutional guidance arrives.

Daily field notes

One episode a day, in your inbox.

Subscribe