The Third Answer to the AI Water Crisis
The viral fight is a false choice. You do not have to pick between clean water and machine intelligence. There is an architecture that refuses the trade entirely, and we built it.
A clip went round the internet this month and lit a fire. One camp planted a flag and did not blink: we can live without AI data centres, we cannot live without clean water. Turn the machines off. The other camp swung just as hard the other way, repeating a dramatic line that circulated widely, the idea that cooling water for the machines should come before human comfort because it buys us a superintelligence that will solve everything in the end. I want to be careful here, because that second line was passed around as a quote from a famous billionaire and there is no reliable confirmation that anyone actually said it in that form. Treat it as the framing that went viral, not as a verified statement from any named person. The framing is what matters anyway, and the framing is the problem.
Because both camps are wrong. Not a little wrong. Wrong at the root. They are arguing inside a box, and the box is the lie.
Two camps, one hidden assumption
Camp one says: shut it all down. Camp two says: pour the water in. Look at what they agree on before you pick a side. Both assume that intelligence must live in a giant, water-cooled, centralised megacentre. Once you swallow that assumption, only two moves remain. You either protect the water and throw away the most important technology of the age, or you protect the technology and quietly decide that a region's people come second to a cooling plant. Shut-it-all-down sacrifices the future. Drain-the-commons sacrifices the people. Both treat the megacentre as destiny.
It is not destiny. It is a design decision, made by a handful of operators who concluded that the only way to do machine intelligence was to do it the way hyperscale clouds already work: enormous, centralised, thirsty, sited wherever land and power are cheapest, with a region's water table routed through one set of cooling towers. That is one architecture. It is not the only one. And it is not ours.
The numbers are real, and they are sobering
Let me be honest about the scale, because hand-waving the figures is how you lose the argument. The International Energy Agency has reported that electricity demand from data centres is on a steep climb and could roughly double across the second half of this decade, driven in large part by AI. Cooling that load takes water. Independent reporting and operator disclosures put the water footprint of large facilities in the range of hundreds of thousands to millions of litres a day, comparable in some cases to the draw of thousands of households. The exact number depends on climate, cooling design and how you count the water consumed at the power plant upstream, so treat any single headline figure with caution. The direction of travel is not in doubt.
Now hold that against the other side of the ledger. The United Nations and the World Health Organisation estimate that around two billion people still lack safely managed drinking water at home, and the World Economic Forum has named water stress among the gravest risks of the decade. So when someone says we should not let machines drink before people do, they are not being hysterical. They are pointing at something true. The mistake is the next step, the step where they conclude the only way to defend the water is to kill the intelligence.
The third answer
Here is the move both camps missed. You do not have to put intelligence in a megacentre at all. You can right-size it. You can run specialised models on hardware the operator owns, sited near where the work actually happens, fully capable of running offline, drawing the modest power and cooling of equipment you could keep in a cabinet rather than a hall the size of a town. Intelligence distributed across many small, owned nodes does not route a region's water table through one cooling plant, for the simple reason that there is no one cooling plant. There is no central hall to drain a river for.
This is the architecture we built, and it has a name. Mickai is a Sovereign Intelligence Operating System. Fifty specialised brains, each a model in its own right, running distributed on the operator's own hardware, from edge devices and workstations up to our flagship Prometheus edge server, and we are actively training our own models now. It is the opposite of a hyperscale cloud by design. The point was never to build a smaller version of the megacentre. The point was to make the megacentre unnecessary.
Sovereignty is the word that does the work. When the model runs on your machine, the water question changes shape. You are not negotiating with a distant operator over how much of your aquifer their cooling towers may take. You are running a right-sized system on your own power and your own footprint, and the marginal cost to the commons is a rounding error next to a hyperscale hall. Keep the water. Keep the intelligence. The trade the viral fight insisted you had to make simply does not exist at this layer.
Where I have to be straight with you
I am not going to tell you Mickai single-handedly drains the thirst out of every data centre already standing. It does not. The megacentres that are already pouring concrete will keep pouring water until their operators choose differently, and that is on them, not on any one architecture. What I am telling you is narrower and, I think, more useful. Mickai is the architecture that does not create the dilemma in the first place. It is the third path the debate is pretending does not exist. Every workload you can move to sovereign, distributed intelligence is a workload that never asks a town to choose between its taps and its future.
That is the honest claim. Not a magic wand. A different road, available now, that the two-camp fight keeps walking past because it is busy shouting across a false binary.
A record you can check
Sovereignty without accountability is just a black box you happen to own, so we built the proof in. Every consequential action a Mickai system takes can be written to an Open Audit Record, signed with FIPS 204 ML-DSA-65, the published post-quantum signature standard from the United States National Institute of Standards and Technology, which we adopted rather than invented. For durability beyond any single machine, our sovereign Layer 1, Pantheon, can commit a cryptographic hash of that record to Bitcoin. Anchoring a hash is not spending and it is not a Layer 2. It is a tamper-evident timestamp, pinned to the most durable public ledger there is, that says this record existed and has not been altered since.
The architecture is evidenced, not just asserted. The work sits behind 101 filed United Kingdom patent applications, around 2,234 claims, owned by Mickai LTD, with myself as the named inventor. I lead with the architecture rather than the paperwork because the paperwork is supporting evidence, not the argument. The argument is that there was always a third answer.
Pick the road that keeps both
So here is where I land, and I will say it plainly. The shut-it-all-down camp is right that the water matters and wrong that we must surrender the most important technology of the age to protect it. The pour-the-water-in camp is right that the technology matters and monstrous in its conclusion that people should come second to a cooling plant for a payoff nobody can guarantee. Both are trapped in the same box. The way out of the box is not to win the fight. It is to refuse the question.
“You were told to choose between the water and the intelligence. That was never the real choice. Build the intelligence small, sovereign and near to where it is used, and you keep both. I am Micky Irons, and that is the road we built.”
About Mickai
Mickai is a Sovereign Intelligence Operating System, fifty specialised AI brains running distributed and offline-capable on hardware the operator owns, from edge devices and workstations to the flagship Prometheus edge server. Its actions are recorded in an Open Audit Record signed with the FIPS 204 ML-DSA-65 post-quantum standard, with durable proof anchored as a hash to Bitcoin through Pantheon, Mickai's sovereign Layer 1. The architecture is held privately by its founder, Micky Irons, and is evidenced by 101 filed United Kingdom patent applications, around 2,234 claims, owned by Mickai LTD.