Watterload: the power and water data centres quietly share.

Data centres in Aragón with different owners depend on the same substations and the same water system, so one drought or grid bottleneck can hit them all at once, and the people financing them can't see it. Watterload uses the operators' own public permits to map which power and water each project shares, and gives investors a sourced report on the limiting factor and the questions to ask before they commit.

Why it matters

Every AI query runs on servers, power and cooling. We know the costs in electricity and water. Heat is a newer one. A 2026 preprint, not yet peer reviewed, found land around data centres warming by up to 9.1°C at some sites, with the effect reaching into nearby communities. It needs more research, but it points the same way as the grid and the water: the costs land locally, on systems other people share.

The point is not to stop AI. It is to build it on systems that can carry it. That means:

  • Data centres that use less energy
  • New renewable electricity, not a bigger share of what is already there
  • Smarter cooling, and putting waste heat to use
  • Planning that accounts for land and nature
  • Honest, public monitoring of environmental impact

The Fixathon

Norrsken selected 50 people for its Fixathon, a 24-hour hackathon in Barcelona in October 2026. Demand for places was high and selection was very competitive. The Fixathon is part of Norrsken's Prompt What Matters initiative.

Our team built a scalable product and business model to stop data-centre development from collapsing grids and water resources: resilient systems for data sovereignty, and a way to assess the true capacity of these data centres. The team was Cédric Lion, Cecilia H. G. Morilla, who brought the water–energy science from Cetaqua, and me.

What we built

A four-step tool with no login. Say who you are, pick one of 22 known Aragón projects or drop a pin for a new one, and see what it depends on: the substation, the water system and the pipe, each with a load bar and the other projects that share it. The last step is a report you can print, with the limiting factor, what could happen and the next step for your role.

  • Lenders and investorsA report on an existing or planned project, with the questions to ask the developer
  • Developers and operatorsWhat a new site would depend on, and three lower-stress areas to consider instead
  • Insurers and reinsurersThe next step: the same shared exposure, seen across a whole portfolio

Notes

Public data goes further than people think. Everything comes from public filings: grid permits, the grid operator's list of congested nodes, and the reservoir and city water system. The risk was never hidden. It was spread across documents nobody reads together.

Every number carries its label. Declared, reported, inferred, likely, not public, our estimate. The tool shows stress, not probability, and invents no losses or prices. For anyone making a decision, knowing how solid a figure is matters as much as the figure.

The buyer changed halfway through. We started with reinsurers in mind. Five rounds of expert feedback moved the first user to the people financing and building the projects, who decide before the risk is locked in. Insurers come next.

A 24-hour prototype. Figures are labelled by source, and its signals are not a verdict on any operator. Back to talks and events