London is writing a data centre policy. Cities need to treat AI compute like heavy industry, not invisible infrastructure
London is moving towards a dedicated policy on new data centres, driven by concerns about energy and water use and the carbon footprint of rapid buildouts.
That should be read as a warning to every city chasing "AI hub" status. Data centres are being treated like clean, invisible infrastructure. In practice they behave like heavy industry: they need land, they strain grids, they compete for water, and they lock in long-term energy demand.
The risk is that planning and climate policy stay abstract while projects arrive one by one. Each application looks manageable. The cumulative load is not. Without a transparent accounting of total demand, cities end up with a slow-motion capacity crisis and a procurement scramble for emergency power that blows up climate targets.
The accountability gap is obvious. The benefits are counted as jobs and investment. The costs show up as higher bills, grid constraints, and local environmental trade-offs. If a policy is serious, it needs hard requirements: measurable efficiency targets, public reporting, water limits, and binding commitments to bring new generation, not just consume it.
We'd like your views:
- Should planning approval require proof of additional low-carbon generation, not just purchase agreements?
- What should be disclosed publicly: energy demand, water usage, emissions, and local grid upgrades?
- Who should pay when data centre growth forces expensive infrastructure expansion?
- Are moratoriums a sensible tool, or a sign of policy failure?
- What is a fair test for whether a data centre is a net benefit to a city?

