The environmental debate around artificial intelligence is often reduced to a single question: how much electricity do data centres consume?
It is an important question, but it is not enough.
Data centres consume power, use water, require grid capacity, generate heat and influence where new energy infrastructure is built. As AI workloads expand, their sustainability performance will increasingly depend not only on computing efficiency, but also on how each facility interacts with the electricity system and the communities around it.
That is the direction of travel behind the European Commission’s new proposed common rating scheme for data centres. The framework is intended to improve transparency around energy and water use, while also considering how facilities support the wider energy system through measures such as waste-heat reuse, clean-energy additions and flexibility.
The proposal applies to individual data centres above 500 kW. The Commission expects the first sustainability labels to appear in 2027, following scrutiny by the European Parliament and Council.
This matters because Europe is looking to triple data-centre capacity over the next five to seven years. At the same time, the Commission cites projections that data centres could account for more than 3% of EU electricity demand by 2030.
The label is therefore not just a reporting exercise. It is a signal that digital infrastructure is becoming energy infrastructure.
A credible data-centre sustainability strategy should answer at least five operational questions:
How efficiently does the facility use electricity?
How much water does it use, and where does that water come from?
Can demand be adjusted when the grid is constrained?
Is waste heat being reused in nearby buildings or heat networks?
Does the facility enable additional clean generation and grid capacity, rather than simply competing for scarce power?
The last question is especially important. A data centre powered by renewable-energy certificates may be able to make an accounting claim. But the system-level outcome depends on location, grid congestion, hourly demand patterns, new generation, storage and local network capacity.
The European Commission notes that reusing around half of the waste heat generated by European data centres could be equivalent to the heating demand of four million households. That figure illustrates the potential, but delivery will depend on geography, heat-network investment, commercial agreements and the proximity of homes or industrial users that can actually use the heat.
For corporate buyers of cloud and AI services, this creates a new procurement challenge. Sustainability teams will need information that is specific enough to guide decisions: facility location, energy mix, power-usage effectiveness, water-use effectiveness, grid conditions, renewable-energy arrangements and carbon-allocation methodology.
For operators and investors, the opportunity is equally clear. The strongest data-centre projects will not be those that merely report a lower footprint. They will be the projects that demonstrate a constructive role in the energy system: flexible demand, efficient design, credible water stewardship, renewable integration and usable heat.
AI may be digital, but its sustainability consequences are physical. Europe’s new rating scheme recognises that the next generation of data centres must be judged not only by computing capacity, but by their contribution to resilient, low-carbon infrastructure.