AI growth puts data centre sustainability under strain

Dame Dawn Childs, a prominent figure in the data centre industry, highlights that AI-driven growth is pushing the limits of data centre sustainability. The industry must now balance rising compute demand with energy efficiency and environmental constraints, requiring significant innovation and collaboration.

• Dame Dawn Childs says AI infrastructure demand is colliding with slower-moving power networks.
• Operators must address absolute electricity demand, cooling, site constraints and grid integration—not efficiency claims alone.

What happened: AI build-out meets physical constraints

At Datacloud Energy Europe 2026, Dame Dawn Childs, then president of Pure Data Centres Group, discussed with Capacity how AI-led expansion is testing the industry's ability to build sustainably. She described sustainability in practical terms: not only lowering environmental impact, but being able to keep adding infrastructure at the pace customers require.

Childs identified a timing mismatch. Digital infrastructure may need to scale over one or two years, while power infrastructure can take five to ten years. She said the immediate bottleneck is often transmission capacity rather than electricity generation, making coordination among data centre operators, grid companies, governments and regulators essential.

She also argued that AI is both part of the demand problem and a possible tool for improving grid and water use. That does not remove the near-term constraints: higher-density computing raises the importance of power delivery, cooling design, equipment supply, site selection and a credible route from interim energy arrangements to lower-carbon operation.

The wider data supports the direction of the warning. The International Energy Agency estimated that data centres consumed about 415 TWh of electricity in 2024, around 1.5% of global electricity use, and identified AI as a major driver of future demand and higher rack power density.

For operators, the issue is therefore not a single sustainability metric. A project can improve energy efficiency per unit of computing while total electricity use, cooling demand or local grid pressure still rises as capacity expands.

Why it matters

The gap between computing demand and available power can delay projects, raise connection and equipment costs, and alter where new capacity is built. In grid-constrained markets, a technically viable data centre may still depend on transmission upgrades, planning approval and agreements with utilities.

Energy efficiency affects both operating expenditure and investment decisions, but claims need clear boundaries. Power usage effectiveness does not by itself describe absolute electricity demand, water use, generation mix or the effect on a local network.

Europe is a particularly important test because infrastructure developers must reconcile AI investment with grid queues, planning rules and environmental commitments. Childs' central point is collaborative: data centres can potentially support flexible grids, but only if operators and public authorities plan the power system and digital build-out together.

The evidence also has limits. The Capacity material records an executive assessment, not a quantified forecast for a specific Pure Data Centres project. The most useful follow-up is to track disclosed power capacity, grid-connection schedules, cooling and water choices, energy sourcing, planning decisions and comparable operating metrics.

AI growth is not proof that sustainability targets are unattainable. It does mean that the sector's claims will increasingly be judged against physical delivery: how much power is available, when it arrives, how efficiently it is used and whether expansion shifts environmental or network costs elsewhere.