bsustainable today
Back to Articles
Industry News5 min read

AI’s Sustainability Test: Can Data-Centre Growth Be Designed Within Planetary Limits?

By bsustainable today
AI’s Sustainability Test: Can Data-Centre Growth Be Designed Within Planetary Limits?

The resource reality

Data-centre electricity demand is expected to grow rapidly. S&P Global projects that global data-centre power consumption could exceed 2,200 TWh by 2030 — comparable with India’s current electricity consumption — and that sector demand could nearly double between 2024 and 2030.

This growth matters because data centres do not operate independently from the power system around them. They require reliable, high-capacity electricity connections, often at a scale that can materially affect local-grid planning, power prices and the pace at which new generation must be developed.

Water is equally important. Data centres may use water directly for cooling, while their electricity consumption can create further indirect water demand, depending on the generation mix. S&P Global reports that 43% of existing data centres are located in areas of high water stress.

This does not make data centres inherently incompatible with climate action. It does mean that development needs to be governed by local resource conditions, rather than generic sustainability claims.

A facility located in a lower-water-stress region, powered by additional clean electricity and designed around efficient cooling has a fundamentally different sustainability profile from one that increases pressure on a constrained grid or watershed.

The additionality question

Technology companies increasingly use renewable-energy certificates and corporate power-purchase agreements to support their clean-power goals. These mechanisms can be valuable tools for financing renewable generation. However, the rapid expansion of AI infrastructure raises a more demanding question: is a company’s electricity strategy adding new clean capacity, or competing for a limited existing supply?

The distinction is significant.

In regions where electricity demand is rising faster than new renewable generation, storage and transmission infrastructure, additional data-centre load can place pressure on the broader system. S&P Global identifies grid strain as a key sustainability challenge and notes that demand growth can contribute to the continued use of legacy gas and coal generation in certain markets.

For organisations with net-zero targets, annual clean-energy matching may therefore become insufficient on its own. The direction of travel is towards more granular approaches that consider:

  • Whether renewable procurement is geographically relevant to consumption

  • Whether it supports new generation and storage capacity

  • Whether power is available when the data centre is actually operating

  • Whether the company contributes to transmission, flexibility and grid resilience

  • Whether backup-power arrangements are compatible with wider decarbonisation goals

The core question is not only whether a company can claim renewable electricity. It is whether its growth helps build a cleaner and more reliable energy system.

Responsible data-centre growth

A credible approach to sustainable data-centre development should be built around five principles.

Responsible location. Site-selection decisions should assess grid carbon intensity, renewable-energy availability, water stress, climate hazards, biodiversity sensitivity and local infrastructure capacity. Location is a sustainability decision, not simply a real-estate decision.

Additional clean energy. Developers and major users should support new renewable generation, storage, transmission and demand-response capacity. This helps ensure that growing demand is matched with a growing supply of low-carbon electricity.

Water stewardship. Operators should measure and disclose water use, assess watershed-level impacts and select cooling systems appropriate to local conditions. In high-water-stress regions, water stewardship should be a pre-development requirement rather than a post-construction communications exercise.

Local value creation. Communities should see transparent commitments on jobs, local procurement, grid investment, tax contributions and reporting. Data-centre projects are more likely to retain public support when local benefits are tangible and verifiable.

Decision-useful transparency. Investors, regulators and customers need comparable information on energy consumption, emissions, water use, renewable procurement, backup generation, location exposure and local impacts.

ERM identifies the rapid but sustainable expansion of data centres as a key 2026 sustainability issue, reflecting the sector’s material energy and water requirements as well as its wider role in the digital economy.

Efficiency still matters

Efficiency cannot replace clean-power procurement, responsible siting or watershed protection. But it remains essential.

Improving server utilisation, optimising AI-model design, using advanced cooling, recovering waste heat and extending hardware life can all reduce the resource intensity of digital services. Companies purchasing AI services should also ask whether a particular use case delivers a material operational, social or environmental benefit proportionate to its electricity and water footprint.

This is ultimately a governance issue.

Boards and executives should apply the same scrutiny to AI-infrastructure decisions that they would apply to any large industrial investment. That means clear demand forecasts, credible energy and water assumptions, risk assessment, community engagement and measurable performance indicators.

The test ahead

AI has the potential to support a more efficient and resilient economy. It can improve forecasting, automate energy management, support climate-risk analysis and accelerate scientific research.

But that potential will be undermined if the physical infrastructure behind AI is treated as invisible.

The next phase of digital leadership will not be defined only by model capability or computing scale. It will be defined by whether AI infrastructure strengthens grids, safeguards water, supports communities and accelerates — rather than delays — decarbonisation.

The sustainability test for AI is therefore straightforward: not whether it can optimise the world, but whether its own growth can be designed within planetary limits.

Sources and further reading