This informal CPD article, ‘AI, Energy, Water and Infrastructure Resilience - Sustainability considerations for data centre growth‘, was provided by IFRS Lab, a leading ESG advisory and training institution committed to advancing sustainability.
Artificial intelligence is increasingly discussed as an economic, industrial and public-policy priority. Its potential benefits include productivity gains, improved data analysis, automation of routine tasks and new tools for scientific and commercial innovation. These benefits help explain why governments, investors and businesses are increasing their focus on AI capability.
At the same time, the expansion of AI depends on physical infrastructure. Large-scale model training, high-volume inference and enterprise deployment require data centres, electricity supply, cooling systems, grid connections, specialist equipment and resilient communications networks. As AI adoption grows, the environmental and infrastructure implications of this expansion require careful assessment.
The central issue is not whether AI should develop. The more practical question is whether energy systems, water resources, planning controls and sustainability governance are being adapted quickly enough to support AI growth responsibly. Recent analysis suggests that data centre expansion is becoming a significant consideration for electricity systems and resource planning in several regions (1).
Electricity demand and grid capacity
Modern AI systems require substantial computational capacity. Training large models and serving them to millions of users can involve continuous operation of high-density computing equipment. This increases electricity demand directly within data centres and indirectly through associated cooling, networking and backup systems.
Estimates cited in recent policy discussion suggest that global data centre electricity demand could reach approximately 1,050 terawatt-hours by 2026, depending on deployment patterns and assumptions about efficiency improvements (1). Such forecasts should be treated with care because demand projections vary across models, but they indicate the scale of infrastructure planning now being considered.
In the United States, electricity consumption is expected to reach record levels in 2026 and 2027, with AI and data centre growth identified as one contributing factor alongside wider electrification and industrial demand (5). This illustrates why AI infrastructure is now relevant not only to technology companies but also to utilities, regulators, transmission planners and local authorities.
Grid capacity is a practical constraint. In some locations, new data centre projects face delays due to interconnection queues, transmission limitations or concerns about local electricity costs. Some jurisdictions have introduced restrictions or moratoriums on new projects while they assess grid capacity, land use and resource impacts. These responses indicate that digital infrastructure can no longer be assessed separately from energy-system planning.
Water use and cooling requirements
Electricity demand is only one part of the sustainability picture. Data centres also require cooling to maintain safe operating temperatures for servers and supporting equipment. Depending on the cooling technology used, local climate and facility design, water consumption can be significant.
Water use is particularly important in regions already experiencing water stress, drought risk, population growth or competing industrial demand. Large facilities may place additional pressure on municipal or regional water systems if cumulative impacts are not properly assessed. A study has highlighted the need to consider data centre growth in relation to both water and energy systems (7).
The water issue also includes indirect consumption. Electricity generation can require water depending on the power source and generation technology. This means the environmental footprint of AI infrastructure may involve both direct cooling water and indirect water use embedded in the electricity supply. Research on AI infrastructure increasingly emphasises the connection between compute demand, energy generation and water systems (4).
A balanced approach should recognise that not all data centres have the same water profile. Air cooling, liquid cooling, closed-loop systems, recycled water use, dry cooling and climate-appropriate design can change the level of water impact. The key governance question is whether developers and authorities are assessing these choices transparently and consistently.
AI infrastructure growth
Implications for emissions and net-zero plans
AI infrastructure growth may also affect emissions pathways. Many organisations have adopted climate targets, renewable-energy procurement plans and net-zero commitments. Higher electricity consumption can complicate these commitments if additional demand is met through fossil-fuel generation or if grid decarbonisation does not keep pace.
Recent commentary has noted that rising data centre demand may encourage some utilities or governments to maintain conventional generation capacity for reliability reasons (3). This does not mean AI growth automatically prevents decarbonisation. It does mean that clean-energy procurement, transmission expansion, demand flexibility and efficiency improvements become more important as compute demand rises.
The emissions impact of AI will depend on several variables, including location, grid mix, renewable-energy availability, data centre efficiency, server utilisation, cooling design and the extent to which AI models become more energy efficient over time. For this reason, simple claims that AI is either inherently sustainable or inherently unsustainable are unlikely to be useful. The more appropriate assessment is context-specific and evidence-based.
For sustainability reporting, organisations using or developing AI may increasingly need to consider whether data centre-related energy use is material to their climate disclosures, value-chain emissions assessments or transition planning. This may be relevant to scope 2 emissions from purchased electricity and, in some cases, scope 3 emissions associated with digital services and cloud infrastructure.
Infrastructure planning and governance
Data centres are often treated as real-estate, technology or inward-investment projects. The scale of AI-related growth suggests that they should also be treated as strategic infrastructure. They can affect grid reliability, water planning, land use, local communities, emissions trajectories and industrial development strategies.
Effective governance requires coordination between several parties: developers, utilities, planning authorities, environmental regulators, water authorities, local communities and corporate customers. Decisions about location, grid connection, cooling systems, backup power, waste heat recovery and renewable-energy procurement can materially influence environmental outcomes.
There are also economic trade-offs. Data centres can support investment, employment, digital capability and national competitiveness. However, these benefits need to be weighed against resource constraints and local impacts. A balanced planning framework should therefore consider both economic value and cumulative environmental pressure.
The World Economic Forum has described data centre energy demand as a strategic infrastructure issue, reflecting the wider shift from viewing digital systems as purely virtual to recognising their physical resource requirements (6). This framing is useful because it encourages long-term planning rather than reactive project-by-project decision-making.
Possible responses and good practice considerations
Several measures can reduce the environmental pressure associated with AI infrastructure. These include improving server utilisation, developing more energy-efficient AI models, using advanced cooling systems, recovering waste heat, procuring additional renewable electricity, locating facilities in areas with suitable grid and water capacity, and designing transparent resource-impact assessments.
Waste heat recovery is one example of a potential circular approach. The European Commission has noted emerging research into the use of AI data centre waste heat for applications such as water purification and carbon capture (2). Such solutions are not universal, and their effectiveness depends on technical feasibility, cost, local demand and infrastructure integration. However, they show how data centre planning can move beyond basic electricity supply toward broader resource efficiency.
Policy responses may also include clearer disclosure requirements, environmental-impact assessment thresholds, water-use reporting, grid-connection planning rules and incentives for low-carbon or low-water design. These measures should be designed carefully so that they support responsible infrastructure development without unnecessarily preventing beneficial innovation.
For businesses, the practical lesson is to treat AI adoption as part of sustainability and risk management. Organisations should understand where AI services are hosted, how cloud providers manage energy and water impacts, and whether digital transformation strategies are aligned with wider climate and resource-efficiency objectives.
Conclusion
AI will continue to expand because it offers significant economic and operational value. However, its growth is increasingly connected to electricity systems, water resources, land-use planning and emissions management. This makes AI not only a technology issue but also an infrastructure resilience issue.
A balanced sustainability discussion should therefore avoid both excessive optimism and excessive alarm. The future environmental impact of AI will depend on choices made now about grid investment, renewable-energy deployment, cooling technology, water management, data centre siting, model efficiency and transparent governance.
The most important shift is to recognise that digital transformation has physical consequences. Responsible AI development will require infrastructure planning that is technically robust, environmentally informed and capable of balancing innovation with resource resilience.
We hope this article was helpful. For more information from IFRS Lab, please visit their CPD Member Directory page. Alternatively, you can go to the CPD Industry Hubs for more articles, courses and events relevant to your Continuing Professional Development requirements.
References
- Brookings Institution. (2026). Global energy demands within the AI regulatory landscape. Brookings. https://www.brookings.edu/articles/global-energy-demands-within-the-ai-regulatory-landscape/
- European Commission. (2026). AI data centre waste heat could be used for water purification and carbon capture. European Commission Environment Directorate-General. https://environment.ec.europa.eu/news/ai-data-centre-waste-heat-could-be-used-water-purification-and-carbon-capture-2026-03-30_en
- Financial Times. (2026). AI-driven electricity demand and fossil fuel pressures. Financial Times. https://www.ft.com/content/7b3f3142-c0f1-47a9-a011-15897dfe50d8
- Nature Sustainability. (2025). The water and energy footprint of artificial intelligence systems. Nature Sustainability. https://www.nature.com/articles/s41893-025-01681-y
- Reuters. (2026, May 12). US power use to beat record highs in 2026 and 2027 as AI use surges, EIA says. Reuters. https://www.reuters.com/business/energy/us-power-use-beat-record-highs-2026-2027-ai-use-surges-eia-says-2026-05-12/
- World Economic Forum. (2026). Data centre energy demand is becoming a strategic infrastructure issue. World Economic Forum. https://www.weforum.org/stories/2026/03/data-centre-energy-demand-strategic-asset/
- World Resources Institute. (2026). US data center growth impacts water and energy systems. World Resources Institute. https://www.wri.org/insights/us-data-center-growth-impacts
