AI, energy and ESG: regulatory considerations for organisations

As the digital transformation of organisations accelerates, artificial intelligence (AI) is moving to the centre of many business strategies. McKinsey’s State of AI 2025 survey notes that in 2025, 88% of organisations have introduced at least one AI application into their operations. The rapid uptake of AI, however, comes with a notable increase in electricity demand. This surge places AI firmly within the ESG conversation and intersects with several EU legislative developments, including the AI Act, the Corporate Sustainability Reporting Directive (CSRD) and the Corporate Sustainability Due Diligence Directive (CSDDD).

AI’s expanding energy footprint

The environmental impact of AI is closely linked to the electricity needed to develop and operate advanced models. These systems are hosted in data centres that run around the clock and require substantial power to support model training and inference. According to projections published by Lawrence Berkeley National Laboratory, by 2028 more than half of the electricity going to data centres will be used for AI. The International Energy Agency has projected that, in a baseline scenario, global electricity consumption from data centres could more than double by 2030, with AI-related workloads expected to be one of the main drivers of this increase.

This increasing demand for electricity is becoming a strategic concern for many organisations, as reporting under the CSRD requires detailed information on electricity consumption. Companies that are subject to the CSRD must perform a double materiality assessment, which may require them to consider whether AI-related energy use is significant enough to feature in their environmental disclosures under the European Sustainability Reporting Standards (ESRS). Where AI substantially influences a company’s overall energy profile, this may have consequences for environmental reporting, climate planning and governance decisions.

Electricity use is determined not only by the design of AI systems, but also by the infrastructure that supports such systems. Data centre operators are implementing measures such as sophisticated cooling systems, server virtualisation and improved hardware efficiency to limit energy consumption. While such innovations contribute to more sustainable operations, they cannot entirely counterbalance the steep rise in demand caused by the rapid deployment of AI technologies.

At the same time, the origin of the electricity used for AI workloads is gaining prominence. Many organisations are examining whether the energy powering their AI tools comes from renewable sources. Under the CSRD, companies must report on their energy mix and related climate information, which means that the reliability and sustainability of the power used for digital operations may become increasingly relevant. As part of this broader shift, businesses are seeking greater visibility into the energy procurement practices of cloud and AI providers to better understand the environmental consequences of their digital supply chains.

Sustainability in AI design and development

The environmental footprint of AI is shaped not only by data centre operations but also by the choices made during model development. Decisions about model architecture, training-data volume and how frequently systems are retrained have a direct effect on electricity use. More streamlined model designs and thoughtful update cycles can significantly limit the energy needed to keep systems performing effectively.

Regulatory developments emphasising transparency and responsible deployment underscore the importance of these design choices. Organisations that are purchasing or deploying AI increasingly expect providers to disclose how models are built and maintained. This shift means that sustainability considerations are becoming part of the evaluation criteria for AI systems, influencing procurement, risk management and governance.

The global expansion of AI also brings social questions to the forefront, particularly regarding access to digital infrastructure. Although emerging economies host a large share of internet users, they account for only a small proportion of global data centre capacity. In regions where power grids are unstable or infrastructure investment remains limited, setting up or maintaining local AI capabilities can be difficult. This imbalance affects access to AI technologies and can limit opportunities for innovation.

Under the CSDDD, companies operating across borders must identify and address significant environmental and social risks in their value chains. Depending on the context, this obligation may extend to assessing how AI-related infrastructure is sourced and managed, including in regions where energy systems are less reliable or less sustainable. In such cases, organisations may need to engage more closely with service providers to ensure responsible business conduct.

Conclusion

AI continues to create opportunities for growth and transformation, but its rising energy demand and broader societal implications call for careful oversight. As legislative frameworks such as the CSRD, CSDDD and the AI Act will create an increased need for transparency, organisations will need to integrate sustainability considerations into every stage of AI adoption.

Factoring energy use into procurement decisions, infrastructure planning, model-development processes and governance structures will be essential for managing regulatory expectations and strengthening responsible business practices. By embedding ESG considerations into their approach to AI, organisations can help shape a more resilient and sustainable digital future.