AI, energy consumption and sustainability reporting

Short answer: AI has a substantial energy and environmental footprint, and it is becoming increasingly visible in regulation. The AI Act requires providers of general-purpose AI (GPAI) models to document information including energy consumption. The CSRD (Directive (EU) 2022/2464) requires large companies to report on their environmental impacts, including the energy use of their AI and IT activities.

Why energy is an AI issue

Training and running large AI models demands heavy compute, and with it electricity, cooling and water in data centres. As AI use scales, the question becomes relevant for both costs and sustainability targets.

Consumption lies not only in one-off training, but above all in daily inference: every query to a model costs energy. At large scale that adds up quickly. Attention is therefore shifting from the training phase alone to the full lifecycle of a model.

AI Act: documentation duty for GPAI

The AI Act focuses on transparency at the top of the chain. Providers of general-purpose AI models must draw up technical documentation that includes the known or estimated energy consumption of the model.

This gives downstream users and supervisors insight into a model's footprint and connects with the regulation's broader documentation and transparency obligations. It is not a consumption cap, but an information duty.

CSRD: reporting on environmental impact

The Corporate Sustainability Reporting Directive broadens the view to the whole company. Large firms must report in standardised form on sustainability matters, including energy consumption and climate impact.

For organisations that lean heavily on AI, the energy use of data centres and models thus becomes part of mandatory reporting. AI here is not a separate track but a line item within broader environmental reporting.

The CSRD works with a double-materiality approach: a company reports both on the impact of its activities on the environment and on the risks that environmental issues pose to itself. For AI that means: not only reporting consumption, but also considering dependence on energy-intensive infrastructure.

Two perspectives that complement each other

The AI Act looks at the product (the model and its footprint); the CSRD looks at the company (its total environmental impact). Combining both gives a fuller picture — similar to how cybersecurity is governed by several frameworks at once.

What to do

• Measure your AI energy use: map training, inference and data-centre use.
• Request GPAI documentation: use the energy information from your model supplier.
• Link AI to your CSRD process: include AI energy in environmental reporting if it applies to you.
• Make sustainable choices: more efficient models, green data centres and reuse of existing models.
• Document assumptions: record measurement methods and estimates.

Energy efficiency and compliance run together here: less consumption means lower costs and an easier story in reporting.