EU report highlights transparency gap over AI data centre energy use

A new European report has highlighted growing concerns around the lack of transparency over energy consumption linked to artificial intelligence and data centre operations, warning that current regulation does not go far enough in addressing the environmental impact of AI infrastructure.

The report identifies a significant gap in how energy use linked to AI “inference” — the operational phase where AI models process requests and generate outputs — is currently monitored and disclosed.

According to the analysis, inference now accounts for approximately 80% to 90% of total AI-related computing activity, yet much of the associated energy consumption data remains commercially undisclosed because services are delivered through proprietary models operated by major technology companies.

The report notes that Europe currently accounts for around 15% of global data centre electricity consumption, compared with approximately 45% in the United States and 25% in China.

Published by the European Environment Agency, the report argues that while the EU has begun introducing transparency requirements through the AI Act, existing measures remain focused primarily on energy consumption linked to AI model training rather than ongoing operational use.

Under the European Union AI Act, providers of general-purpose AI models will be required to document and report energy usage as part of broader transparency obligations.

However, the report states that inference-related energy demand remains “only partially regulated”, limiting access to meaningful data about AI’s overall environmental footprint and creating challenges for wider climate policy objectives.

The report also references the forthcoming Cloud and AI Development Act, which is expected to establish conditions for the expansion of data centre capacity across Europe while incorporating sustainability safeguards.

The findings come amid growing scrutiny of the rapid expansion of AI-driven data centre infrastructure, with governments and industry increasingly balancing digital growth ambitions against rising concerns around energy demand, grid capacity and carbon reduction targets.

The issue is also becoming increasingly relevant to the property and construction sectors as developers seek to deliver larger and more power-intensive data centre campuses across Europe and the UK, often requiring substantial grid connections, cooling infrastructure and long-term energy strategies.

The report suggests that greater operational transparency will become increasingly important as AI adoption accelerates and the environmental impact of inference workloads continues to grow.