Overview
ISO/IEC TR 20226:2025 - Information technology - Artificial intelligence - Environmental sustainability aspects of AI systems - is a Technical Report that provides a structured overview of environmental sustainability considerations for AI systems across their life cycle. It summarizes relevant topics (energy, water, cooling, carbon, waste, location, transportation), outlines potential metrics for measurement, and positions those aspects within AI ecosystems, life cycles and supply chains. As a TR, it is informative guidance rather than a prescriptive standard.
Key technical topics and considerations
The report organizes environmental sustainability topics into clear, actionable areas, including:
- Energy consumption and sources
- Types and sources of energy, renewable procurements (e.g., PPAs, offsets)
- Energy accounting for CPUs, GPUs and accelerated processing units (APUs)
- Energy efficiency metrics and measurement approaches
- Carbon footprint and accounting
- Measurement vs. estimation considerations
- Carbon leakage and applicable carbon accounting methods
- Water use and cooling
- Cooling types, water consumption and post-cooling reuse
- Geographic and infrastructure considerations
- Location impacts, distribution of energy sources, freight and handling
- Waste and e‑waste
- Types of waste, device abandonment, collection, recovery and disposal practices
- AI system perspectives
- Ecosystem view (cloud, edge, data sources, resource pools)
- Life cycle stages: inception, design, training, deployment, operation, continuous validation, retirement
- Supply chain elements
- Mines, smelters/refiners, component manufacturers, assemblers, distributors, internet/domestic infrastructure
- Measuring and metrics
- Candidate metrics and measurement points for system-level, component-level and process-level assessment
- Contextual and regulatory linkage
- Relationship to life cycle assessment standards (ISO 14040/14044), ISO/IEC 5338 life cycle framework, the EU AI Act and international initiatives (e.g., London Declaration)
Note: The TR explicitly does not prescribe how AI can be used to improve sustainability outcomes; it focuses on impacts and measurement.
Practical applications and who uses it
ISO/IEC TR 20226:2025 is useful for:
- AI developers and architects - to design systems with reduced environmental footprint and select efficient compute and cooling strategies.
- Data center operators and cloud providers - for benchmarking energy use, cooling choices and renewable sourcing.
- Sustainability officers and LCA practitioners - to align AI system assessments with established life cycle and carbon accounting practices.
- Procurement teams and IT asset managers - to consider supply chain impacts, device end‑of‑life and e‑waste processes.
- Regulators, auditors and policy makers - to inform guidance, conformity assessment and policy on AI environmental impacts.
- Researchers - to identify measurement gaps and harmonize metrics for energy and carbon in AI.
Related standards and keywords
Related standards and frameworks referenced include ISO 14040 / ISO 14044 (life cycle assessment) and ISO/IEC 5338 (AI system life cycle).
SEO keywords: ISO/IEC TR 20226:2025, environmental sustainability, AI systems, carbon footprint, energy consumption, life cycle assessment, e‑waste, AI supply chain, AI metrics.