Chapter 09 · Sustainability & AISustainability & AI

Life-cycle assessment of AI

Meaning statusEstablishedSource recordDirect document linkedWhy these are different

Definition

Applying life-cycle assessment methods to AI systems — quantifying environmental impacts from raw-material extraction through hardware manufacture, training, use and end of life.

References

ITU-TL.1410 — Methodology for environmental life-cycle assessments of ICT goods, networks and services

This source provides part of the technical or institutional basis for the definition.

arXivLuccioni et al. — Estimating the Carbon Footprint of BLOOM

This source supports the explanation of how the term is applied, measured or governed in practice.

Overview

What it means

LCA extends AI accounting beyond electricity: embodied emissions and resource use of chips and servers, infrastructure construction, and disposal all enter the boundary. ITU-T L. 1410 provides the methodology for ICT goods, networks and services; academic studies of large models (e. g. Luccioni et al. on BLOOM) demonstrate cradle-to-grave estimates.

How it is used

Practitioners use AI LCA to compare deployment options, inform procurement and avoid burden-shifting — for example, cutting operational energy by using more hardware. Corporate inventories apply it to ICT estates generally, with AI workloads an increasing share.

Why it matters

Operational-only accounting misses a large and growing slice of AI's impact. LCA disciplines the comparison between AI's environmental costs and benefits — and exposes when 'efficient AI' claims ignore what it took to build the machine.