Monitoring the energy sustainability of artificial intelligence: the Sustainable AI–Renewable Energy Alignment Indicator

Artificial intelligence (AI) is becoming an energy-system load and a potential instrument of decarbonisation. Existing assessments usually examine model efficiency, data-centre electricity, renewable procurement, water demand or governance separately, making it difficult to judge whether AI expansion is compatible with energy-transition objectives. This study develops the Sustainable AI–Renewable Energy Alignment Indicator (SA-REAI), a five-pillar composite framework covering AI energy pressure, renewable-grid alignment, efficiency–rebound balance, water–material constraints and governance–social readiness. The protocol combines directionally consistent normalisation, explicit evidence-quality and coverage rules, hybrid entropy–CRITIC–Best–Worst weighting, geometric aggregation, environmental penalties and Monte Carlo uncertainty analysis. Empirical feasibility is examined in two stages. A five-country European diagnostic shows what can be supported by current public data without converting incomparable proxies into a ranking. A fully reproducible synthetic organisational case then executes the complete pipeline across 15 indicators and a 30-unit benchmark. For the focal portfolio, the weakest pillar scores 55.1 and the strongest scores 73.8; separately, the post-penalty baseline score is 54.9, and the Monte Carlo median is 55.1 with a 90% interval of 53.2–56.9. Penalty parameters and normalisation thresholds account for most of the worked-example variance, while water–material performance is the weakest pillar. The worked example verifies computational operability but is not external validation. SA-REAI contributes a decision-oriented method that connects AI’s environmental burden with the conditions under which AI may assist low-carbon energy transitions, while keeping assumptions, missing evidence and uncertainty visible.