Uncertainty-aware Carbon Accounting for Large-scale AI models with Market-based Attribution

ICLR 2026 Conference Submission17185 Authors

Keywords: embodied carbon, carbon footprint, large-scale AI models

Abstract: The computational demands of large-scale AI models raise significant concerns about their carbon footprint. Current carbon accounting methods for large-scale AI models suffer from three key limitations: they overlook embodied carbon (from hardware manufacturing) or model it simplistically, rely on location-based carbon attribution that fails to reflect individual corporate efforts to decarbonize (e.g., via Power Purchase Agreements (PPAs)), and are deterministic, ignoring inherent uncertainties. This paper proposes CarbonPPA, an uncertainty-aware carbon accounting model with market-based attribution for large-scale AI models. CarbonPPA integrates market-based carbon intensity to accurately account for the impact of PPAs and employs probabilistic modeling to capture uncertainties in the carbon accounting for AI models arising from spatiotemporal variations in manufacturing and operation, as well as evolving efficiency. We develop a comprehensive carbon dataset by aggregating related data from diverse sources, then we implement a simple yet effective Kernel Density Estimate (KDE) to the distribution of the parameters from the collected dataset. We compare CarbonPPA with LLMCarbon, the state-of-the-art carbon accounting model. The deviation of the accounting result is significant, reaching up to 251.58%

Primary Area: alignment, fairness, safety, privacy, and societal considerations

Submission Number: 17185