Carbon Emissions from Model Training

When the energy consumption and carbon emissions from training large models are considerable and grow rapidly with model scale; due to unmitigated control gaps, under decarbonization pressure, an organization's AI-related carbon footprint faces increasing scrutiny from investors, customers, and regulators, yet is often excluded from existing environmental disclosure scope, triggering compliance exposure and operational reputational costs.