Quantifying the Climate Risk of Generative AI: Region-Aware Carbon Accounting with G-TRACE and the AI Sustainability Pyramid

TL;DR

GenAI introduces climate-risk through scalable, regionally varied energy use across training and especially inference. The authors present G-TRACE, a cross-modal, region-aware carbon accounting framework, and calibrate it with a viral #Ghibli case to quantify energy and CO2 across platforms and regions; they then operationalize findings via the AI Sustainability Pyramid to guide governance from measurement to stewardship. The work highlights that identical workloads can yield vast regional emission disparities and that viral user behavior amplifies emissions, underscoring the need for carbon-aware routing, energy-aware defaults, and standardized telemetry. Overall, the paper provides a concrete, data-driven path to align GenAI deployment with decarbonization goals by linking regional factors, per-output energy, and governance gates.

Abstract

Generative Artificial Intelligence (GenAI) represents a rapidly expanding digital infrastructure whose energy demand and associated CO2 emissions are emerging as a new category of climate risk. This study introduces G-TRACE (GenAI Transformative Carbon Estimator), a cross-modal, region-aware framework that quantifies training- and inference-related emissions across modalities and deployment geographies. Using real-world analytics and microscopic simulation, G-TRACE measures energy use and carbon intensity per output type (text, image, video) and reveals how decentralized inference amplifies small per-query energy costs into system-level impacts. Through the Ghibli-style image generation trend (2024-2025), we estimate 4,309 MWh of energy consumption and 2,068 tCO2 emissions, illustrating how viral participation inflates individual digital actions into tonne-scale consequences. Building on these findings, we propose the AI Sustainability Pyramid, a seven-level governance model linking carbon accounting metrics (L1-L7) with operational readiness, optimization, and stewardship. This framework translates quantitative emission metrics into actionable policy guidance for sustainable AI deployment. The study contributes to the quantitative assessment of emerging digital infrastructures as a novel category of climate risk, supporting adaptive governance for sustainable technology deployment. By situating GenAI within climate-risk frameworks, the work advances data-driven methods for aligning technological innovation with global decarbonization and resilience objectives.

Figures (4)

• Figure 1: Overview of generative tasks and modalities (Multimodal, Video, Audio, Image, and Text).
• Figure 2: GenAI Model Performance Evolution: MMLU Performance Benchmarks vs. Model Release Timeline (2022–2025) with Market Adoption Indicators. Models are categorized into three performance tiers based on MMLU scores: Emerging, Human-Level, and Superhuman. This figure shows the accelerating advancement of GenAI models toward superhuman performance within a three-year development cycle.
• Figure 3: G-TRACE framework architecture. Three stages—Trend Tracker, Device Simulator, and the region-aware CO2 Estimator—convert social activity signals and workload metadata into energy and CO2 estimates with uncertainty bounds. Device tiers span Laptop with and without GPUs and mobile classes (H/M/L). Region-specific grid factors enable fair, cross-modal comparisons and sensitivity analyses.
• Figure 4: Operationalizing G-TRACE via the AI Sustainability Pyramid. L1–L2: measurement; L3–L4: optimization & portfolio governance; L5–L6: innovation & standards; L7: climate stewardship. At each level, gates act as checkpoints using G-TRACE indicators (such as CO2 per output, lifecycle contributions, and regional energy factors) to decide whether systems are ready to advance to the next stage of sustainability.