A probabilistic Carbon Valley dataset for AI-infrastructure pressure on global carbon budgets

Temporal Carbon Debt in the AI Transition: Carbon Valley Model, Supplementary Information, and Processed IEA Trajectories This archive contains the reproduction package and supplementary materials for the study “Temporal Carbon Debt in the AI Transition.” It provides the code, processed scenario trajectories, documentation, and Supplementary Information required to reproduce and interpret the system-dynamics analysis of the temporal mismatch between immediate AI-related infrastructure emissions and delayed system-level emissions savings. The model quantifies cumulative carbon debt arising during rapid AI infrastructure expansion under alternative International Energy Agency (IEA) Energy and AI (2024) scenarios. It includes representations of: electricity-demand growth under the IEA scenario framework, exponentially declining grid carbon intensity, marginality-adjusted operational emissions, embodied emissions from discrete hardware refresh cycles, delayed logistic diffusion of mitigation-oriented AI savings, mitigation-coupling sensitivity, cumulative carbon debt and flow-level breakeven, Jevons rebound dynamics, and stylized policy levers for operational flexibility, including temporal load shifting and dynamic power capping. The archive is intended to support full transparency and reproducibility of the results reported in the manuscript and Supplementary Information. Contents Reproduction script implementing the full Carbon Valley system-dynamics model Supplementary Information describing the mathematical formulation, parameterization, scenario construction, numerical implementation, uncertainty propagation, policy-lever formulation, and replication workflow Processed IEA scenario trajectories used for annual interpolation and model calibration README with setup instructions and execution workflow requirements.txt listing Python dependencies LICENSE metadata and output structure for reproducibility Input data The model is calibrated using the International Energy Agency (IEA, 2024) Energy and AI scenario framework. Where applicable, the archive includes processed scenario trajectories derived from the IEA data annex. Users should consult the README for input-data requirements and execution instructions. Environment Python 3.10 Standard numerical libraries as listed in requirements.txt