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

This repository contains the submission-ready dataset and reproducibility package for the Data Descriptor “A probabilistic Carbon Valley dataset for AI-infrastructure pressure on global carbon budgets.” The dataset supports reproduction and extension of the Carbon Valley cumulative carbon-accounting framework used in the associated study “Rapid artificial intelligence deployment increases near-term pressure on global carbon budgets.” It provides harmonized scenario inputs, Monte Carlo parameter draws, annual emissions and avoided-emissions outputs, embodied-emissions refresh-cycle pulses, marginality-adjusted operational emissions, truncated cumulative carbon-debt trajectories, flow-level breakeven indicators, derived sensitivity outputs, validation records, metadata and executable reproducibility code. The archive covers four artificial-intelligence infrastructure deployment pathways: Lift-Off, Base, Headwinds and High Efficiency, over the period 2024–2035. It includes 40,000 scenario-level Monte Carlo draw records and expanded annual outputs across the full-coupling reference case, μ = 1.0, and mitigation-coupling sensitivity cases, μ = 0.50, μ = 0.25 and μ = 0.10. The resulting annual-output archive contains 1,920,000 annual records and 1,920,000 truncated cumulative-debt records. The reference-output and validation files verify that the Lift-Off full-coupling reference path reproduces the published calibration target of approximately 2.85 Gt CO₂e of truncated cumulative carbon debt and a flow-level breakeven year of approximately 2031.8. Cumulative carbon debt is represented as a non-negative truncated stock and should not be interpreted as a signed cumulative balance or cumulative repayment trajectory. The package includes: harmonized scenario input records; Monte Carlo parameter draws; annual emissions, avoided-emissions and positive-pressure outputs; truncated cumulative carbon-debt trajectories; reference-output paths for mitigation-coupling cases; derived indicators for breakeven timing, peak debt, lag penalty and sensitivity; validation checks for accounting consistency, monotonicity, calibration and reproducibility; metadata dictionaries and manifest files; figures and Python code required to regenerate the dataset. The raw IEA Energy and AI annex is not redistributed in this package; only processed scenario trajectories used for the Carbon Valley dataset are included.