The AI Energy-Demand Forecast Scorecard: a reproducible audit of how the field forecasts data-centre electricity demand

A reproducible audit of how the field forecasts AI/data-centre electricity demand. Eleven published forecasts (IEA, Gartner, Deloitte, McKinsey, LBNL, EPRI, S&P, BloombergNEF, Goldman Sachs, BCG) are scored on four signals: transparency (confirmable-at-source vs reproducible-from-method), dispersion, self-revision behaviour, and heterogeneity of units/scope/horizon. Findings: only 3 of 11 figures are reproducible from public method and data; the single highest figure (BCG ~1,050 TWh) cannot be verified at all; the forecasts use six incompatible formats and disagree by multiples without sharing a scope definition; and every documented self-revision is upward. The scorecard measures forecaster behaviour — it is not itself a forecast or a market call. Full dataset and stdlib-only scripts included; edit the CSV and re-run to reproduce every number and the chart.