Method

How we flag

Stable cite for the three descriptive flags on the AI Datacenters investigation. These rules are computed from the live dataset. This page restates the formulas already used by the dashboard — no new findings, no editorial claims.

Patterns warrant review, not accusations. These flags surface entries that deviate from peers or expose data gaps. They are descriptive, not editorial. Investigative judgments belong in published findings with sourcing and right of reply.

Subsidies > $1M per claimed job

→ benefit-cost review

Disclosed subsidy ÷ claimed jobs exceeds $1M.

Formula

amount_usd / claimed_jobs > 1000000

Applies when

claimed_jobs IS NOT NULL AND claimed_jobs > 0 AND amount_usd IS NOT NULL

Source app.subsidy

No power source disclosed

→ data gap

No primary energy source recorded — cannot assess emissions from the record.

Formula

verified = true AND primary_power_source IS NULL

Source app.datacenter

Greenwash gap

→ claim verification

Operators ≥30pp behind commitment (commitment % − actual carbon-free MW %).

Formula

gap_pp = commitment_pct − (carbon_free_mw / total_mw × 100)

Flag when

gap_pp >= 30

Universe: verified facilities; operators with a public commitment_renewable_pct; total MW > 0. MW is capacity_mw + capacity_mw_planned. solar, wind, hydro, geothermal, nuclear.

Source app.brand + app.datacenter

Sources

Rules are those already used by the live investigation dashboard (investigationStats red flags and phase5RedFlags).

This page documents the computational rules. Counts on the live investigation change as the dataset is updated.

Cite this URL for the method. Cite the live investigation for current counts.

Live investigation