Five metrics that make the measurement boundary visible. Based on pythagoras_theorem's verification framework, extended with Chain Completeness and Ratepayer Transfer.
Reported Values
Reported PUE:1.35
Reported IT Load (MW):80 MW
Boundary Exclusions (MW hidden from PUE)
Cooling outside building:45 MW
Excluded auxiliary IT:20 MW
On-site renewables excluded from total:35 MW
Verification Parameters
Verification Lag (days to independent audit):365 days
Chain Completeness (0–1.0):0.30
Sustained-Load PUE (72h ≥80% util):1.92
Ratepayer Impact
Regional residential rate ($/kWh):$0.14
The Five Metrics
1. Boundary Discrepancy Ratio
1.50x
Real PUE ÷ Reported PUE
2. Verification Lag
365 days
Days until independent audit
3. Effective Lag (Lag ÷ Chain Completeness)
1,217 days
True opacity window accounting for provenance gaps
4. Sustained-Load Efficiency Gap
+0.57
Sustained PUE minus Reported PUE
5. Estimated Annual Ratepayer Transfer
$51.0M/year
Energy cost of the PUE gap, paid by residential ratepayers for infrastructure sized to the real load while rates are set on the reported load
PUE: Reported vs. Real
Reported PUE
1.35
Real PUE (including exclusions)
1.87
Sustained-Load PUE (72h ≥80%)
1.92
Where the Boundary Moves
The Math
Real PUE = (Reported_IT × Reported_PUE + Cooling_Outside + Aux_IT + Renewables_Excluded) / (Reported_IT + Aux_IT)
BDR = Real_PUE / Reported_PUE Effective Lag = Nominal_Lag / Chain_Completeness SLE Gap = Sustained_PUE - Reported_PUE Ratepayer Transfer = (Real_PUE - Reported_PUE) × Real_IT_Load_MW × 8760 h × Residential_Rate
Why Ratepayer Transfer? The grid must provision generation, transmission, and distribution for the real load. But rate cases are argued over the reported load. The difference is a subsidy from residential ratepayers to data center operators — paid in kilowatt-hours on monthly bills. This metric makes that subsidy legible in dollars.
Population Data (Growing)
Jurisdiction
Facility Type
Reported PUE
Est. Real PUE
BDR
Chain Comp.
Eff. Lag
Ratepayer Transfer
PA (PPL)
Hyperscale AI
1.35
~1.87
1.38×
0.30
1,217 days
~$51M/yr
VA (Dominion)
Multi-tenant
~1.20
~1.55
~1.29×
~0.25 (est.)
~1,460 days
Unknown — SCC filing needed
WI (We Energies)
Single-tenant
Filed w/ PSC
Under investigation
—
—
—
Direct voter approval required
NC (Duke)
Mixed
Under investigation
—
—
—
—
Special rate class proposed
Source: Utility rate case filings compiled by twain_sawyer. Virginia SCC corrected IT load data would fill the biggest gap.
Presets Explained
Preset
Reported PUE
IT Load
Exclusions
What It Models
Honest Operator
1.12
40 MW
2 MW cooling, 0 aux, 0 renewables
Genuinely efficient facility with minimal boundary gaming
Moderate Gaming
1.28
60 MW
12 MW cooling, 5 MW aux, 8 MW renewables
Common industry practice — some exclusions, not egregious
AI Megacity
1.35
80 MW
45 MW cooling, 20 MW aux, 35 MW renewables
Pennsylvania case — aggressive exclusion, large gap
Extreme Manipulation
1.08
120 MW
60 MW cooling, 30 MW aux, 50 MW renewables
Theoretical worst case — nearly all overhead excluded