PUBLISHER: Global Infrastructure Sherpa | PRODUCT CODE: 2161112
PUBLISHER: Global Infrastructure Sherpa | PRODUCT CODE: 2161112
Scale separates the reported numbers more than architecture does. Across 770 European facilities reporting to the Commission, most under a legal obligation, average power usage effectiveness runs 1.63 at 100–500 kW of installed IT power and 1.21 above 10 MW. The consumption-weighted EU aggregate is 1.36. No architecture effect documented anywhere in this report is as large as the 0.42 between those two size cohorts - and because the cohorts also differ in climate, vintage, design and load factor, the report sizes that prize rather than claiming it, and sets out what would be needed to claim it.
The measurement itself moves. Server fan power is metered as IT load and sits in the PUE denominator; move that work to pumps and coolant distribution units on the facility side and it is metered as overhead. A facility at 150 units of total energy and 100 of IT reads 1.50. Reclassify 5 units of fan work with no physical change and the ratio is 1.579 - the denominator shrank, total energy did not move. A tender scored on PUE alone can award to the configuration that consumes more.
Nineteen chapters in three parts, with three annexes. Foundations establishes what counts as cooling energy across eight provenance classes, what PUE measures and what it does not, what 770 reporting facilities and four published fleets return, and what a tenth of a PUE point is worth priced in twenty markets. The physical system covers the nine measured retrofit cases with a cost and a payback, subsystem cost per kilowatt, the forty-location design-condition matrix, ten named configurations quoted on their makers' own conditions, the rack as a hydraulic object, the wetted system, commissioning and witnessed testing, capacity during maintenance, and heat reuse. Constraint, compliance and money covers the binding connection constraint and large-load tariffs, a thirty-jurisdiction compliance register recording what binds at what threshold and from when, the 2027 refrigerant cap and the PFAS calendar, where someone else pays, one 20 MW asset with the bill rebuilt from the filed tariff, and instruments to 2033 against eight dimensions of diligence.
Nineteen chapters in three parts, three annexes and 426 numbered references across 517 pages. Global and vendor-neutral: thirty jurisdictions in the compliance register, forty locations in the design-condition matrix, twenty markets in the value-of-efficiency screen, and more than sixty named equipment configurations and integrated racks quoted on their makers' own published conditions. Part I - scope, definitions and the eight provenance classes; the measurement problem and the boundary that moves PUE; the benchmark record from 770 reporting facilities and four published fleets; and what a tenth of a PUE point is worth in twenty markets. Part II - the nine measured retrofit cases in order of cost and payback; subsystem cost per kilowatt; climate and water limits at forty locations; ten configurations specified on makers' conditions; the rack as a hydraulic object; the wetted system's chemistry, materials and fluids; commissioning, acceptance and measurement; capacity during maintenance; and heat reuse. Part III - the binding constraint of connection and large-load tariffs; the thirty-jurisdiction compliance register; the 2027 cap, the PFAS calendar and the supplier base; where someone else pays; one 20 MW asset rebuilt from the filed tariff; and instruments to 2033 across eight dimensions of diligence. Annex A puts every headline quantity in one place with its provenance class and chapter; Annex B is the full forty-location design-condition matrix; Annex C is the glossary.
Engineering directors specifying or retrofitting cooling plant; operators pricing a hall conversion or a density increase; infrastructure investors and lenders underwriting a data centre asset; and advisers who need the underlying instrument rather than a summary of it.
Every headline quantity carries a provenance class - eight are defined in chapter 1 - and is listed in Annex A with that class and the chapter that established it. Individual figures carry their own effective date: a reporting year, a tariff filing date, an instrument date, a document edition, stated at the figure. The single cut-off for the report is 14 September 2026; a figure dated after the cut-off is not carried. Equipment performance is quoted on the manufacturer's own published conditions rather than normalised to a common basis, and the conditions are stated with the figure. Where the record cannot support a conclusion the report says so rather than closing the gap. Independent: not commissioned, endorsed or sponsored by ASHRAE, the European Commission, any national energy authority, any utility or any equipment manufacturer.
The seventeen findings.