Introducing Power Compute Effectiveness (PCE), a Provisioned Capacity Allocation Metric for the New Data Center Era (October 2026)

Overview

The paper starts from the observation that, as rack densities and AI demand grow, operators are constrained less by floor space than by how provisioned electrical capacity is distributed inside existing infrastructure. In many facilities a significant share of contracted and installed power goes to cooling, electrical losses and auxiliary loads before reaching IT equipment. Existing metrics mostly evaluate operating efficiency, so they do not expose this structural constraint at design or planning stage. PCE is proposed to fill that gap.

The metric is PCE = P_ITcompute,prov / P_facility,prov under a declared facility boundary and redundancy basis (N, N+1, 2N), a dimensionless value between 0 and 1. P_ITcompute,prov is derived, not measured: provisioned facility capacity minus provisioned cooling and heat-rejection power, electrical conversion and distribution losses, and auxiliary loads. Two declarations are mandatory before calculation, the electrical boundary and the redundancy basis, and every variable is evaluated on that basis.

Source rules for the four inputs are detailed. Facility capacity rests on the single-line diagram, commissioning report and nameplates, with one disclosed basis such as the utility interconnection agreement, transformer rating, main switchgear nameplate (kVA converted to kW at a declared power factor) or aggregate UPS output. Cooling comes from mechanical schedules, OEM submittals and panel schedules, with motor horsepower converted at HP x 0.746, and modelled values without OEM documentation are not permitted. Metered operating data may validate but never serve as the primary input.

The paper compares PCE with PUE, DCiE, IUE, TUE, CUE and WUE and stresses that it complements them: the others address operating energy, utilization, emissions or water, while PCE addresses design-level capacity allocation. Two worked examples are given. In a 20 MW N+1 Tier III colocation site, 4.8 MW cooling, 0.7 MW losses and 0.4 MW auxiliaries leave 14.1 MW for IT, PCE 0.705. In a 100 MW hyperscale campus, 16.0 MW cooling, 3.5 MW losses and 1.5 MW auxiliaries leave 79.0 MW, PCE 0.790.

The scenarios are explicitly representative composites that illustrate the method; they set no target values and prescribe no design. Suggested applications are retrofit prioritization, cooling-architecture comparison, capacity planning and portfolio analysis. The paper's stated limits and positioning: PCE is independent of workload type and actual utilization, needs no new metering, and can be computed before commissioning. It is licensed CC BY-SA 4.0 and dated October 2026.

Key points

1. PCE = P_ITcompute,prov / P_facility,prov, dimensionless between 0 and 1; P_ITcompute,prov = P_facility,prov - (P_CoolReject,prov + P_ElecLoss,prov + P_Aux,prov).
2. Two declarations are mandatory before calculation: the facility electrical boundary and the redundancy basis (N, N+1, 2N).
3. Inputs must come from utility agreements, nameplates, issued-for-construction drawings, commissioning reports and OEM submittals; modelled values without OEM documentation are not allowed.
4. Motor horsepower converts at HP x 0.746; kVA ratings convert to kW at the declared power factor.
5. Planning conditions (IT load percentage, ambient temperature, supply/return temperatures) must be disclosed with the PCE value.
6. Metered operating data may validate the result but cannot be the primary input.
7. PCE differs from PUE, DCiE, TUE, CUE and WUE: those measure operating energy or resources, PCE measures provisioned capacity allocation; it complements rather than replaces them.
8. Example 1: 20 MW N+1 colocation; 4.8 MW cooling, 0.7 MW losses, 0.4 MW auxiliaries, 14.1 MW for IT, PCE 0.705.
9. Example 2: 100 MW hyperscale campus; 16.0 MW cooling, 3.5 MW losses, 1.5 MW auxiliaries, 79.0 MW for IT, PCE 0.790.
10. Applications: retrofit prioritization, cooling-architecture comparison, capacity planning and portfolio-level analysis.

Using it in design

• Compute PCE automatically from the design model: utility/transformer capacity as P_facility, peak cooling electrical input at design ambient, UPS/transformer/busway losses and auxiliary panel loads, all on the declared redundancy basis.
• Show stranded capacity (MW) next to PUE so users see the trade-off between cooling choices (chiller kW/ton, liquid cooling) and sellable IT MW.
• Require boundary and redundancy declarations in the report header.
• Take cooling inputs from equipment schedules and OEM submittals; flag modelled values and exclude them from any PCE claim.
• Use the two worked examples (0.705 and 0.790) as calculation tests for the tool.
• For retrofit prioritization, rank portfolio sites by PCE and by stranded MW.

Who should read it

Data-centre designers, capacity planners and investment decision makers evaluating capacity allocation at design stage.

Document contents

1. Introduction
2. Compliance with Open Compute Project Tenets
3. PCE Framework and Methodology
4. Metric Definition
5. Required Declarations
6. Variable Definitions
7. Six requirements to claim PCE compliance
8. Relationship to Existing Metrics
9. Applied Scenarios
10. Enterprise Colocation: 20 MW Tier III Facility (N+1)
11. Hyperscale Campus: 100 MW Facility
12. Observations and Adoption
13. Operational Applications of PCE
14. Conclusion
15. Glossary
16. References
17. License

Related standards

• ISO/IEC 30134-2 (PUE)
• The Green Grid DCiE / TUE / ITUE
• OCP Sustainability Metrics Considerations - PUE

Read the document

Publication details

Authors

Chen, Y.; Quigley, P.; Roberts, J. (Airsys)

Publisher

Open Compute Project Foundation

Year

2026

Document language

English

Source type

Industry body

Access

Open licence — downloadable

Licence

CC BY-SA 4.0 — share with attribution under the same licence

Licence evidence

www.opencompute.org

Citation