Developing Methodology · First Published April 21, 2026

A Developing Standard for
Compute Energy
Measurement.

CEH, or Compute Energy Hour, is a developing measurement framework designed to help compare compute output, energy consumption, infrastructure efficiency, and related cost inputs across hardware, facilities, and energy environments.

Compute + Energy
Framework Scope
Output, cost, and efficiency context
Hardware Comparison
Benchmark Lens
Cross-system evaluation support
Developing
Publication State
Subject to validation and iteration
Oak Ridge
Institutional Home
Methodology stewardship and publication

Overview

What CEH Is
And What It Is Not

CEH is not a certified public benchmark, not an exchange-traded index, and not investment advice. It is a developing methodology intended to help create a more consistent way to discuss compute-energy efficiency.

The framework is designed to support comparison across different compute systems by linking power consumption and infrastructure conditions to output assumptions in a form that can be reviewed, challenged, and iterated.

CEH may provide a useful shared language for operators, researchers, and capital partners evaluating how energy, utilization, and hardware characteristics interact.

Open Methodology Intent
Designed to support replication, critique, and refinement rather than closed black-box scoring.
Infrastructure Context
Intended to help compare compute performance under different facility and energy assumptions, including grid and behind-the-meter scenarios.
Versioned Development
Published as a developing framework subject to revision as market data, validation work, and deployment feedback evolve.

Formula

Core CEH Expressions

These expressions are designed to support discussion of compute-energy measurement. They should be read as a developing methodological framework, not as a finalized certified standard.

Primary Expression
CEH = kWh consumed / Compute Output Units per hour
Expanded Operational Form
CEH = (TDP x Units x Util% x PUE) / 1,000 / Output Units per Hour
Cost Layer
CEH Cost = CEH x $/kWh
Carbon Layer
CEH Carbon = CEH x kg CO2/kWh

Example BTM assumptions such as $0.035/kWh are methodological benchmark inputs only. They are not guaranteed rates and may vary materially by generation type, geography, project structure, and contract vintage.

Benchmarks

Illustrative Benchmark Table

These rows are presented as benchmark examples within a developing framework. They are intended to support comparative discussion, not to represent certified market-wide rankings.

HardwareCEH-BaseGrid $/M tokBTM $/M tokGrade
NVIDIA H200 SXM5 141GB0.00191$0.153$0.067S
AMD MI300X 192GB0.00214$0.171$0.075S
NVIDIA H100 SXM5 80GB0.00243$0.194$0.085S
NVIDIA H100 PCIe 80GB0.00289$0.231$0.101A
NVIDIA A100 SXM4 80GB0.00412$0.330$0.144A
NVIDIA L40S 48GB0.00661$0.529$0.231B
S · blueA · greenB · emeraldC · amberD · orangeF · red

Market Context

“The demand for computing power is so large that a new asset class will spring up.”

— Larry Fink, Chairman & CEO, BlackRock
Bloomberg Television / Milken Institute

Bloomberg Television. Statement cited as independent market context. CEH is not affiliated with BlackRock or Larry Fink, and this statement does not constitute an endorsement of CEH.

Access

Intended Use Cases

CEH is being developed for readers evaluating compute-energy performance, infrastructure efficiency, benchmarking frameworks, and adjacent underwriting context.

Operators & Infrastructure Teams
CEH is designed to support internal comparison across hardware mixes, facility designs, and changing utilization assumptions.
Capital & Underwriting
CEH may provide a common framework for discussing power cost sensitivity, infrastructure efficiency, and compute output assumptions.
Research & Standards
The methodology is intended to help create a more auditable vocabulary for compute-energy benchmarking and future external validation.