Abstract
The Compute Heat Rate (CHR) is a metric that quantifies the maximum electricity price a data center operator can rationally pay before the computation running on that electricity becomes uneconomic. It functions as the demand-side analogue to the traditional gas heat rate, which has served as the primary supply-side price-setting mechanism in wholesale electricity markets for decades. While the gas heat rate converts fuel prices into electricity prices via thermal efficiency, the CHR derives the electricity price ceiling from the economic value of the compute workload consuming that electricity. This paper introduces the CHR framework, presents the formal methodology, reports initial calculations across AI workload types, and discusses implications for wholesale price formation in regions experiencing significant AI-driven demand growth.
1. Motivation
Wholesale electricity markets in the United States have historically been governed by a well-understood price formation mechanism: the marginal generator, typically a natural gas combined-cycle plant, sets the clearing price based on its fuel cost and thermal efficiency (the gas heat rate). At a gas price of $3.50/MMBtu and a heat rate of 7,000 BTU/kWh, the implied marginal cost is approximately $24.50/MWh. This mechanism has kept wholesale prices largely within the $20-$60/MWh range across most U.S. markets for the past two decades.
A critical, underexamined feature of this equilibrium is the role of demand-side price sensitivity. Traditional large electricity consumers, including aluminum smelters, steel producers, chemical manufacturers, and other heavy industry, become uneconomic at relatively low price thresholds ($40-$120/MWh depending on the industry). When wholesale prices rise, these consumers curtail, reducing demand and restoring price equilibrium. This demand-side brake has functioned as a self-correcting mechanism for decades.
The rapid growth of AI-driven data center electricity demand introduces a new category of consumption with fundamentally different price elasticity characteristics. The economic value generated per megawatt-hour of AI compute is orders of magnitude higher than any prior industrial electricity use case. This paper introduces the Compute Heat Rate as a formal metric for quantifying this difference and analyzing its implications for wholesale price formation.
2. Definition
The Compute Heat Rate is defined as the maximum electricity price at which a given AI compute workload remains profitable for the operator, after accounting for all non-electricity costs and a required return margin.
- CHRw
- Compute Heat Rate for workload type w, expressed in $/MWh. Represents the maximum electricity price at which the workload remains economic.
- Rw
- Gross revenue per MWh of electricity consumed by workload type w. Derived from API pricing, cloud compute rates, or enterprise contract values, converted to revenue per MWh using GPU power consumption and throughput data.
- Cnon-elec
- Non-electricity operating costs per MWh, including GPU amortization, facility overhead, cooling, networking, and maintenance. Estimated from published infrastructure cost data and industry benchmarks.
- m
- Required return margin (expressed as a decimal). Represents the minimum profit margin the operator requires. Baseline assumption: 0.30 (30%).
The CHR is conceptually analogous to the gas heat rate but operates on the demand side of the market rather than the supply side. Where the gas heat rate converts a fuel input price into a generation cost (setting a floor for wholesale prices), the CHR converts a compute revenue output into a maximum tolerable electricity input cost (setting a ceiling on the price at which AI demand curtails).
3. Input Data Requirements
The CHR is calculated from publicly available data sources. No proprietary or confidential data is required.
| Input | Source | Role in Calculation |
|---|---|---|
| GPU power consumption | NVIDIA published specifications; MLPerf benchmarks | Converts compute throughput to electricity consumption per unit of output |
| Power Usage Effectiveness (PUE) | Industry benchmarks (Uptime Institute, DOE) | Scales IT load to total facility electricity consumption |
| API / inference pricing | OpenAI, Anthropic, Google published pricing | Establishes revenue per token/request; converted to Rw per MWh |
| Cloud compute rental rates | AWS, GCP, Azure published pricing | Alternative revenue basis for non-API workloads |
| GPU throughput | NVIDIA documentation; independent benchmarks | Links power consumption to revenue-generating output |
| Infrastructure costs | Hyperscaler SEC filings; industry reports (JLL, CBRE) | Basis for Cnon-elec estimation |
| Workload mix estimates | Industry analyst reports; hyperscaler earnings guidance | Weighting for blended CHR calculation |
Table 1. Input data sources for CHR calculation. All sources are publicly available.
4. Initial Results (2026)
The following table presents CHR calculations across major AI workload categories using current (Q1 2026) pricing and infrastructure cost data. The gas heat rate benchmark of approximately $50/MWh is provided for comparison.
| Workload Type | Rw ($/MWh) | Cnon-elec | CHR ($/MWh) | Multiple vs. Gas HR |
|---|---|---|---|---|
| Frontier Inference | $49,300 | $4,250 | $34,650 | ~690x |
| Enterprise Contracted | $9,800 | $4,250 | $4,270 | ~85x |
| Enterprise Agentic AI | $15,000 | $4,250 | $8,270 | ~165x |
| Commodity Inference | $1,850 | $4,250 | ~$800* | ~16x |
| Frontier Model Training | $2,000† | $4,250 | ~$500† | ~10x |
| Blended Average (2026) | $8,500 | $4,250 | $3,270 | ~65x |
Table 2. Computed CHR by AI workload type, Q1 2026. Assumes 30% required margin. *Commodity inference approaches break-even at current non-electricity costs; operators cross-subsidize with higher-margin workloads. †Training revenue amortized over model lifetime; reflects revealed willingness-to-pay. Gas heat rate benchmark: ~$50/MWh.
The blended CHR of approximately $3,270/MWh implies that AI data center demand will not curtail at electricity prices below roughly 65 times the current wholesale average. Even the lowest-margin workloads exhibit tolerance ceilings 10-16 times above the gas heat rate. This asymmetry is the central finding: the demand-side brake that has historically governed wholesale price formation does not apply to AI compute workloads at any price level relevant to current market dynamics.
5. Applications & Use Cases
The CHR framework supports several analytical applications independent of any specific wholesale price projection:
5.1 Demand-Side Elasticity Measurement
The CHR provides a quantitative basis for comparing the price elasticity of different demand classes on the same grid. By computing the CHR for AI workloads and comparing it to the curtailment thresholds of traditional industrial consumers, analysts can identify the "elasticity gap" that determines how price formation changes as AI demand grows as a share of total load.
5.2 Structural Demand Classification
The CHR enables classification of electricity demand into structural categories based on price tolerance rather than sector or SIC code. A demand class with a CHR of $3,270/MWh behaves fundamentally differently in a wholesale market than a demand class with a curtailment threshold of $80/MWh. This classification has implications for capacity planning, transmission investment, and regulatory design.
5.3 Technology Transition Tracking
As AI hardware improves in efficiency and workload economics evolve, the CHR provides a time-series metric for tracking whether the net effect is demand growth (via Jevons paradox dynamics) or demand reduction. Declining per-token costs may coincide with rising or stable CHR values if frontier workloads command premium pricing and the most valuable applications concentrate at the capability frontier.
6. Analogues in Energy Economics
| Metric | Function | Relationship to CHR |
|---|---|---|
| Gas Heat Rate | Converts fuel price to generation cost; sets marginal wholesale price | Supply-side counterpart. CHR is the demand-side analogue. |
| CONE (Cost of New Entry) | Long-run equilibrium price for new generation capacity | Sets the price floor in supply-constrained markets. CHR sets the ceiling. |
| LCOE | Levelized cost of electricity from a generation source | Supply-side metric. CHR measures demand-side willingness-to-pay. |
| VIX | Implied volatility derived from options pricing | Derived metric (not directly observed) that became an industry standard and tradable index. |
Table 3. Positioning of CHR among established energy and financial metrics.
7. Research Agenda
The CHR framework introduced here represents an initial formulation. Several extensions are under active development:
Hub-Level CHR Calculation
Computing location-specific CHR values at individual settlement points by incorporating regional workload mix data, local infrastructure cost differentials, and geographic demand concentration patterns.
CHR Time Series & Index Construction
Developing a methodology for periodic (quarterly or monthly) CHR publication to enable tracking of the metric over time. This includes defining index weighting methodology, data update procedures, and publication standards suitable for use as a market reference.
CHR Spread Analysis
Examining the difference between the CHR tolerance ceiling and actual wholesale prices at specific settlement points as a measure of latent repricing potential, with applications to forward curve analysis and risk assessment.
International Application
Extending the CHR framework to European and Asian electricity markets, accounting for different market designs, regulatory structures, and data center demand patterns.
Related Publications
- Royal, H. (2026). "The Compute Heat Rate: A Preview of What's Coming for Electricity Markets." LinkedIn, February 24, 2026. Link
Additional publications forthcoming. This page will be updated as new research is released.
Suggested Citation
Working paper. Available at: https://computeheatrate.com
Data Sources
All calculations in this framework are derived from publicly available data sources, including: NVIDIA published GPU specifications and documentation; OpenAI, Anthropic, and Google published API and model pricing; AWS, GCP, and Azure published cloud compute pricing; hyperscaler capital expenditure guidance from SEC filings; U.S. Energy Information Administration (EIA) electricity market data; PJM Interconnection and ERCOT published market data; Lazard Levelized Cost of Energy reports; LevelTen Energy PPA Price Index; CBRE, JLL, and Cushman & Wakefield data center market reports; and Uptime Institute and DOE data center efficiency benchmarks.