Research methodology

Where every number comes from.

Authoritative public sources, a clearly labeled archival directory snapshot, and configured feeds — kept distinct so current research does not inherit stale reference data silently.

Source detectedProbability movedDesk reviewedBrief ready
Current status: the source families below describe inputs the research workflow can use. Compute Index code and Data API read wrappers exist, but customer-serving benchmark publication remains gated on persisted, methodology-cleared records and deployed freshness proof. This page is not proof that every source is connected or that a live benchmark is operating.

Representative Source Families

These categories are research inputs, not a claim of continuous connection, complete coverage, or current licensing for every named source.

Government & Regulatory

  • • Department of Energy (DOE): National energy consumption data, data center electricity projections
  • • Lawrence Berkeley National Laboratory (LBNL): Data center energy research and forecasts
  • • FERC: Federal energy regulations, Order 2023 interconnection rules
  • • State PUCs: Utility tariffs, rate cases, data center-specific regulations
  • • EIA: Energy Information Administration statistics and pricing data

Grid Operators (ISOs/RTOs)

  • • PJM: Interconnection queue data, capacity auction results, congestion pricing
  • • ERCOT: Real-time grid conditions, resource adequacy reports
  • • CAISO: Renewable integration data, transmission constraints
  • • MISO, SPP, NYISO, ISO-NE: Regional grid data and queue status

Industry & Financial

  • • Corporate filings: 10-K/10-Q reports from public data center operators
  • • Earnings calls: Management commentary on energy costs and expansion plans
  • • Licensed transaction databases: Infrastructure-deal data when configured for an engagement
  • • Permit filings: Construction permits and environmental assessments

The GreenCIO Compute Index

The repository defines the GreenCIO Compute Indexas a gated GPU-hour benchmark framework for spot and forward compute-price records across SKU × region × tenor. Implemented records use named series identifiers, such as GCI-H100-USE-SPOT. Customer-facing values must come from fresh, persisted, non-provisional, methodology-cleared rows.

Candidate settlement method

  • • Eligible inputs: source-traced venue rates and contributed quotes only after provenance and usage rights are established.
  • • Filter & weight: the candidate rulebook defines outlier, freshness, and source-weight handling; a methodology version travels with each cleared record.
  • • Settle & band: implemented records can retain a value, uncertainty band, source counts, and accepted/rejected input lineage.
  • • Basis spreads: implemented read routes can expose persisted region, SKU, and tenor comparisons after the same publication gates pass.

Publication and governance gates

  • • Versioned records: the schema supports methodology versions, rulebook hashes, revisions, and input lineage; deployed completeness and replay still require proof.
  • • Human publication: production candidates require explicit approval before they can become customer-serving records.
  • • External assurance: GreenCIO does not currently claim IOSCO compliance, independent oversight, or an operating public benchmark.
  • • Business boundary: GreenCIO does not operate an exchange, clearing house, or brokerage. That posture is not a substitute for an independent governance program.

Internal Compute Desk and external Data API read paths exist. Customer-serving access remains gated until the linked schema, source rights, persisted data, scheduler, freshness, and deployment smoke checks pass. The MCP server remains a preview/design-partner surface.

Update and Freshness Boundaries

The frequencies below are workflow targets. Repository jobs are not evidence that a live scheduler is enabled, and each output remains subject to source-specific freshness, provenance, and publication checks.

Configured Feed Cadence

  • • News and press releases
  • • M&A announcements
  • • Regulatory filing alerts
  • • GCS-backed feed updates when storage is connected

Source-Dependent Updates

  • • Interconnection queue changes
  • • Tariff modifications
  • • Permit applications
  • • Weather and climate data where connected

Potential Periodic Analysis

  • • Queue progression analytics
  • • Regional trend reports
  • • Policy change summaries
  • • Market sentiment indicators

Analyst Review Targets

  • • Comprehensive market reports
  • • Regulatory landscape updates
  • • Infrastructure development tracking
  • • ESG metrics compilation

Illustrative Analytical Framework

Risk Scoring Inputs

A configured engagement can combine factors such as:

  • • Grid risk: Queue position, interconnection delays, and capacity constraints
  • • Regulatory risk: Current tariffs, proposed changes, and jurisdiction context
  • • Energy-cost risk: Price volatility, renewable availability, and demand charges
  • • Environmental risk: Water stress, extreme-weather exposure, and carbon intensity

Weighting and thresholds are engagement-specific unless a reviewed methodology version explicitly says otherwise.

Scenario Outputs

Models and rules can frame hypotheses for analyst review, including:

  • • Interconnection approval timelines
  • • Energy price trajectories
  • • Regulatory change probability
  • • Infrastructure build-out patterns

These outputs are not described as calibrated forecasts until a current, outcome-linked validation record proves that claim.

Required Quality Controls

Before Customer Use

  • Source identity and usage rights recorded for supported inputs
  • Freshness and persistence state checked explicitly
  • Outliers and rejected inputs retained for reviewer inspection
  • Human approval for customer-serving Compute Index publication

Output Boundaries

  • Source attribution where the underlying record supports it
  • Confidence or uncertainty shown only when the method supplies it
  • Material assumptions available for analyst review
  • Methodology and data-state changes distinguished from live proof

Key Differentiators

1. Evidence-First Research: Outputs should preserve source attribution, confidence, and assumptions so analysts can verify before acting.

2. Unified View: Grid, regulatory, financial, and environmental sources can be reviewed in a single risk framework.

3. Forward-Looking Scenarios: Historical and current inputs can support scenario comparisons; they are not verified outcome forecasts by default.

4. Actionable Briefs: Raw data is transformed into research briefs tied to your portfolio and investment strategy where client data is connected.