CCIR — Compute Credit Index Research

CRI-H100

CRI-H100 is a weekly index measuring the trailing 7-day median $/GPU-hour for NVIDIA H100 SXM on-demand rental listings on the US marketplace. It is designed for reference citation in credit agreements, loan covenants, and collateral valuation frameworks.

The index is calculated from publicly available data, published under an open methodology, and independently verifiable by any third party. Governance is designed with reference to the IOSCO Principles for Financial Benchmarks.

CRI-H100 — Current Status

Burn-in Period

Data collection commenced February 2026

The index is in its initial data collection period. Daily snapshots are being archived across 9 GPU models. The first published index value will be announced here when the minimum observation threshold is met.

Methodology

Data source	Vast.ai public API — no authentication required
Target GPU	NVIDIA H100 SXM
Geography	United States
Quality filters	Reliability ≥ 0.90, active within 7 days, unrented
Outlier removal	Trimmed mean ± 2.5σ
Calculation	Trailing 7-day median $/GPU-hour
Publication	Weekly — Thursdays
Reproducibility	Fully open-source — any party can independently verify any published value

Documents

CCIR Methodology

Full technical specification for CRI-H100 index calculation

v1.1.0
Governance Framework

Administrator framework, change control, fallback waterfall, IOSCO alignment

v1.0
Data Repository

Daily snapshots, calculation pipeline, verification script, audit trail

GitHub

Multi-Model Data Archive

CCIR collects daily snapshots across multiple GPU generations to support cross-generational depreciation curve construction. The Vast.ai API provides real-time snapshots only — historical data cannot be collected retroactively.

H100 SXM	Primary index — CRI-H100
A100 SXM / PCIe	Depreciation curve — A100→H100 transition
H200 / H200 NVL	Next-generation transition tracking
V100	Long-horizon depreciation (2+ generations)
L40S	Inference-class GPU market
RTX 4090	Consumer GPU market context

Use Cases

CRI-H100 is designed for credit market applications where an independent, auditable rental rate reference is required:

Covenant triggers	Mechanical, non-discretionary triggers linked to an external market reference
Collateral valuation	Standardized methodology for marking GPU collateral to market
Stress testing	Floor reference for scenarios where borrowers lose enterprise contracts
ABS structuring	Performance monitoring, early amortization triggers, overcollateralization tests
Rating agency input	Independent data source meeting post-LIBOR governance standards

Contact

research@ccir.io

Licensing inquiries, pre-transaction audit coordination, methodology questions, and complaints should be directed to this address.