Independent compute market data

A pricing tape for the AI compute market.

Compute Index turns fragmented GPU rental supply into comparable market data: live inventory, provider-level pricing, H100-focused index signals, Blackwell coverage, and transparent source quality.

Market focus

Cloud GPU spot pricing

Primary benchmarkH100 spot market
Emerging coverageBlackwell Watch
MethodologyProvider-weighted medians
Quality controlsFreshness and source confidence

Comparable pricing

Normalize fragmented GPU offers into provider-level market signals, so buyers can compare compute on the same economic basis.

Source quality first

Separate live API inventory from reference prices, stale rows, hardcoded fallbacks, and data that should not enter the primary index.

Market structure, not screenshots

Track price dispersion, inventory depth, freshness, and provider concentration instead of treating one low listing as the whole market.

Why it exists

Compute is becoming a traded input. The data layer is still fragmented.

GPU pricing is spread across marketplaces, cloud providers, availability pages, private quotes, and changing machine configurations. Buyers need a cleaner view of what is live, what is comparable, and where pricing pressure is moving.

Compute Index is built to make that market legible. The goal is not to claim every private enterprise contract is visible. The goal is to publish a transparent public signal with clear inclusion rules, source health, and methodology.

Who uses it

Built for teams that need compute decisions to be grounded in data.

AI labs and startups planning training or inference capacity
Infrastructure teams benchmarking cloud GPU procurement
Investors and analysts tracking AI compute market structure
GPU cloud providers monitoring public price competition

Independent, transparent, and still improving.

Compute Index is operated as an independent market-data project by AI infrastructure builders. We publish methodology and quality pages because the credibility of an index depends on what it excludes as much as what it includes.

If you are a provider, buyer, researcher, or investor and see a data issue, missing source, or better normalization method, we want to hear from you.

Contact Compute Index

For provider data corrections, methodology feedback, partnerships, or market-data questions.

ai.compute.index@gmail.com