Computable GPU Index Makes Compute Prices Verifiable
Computable GPU Index (CGI) publishes transparent USD-per-GPU-hour prices for H100, H200, B200, and B300 compute based on published provider rates. Its open-source repository, REST API, flat-file history, and MCP server make each index print independently reproducible.
CGI turns opaque GPU-cloud pricing into usable market infrastructure, though its fixed provider panel means coverage and methodology choices still shape the signal.
- –Robust aggregation and liveness weighting aim to reduce outliers, stale listings, and provider concentration
- –The API and MCP server make current prices accessible to dashboards, procurement tools, and AI agents
- –Reproducible historical records could help teams forecast training and inference costs instead of relying on ad hoc quotes
- –Coverage currently focuses on a limited set of NVIDIA accelerators, leaving AMD and broader regional pricing gaps
- –Apache-licensed code improves adoption, while the published data’s CC BY-NC license may constrain commercial reuse
DISCOVERED
2h ago
2026-09-01
PUBLISHED
7h ago
2026-09-01
RELEVANCE
AUTHOR
[REDACTED]
