CT-H100 Index
Daily median sourced H100 on-demand pricing with history and downloadable data.
Charts
Charts for the AI compute market
A chart index for source-backed AI compute price, capacity, and power signals.
Start here when you want the visual surfaces behind the market read: current GPU price bands, the CT-H100 trend, provider capacity, and power constraints. ComputeTape links only charts and data assets with source, date, and methodology context.
4Live assets
H100 Index, Market, GPU Pricing, and Power Watch carry source-backed visual reads.
Source-labeledSource standard
Every public value includes source, timestamp, and methodology context.
Live chart assets
Open the charts behind the market read
These are the current source-backed chart surfaces. They are linked instead of duplicated so each dataset has one canonical home.
GPU Pricing
Sourced H100, H200, and B200 price bands plus per-provider rate tables.
Market Signals
Capacity-led AI compute market page with the H100e market-rate signal and provider context.
Power Watch
Regional power constraints, data-center queue signals, and power deals affecting usable compute capacity.
Chart categories
What each chart lane is for
- Price charts show public GPU-hour list prices and H100-equivalent comparisons; they are not quotes or reserved terms.
- Capacity charts show availability, access friction, and provider signals that affect whether buyers can actually obtain GPUs.
- Power charts show grid, energization, and power-sourcing constraints that can delay compute supply even when chips are available.
- Market-signal charts combine price, capacity, and power context so readers can see whether supply is tightening or loosening.
Planned chart lanes
Chart lanes still gated on data quality
- H100 vs H200 — the transition between current and transition-generation supply.
- H100 vs B200 — current capacity versus the next accelerator cycle.
- Spot vs forward — term structure for near-term scarcity and forward expectations.
- Cloud vs neocloud — hyperscaler and specialist operator capacity signals.
- GPU vs stocks — public-equity context for compute-linked names.
- Power vs compute — infrastructure constraints on usable capacity.
A lane publishes only when the underlying dataset meets ComputeTape source, freshness, and caveat standards.