Same access mode
Every row here is public on-demand list price. A reserved or spot quote belongs in a different comparison — the tiers price different commitments.
GPU Pricing
H100 Cloud Comparison
Sourced H100 on-demand rates arranged for a buying decision — provider type, region, availability, price dates, and estimate prefills.
This is a guide to reading provider price pages, not a quote engine. The rows are the same rights-vetted observations the directory and the CT-H100 Index use; the columns are the ones that decide whether two H100 offers are actually comparable.
7Providers compared
Sourced on-demand H100 rows that cleared editorial review
$2.40–$12.29 (7 providers)On-demand band
Lowest to highest list price as of Jun 11, 2026
$4.29 / H100-hrCT-H100 median
The single-number market read behind the spread below
Side by side
H100 providers compared
Lowest list price first. Use the Estimate cost link to push a row’s rate into the GPU-Hour Cost Calculator with your own runtime and utilization.
| Provider | Provider type | Region | List $/GPU-hr | H100e $/hr | Availability | No change in | Estimate cost | Source |
|---|---|---|---|---|---|---|---|---|
| Hyperstack | Developer GPU clouds | Multi-region | $2.40 | $2.40 | Available Limited | 18 days | Prefill calculator | Open price page |
| RunPod Secure | Developer GPU clouds | Multi-region | $3.29 | $3.29 | Available Limited | 14 days | Prefill calculator | Open price page |
| Crusoe | AI-native clouds | Multi-region | $3.90 | $3.90 | Available Limited | 14 days | Prefill calculator | Open price page |
| Lambda | Developer GPU clouds | Multi-region | $4.29 | $4.29 | Available | 14 days | Prefill calculator | Open price page |
| CoreWeave | AI-native clouds | us | $6.16 | $6.16 | Available Limited | 26 days | Prefill calculator | Open price page |
| AWS | Hyperscalers | us-east-1 | $12.29 | $12.29 | Quota Gated | 26 days | Prefill calculator | Open price page |
| Azure | Hyperscalers | eastus | $12.29 | $12.29 | Quota Gated | 26 days | Prefill calculator | Open price page |
Public on-demand list prices from provider price pages — not negotiated quotes, reserved terms, or spot rates. Each row carries its own source link and price date. How we source, label, and date evidence: methodology.
Buyer checklist
Make the rows comparable before you compare them
Two H100 rates rarely describe the same product. Before treating one row as cheaper, line the offers up on the four axes that actually move delivered cost.
Same region and deployment
Multi-region rows mean the source does not publish a single-region price. Confirm the region you would actually deploy in before budgeting a row.
Same cluster shape
A single H100 VM and an 8-GPU NVLink node price differently per GPU-hour because they are different products. Check the SKU behind the rate on the source page.
Total cost, not GPU-hour cost
Storage, egress, networking, and support land on the same invoice. A slightly higher GPU rate with free egress can be the cheaper offer.
Next step
From comparison to plan
How to compare GPU cloud quotes
The full lesson on normalizing quotes — terms, interruption risk, and completed-workload cost.
Quote comparison checklist
A printable line-by-line checklist for putting two provider quotes on the same footing.
GPU-Hour Cost Calculator
Estimate monthly cost from any rate above with your own utilization and overhead assumptions.
Provider Directory
Every reviewed provider row across chips, with profiles and machine-readable exports.