AI builders
Find GPU capacity
TinySet compares GPU clouds on H100 price, regions and commitment. TinyResearch finds the best-value reliable capacity.
TinySet
Dataset · 50 rowsGPU cloud providers, with H100 hourly price, regions and minimum commitment
| Provider | H100 hourly price | Regions | Minimum commitment |
|---|---|---|---|
| Azure | 6.98 | East US, East US 2, West US 2 | none |
| AWS | 6.88 | us-east-1, us-east-2, us-west-2, ca-central-1, eu-west-1, eu-central-1, eu-north-1, ap-northeast-1, ap-northeast-2, ap-south-1, ap-southeast-1, ap-southeast-2 | none |
| Oracle Cloud | 10 | global | none |
| VERDA | 4.79 | FIN-01, FIN-02 | none |
| ThunderCompute | 3.2 | quebec | none |
| Spheron | 2.64 | North America, Europe, Asia | none |
| RunPod | 1.99 | global | hourly |
| CoreWeave | 6.16 | North America, Europe | none |
TinyResearch
ReportWhere is the best-value reliable H100 capacity for a startup right now?
H100 GPU compute is widely available but highly stratified by price and reliability. On-demand rates span from roughly $1.80/GPU-hour (small niche providers) to $12.29/GPU-hour (Azure hyperscaler) — a 7× range for nominally the same chip [3][11][26]. The median across 47 providers with published on-demand pricing sits at $3.75/GPU-hour [11]. For most startups, the best combination of price, reliability, and operational simplicity in November 2026 falls into one of three tiers:
