Clusters and reserved terms
For jobs bigger than one node or longer than a few weeks. Published list prices, plus live reserved prices from providers that publish them in their APIs.
Clusters and reserved terms
Published list prices from provider pricing pages, checked 2026-11-10. These are not live stock. All cluster and reserved offers
| Provider | GPU | Type | Term | GPUs | Per GPU-hour | Notes | |
|---|---|---|---|---|---|---|---|
| HyperstackThrough sales | H100 PCIe | Reserved | Not stated | — | $1.82 | "Starting from" reservation price. On-demand is $2.60. | Provider page |
| HyperstackThrough sales | H100 SXM | Reserved | Not stated | — | $2.72 | "Starting from" reservation price. On-demand is $3.20. | Provider page |
| Together AIThrough sales | H100 SXM | Reserved | 91 to 180 days | — | $3.19 | InfiniBand. Paid upfront. Longer terms are quote only. | Provider page |
| DigitalOceanThrough sales | H100 SXM | Reserved | 12 months | — | $3.26 | 12-month commitment through sales. On-demand is $4.41. | Provider page |
| Together AIThrough sales | H100 SXM | Reserved | 31 to 90 days | — | $3.45 | InfiniBand. Paid upfront. | Provider page |
| Together AIThrough sales | H100 SXM | Reserved | 3 to 30 days | — | $3.69 | InfiniBand. Paid upfront. | Provider page |
| Together AISelf-serve | H100 SXM | Cluster | No commitment | 8–256 | $3.99 | InfiniBand. On-demand HGX H100 clusters up to 256 GPUs. | Provider page |
| RunPodSelf-serve | H100 SXM | Cluster | No commitment | Up to 64 | $4.79 | Instant Clusters up to 64 GPUs. | Provider page |
| LambdaThrough sales | H100 SXM | Cluster | 2 weeks to 1 year | 256+ | $5.54 | 1-Click Clusters, 256+ GPU tier. Plus tax. The page says new prices take effect November 16, 2026. Booked through the sales team. | Provider page |
| LambdaThrough sales | H100 SXM | Cluster | 2 weeks to 1 year | 64+ | $5.85 | 1-Click Clusters, 64 GPU tier. Plus tax. The page says new prices take effect November 16, 2026. Booked through the sales team. | Provider page |
| CoreWeaveThrough sales | H100 SXM | Cluster | On-demand | 8+ | $6.16 | $49.24 per 8-GPU instance-hour, divided by 8. | Provider page |
| LambdaThrough sales | H100 SXM | Cluster | 2 weeks to 1 year | 16+ | $6.16 | 1-Click Clusters, 16 GPU tier. Plus tax. The page says new prices take effect November 16, 2026. Booked through the sales team. | Provider page |
| CoreWeaveThrough sales | H100 SXM | Reserved | Committed usage | — | Quote only | Quote only. The page says committed usage gets up to 60% off on-demand. | Provider page |
| CrusoeThrough sales | H100 SXM | Reserved | Not stated | — | Quote only | Quote only, all GPUs. | Provider page |
| Massed ComputeThrough sales | H100 SXM | Cluster | Not stated | — | Quote only | InfiniBand. Quote only. | Provider page |
| Voltage ParkSelf-serve | H100 SXM | Cluster | No minimum term | 8–1,016 | Quote only | InfiniBand 3200 Gbps. Quote only. | Provider page |
| RunPodThrough sales | H100 SXM | Reserved | 1 to 12+ months | — | Quote only | Reserved Clusters are quote only for B200, H200, H100, A100, B300 and L40S. | Provider page |
| HyperbolicThrough sales | H100 SXM | Reserved | Under 1 year, prepaid | Up to 1,000 | Quote only | InfiniBand or Ethernet. Quote only. | Provider page |
| NebiusThrough sales | H100 SXM | Reserved | Multiple months | — | Quote only | Quote only for H100, H200, B200 and B300 clusters. | Provider page |
| Denvr DataworksThrough sales | H100 SXM | Reserved | Multi-month or annual | Up to 512 | Quote only | Quote only. The page says up to 20% off with commitments. | Provider page |