Pricing
Simple compute pricing.
IndexCube is designed around an all-in hourly commercial model. Final rates depend on architecture, configuration, commitment, location, networking and storage.
| GPU | Service | Availability | Indicative Price | |
|---|---|---|---|---|
| NVIDIA H100 | GPU / Node / Cluster | ENQUIRE | Request rate | Get H100 Pricing |
| NVIDIA H200 | GPU / Node / Cluster | ENQUIRE | Request rate | Get H200 Pricing |
| NVIDIA B200 | Node / Cluster | CAPACITY REQUEST | POA | Request B200 |
| NVIDIA B300 | Node / Cluster | CAPACITY REQUEST | POA | Request B300 |
| AMD MI325X | GPU / Node / Cluster | ENQUIRE | Request rate | Get MI325X Pricing |
| AMD MI355X | Node / Cluster | CAPACITY REQUEST | POA | Request MI355X |
Pricing can be structured for hourly consumption, reserved capacity or dedicated infrastructure. Object storage, private connectivity and custom cluster architecture can be added to the commercial model.
Reserved
Committed capacity for predictable demand and better long-term economics.
Dedicated Cluster
Private GPU infrastructure with network and storage tailored to your workload.
Large Scale
Talk to IndexCube64, 256 or 1,000+ GPUs?
For large deployments, IndexCube can structure dedicated infrastructure around GPU architecture, power, network, storage, data centre location and contract term.