Every number,
published.
Most providers at our stage say "contact us." Here is the whole rate card instead — per GPU-hour, in USD, billed per second. Reserve and it is fixed for your term.
Indicative launch pricing, exclusive of tax, for clusters currently in build. Not a quotation and not an offer capable of acceptance — your rate is fixed in a signed reservation.
USD / GPU-hour · billed per second
| Accelerator | VRAM | Interconnect | vCPU / RAM | Availability | Rate |
|---|
Full machine · bare metal
Whole nodes, NVLink intact.
Eight SXM accelerators, no hypervisor. Rates follow the same commitment ladder above.
| Node | vCPU | System RAM | Interconnect | Availability | Rate |
|---|
Ingress, egress and inter-node transfer. On every plan, including on-demand.
NVMe-backed persistent volumes. First 500 GB per node included.
WEKA or Lustre tier, up to 1.2 TB/s aggregate read.
Cost model
Do the math
before the call.
Set your fleet and utilisation to see the monthly bill. Enter what you pay today and it will show the delta — we compare against your number, not a figure we picked.
Your blended effective rate, including any committed-use discount. Leave blank to skip the comparison.
- Compute
- $0
- Parallel storage
- $0
- Egress & transfer
- $0.00
- Setup & onboarding
- $0.00
Estimate only, based on indicative rates and a 730-hour month. Not a quotation.
In the price
Nothing below is an add-on.
Ready when you are.
Send us the workload shape. Firm quote back within one business day.