Skip to content

Add provider: VoltageGPU (confidential GPU instances, Intel TDX) #129

Description

@Jabsama

Provider to add: VoltageGPU

Cloud: VoltageGPU
Website: https://voltagegpu.com
Live JSON price feed (no key, no rate limit): https://voltagegpu.com/api/pricing/snapshot

Not to be confused with Voltage Park, which is a different and unrelated company. We are VOLTAGE EI, a French company, SIREN 943 808 824.

Why this might be worth a row

Every instance below runs sealed inside an Intel TDX hardware enclave, with guest memory encrypted from the host. As far as I can tell the CSV has no column for that today, so these rows would just look like ordinary Hopper and Blackwell instances at an above-median price. That is the honest framing: we are not competitive on price per FLOP and I am not pretending otherwise. If confidential compute is out of scope for this table, close this and I will not push back.

Proposed rows

Cloud,GPU Type,GPU Arch,GPUs,GPU RAM,vCPUs,RAM,On-demand,Per-GPU,Spot,Name
VoltageGPU,H100 (80 GB),Hopper,1,80,,,6.95,6.95,,h100-small
VoltageGPU,H100 (80 GB),Hopper,8,640,,,55.62,6.95,,h100-xlarge
VoltageGPU,H200 (141 GB),Hopper,1,141,,,8.08,8.08,,h200-small
VoltageGPU,H200 (141 GB),Hopper,8,1128,,,64.62,8.08,,h200-xlarge
VoltageGPU,B200 (180 GB),Blackwell,8,1440,,,96.30,12.04,,b200-xlarge
VoltageGPU,RTX 6000B (48 GB),Blackwell,1,48,,,3.80,3.80,,rtx6000b-small

All prices are USD per hour, on demand, no spot tier. There is no minimum commitment and no reservation.

Two things I should flag rather than have you discover them

vCPUs and RAM are blank. I did not want to invent them. Our public API does not expose per-instance CPU and host RAM, and I would rather leave two cells empty than fill them with a guess that ends up in a table people quote. If you need them filled before merging, say so and I will get the exact figures from the machine definitions and send them.

B200 GPU RAM is 180 GB, not 192 GB. This is a common error, including in some vendor material. An NVIDIA DGX B200 carries 1,440 GB of HBM3e across eight GPUs, which is 180 GB per GPU. I mention it because if the number ever lands in this CSV as 192, it will get copied onward.

On the rest of our catalogue

We also sell a standard tier with no enclave, cheaper, but its prices move during the day against live availability. Putting those in a static CSV would make the row wrong within hours, so I am deliberately not proposing them. The JSON feed above always carries the current figures and an updatedAt timestamp if you ever automate this.

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions