OpenRelay · Bring your own capacity
We run your GPU fleet for nothing, and pay you for the hours it sits idle.
What it costs
Nothing. No platform fee, no per-node charge, no support tier.
How we make money
We serve inference on your idle hours and take a share. You take the rest.
Where it runs
On-prem, colo, or reserved instances in your own cloud account.
Network changes
Nothing inbound. Nodes dial out over QUIC, so egress needs HTTPS and UDP 443.
What your team gets
Idle capacity
nothing to configure
There is no switch to flip and no policy to write. A chip nobody on your side is using picks up inference, and the moment your team wants it back it is theirs again, near-immediately.
What we operate
Included · $0
- Security patching, with the kernel held so apt cannot reboot a node under a running job
- GPU health checks, with unhealthy cards delisted before they can fail a job
- Node agent updates on a jittered timer, so the fleet never restarts at once
- VFIO binding and IOMMU config, boot-persistent across reboots
- Base image and VM template distribution, content-addressed and checksum-verified
- Scheduling, routing, and metering per user and per project
- The reconciliation that cleans up after all of it
What you see
Nodes onboarded
24
3 sites
Your utilization
69%
trailing 7 days
Idle hours served
1,284
this month
| Node | GPUs | Running | Utilization |
|---|---|---|---|
| gpu-ord-01 | 8x H100 | vision-lab / train-7bKubernetes | 98% |
| gpu-ord-02 | 8x H100 | vision-lab / train-7bKubernetes | 97% |
| gpu-ord-03 | 8x A100 | nlp / eval-sweepPod | 62% |
| gpu-ord-04 | 4x L40S | platform / ci-runnersVM | 31% |
| gpu-sfo-01 | 8x H100 | OpenRelay inferenceEarning | 88% |
| gpu-sfo-02 | 8x H100 | OpenRelay inferenceEarning | 91% |
Onboarding
- 01
Run one command on a clean Ubuntu install.
- 02
The node dials out and reports its GPUs and topology.
- 03
Your teams schedule against it.
- 04
Idle hours start earning, with nothing to configure.
Requirements: x86_64 with IOMMU, 32 GB RAM or more, NVMe, a clean Ubuntu install, outbound HTTPS and UDP 443.
Tell us the GPU models, the node count, and roughly how busy they are, and we will come back with what the fleet earns.