About OpenRelay
Resilient compute,
for everyone.
OpenRelay is a distributed overlay network for GPU inference. We connect consumer GPUs into a fault-tolerant mesh. Nodes can fail — we don't care. Traffic automatically routes around failures. Anyone can join.
The thesis
Designed for failure.
Centralized clouds treat a failed node as an outage. We treat it as a Tuesday. The network expects nodes to drop — and routes around them before you notice.
→ rerouted 312 req to node-iad-04 in 38ms
Detect
Health checks flag an unreachable node within milliseconds — no waiting on timeouts.
Reroute
In-flight and new requests shift to the nearest healthy GPU automatically.
Heal
The failed node drains and rejoins when it recovers. No manual intervention.
Our values
The principles that guide everything we build.
Fault Tolerance First
We design for failure. Nodes go down, networks partition — your workloads keep running. Resilience is the default, not a feature.
Decentralized by Design
No single datacenter, no single point of failure. Our overlay spans the globe.
Open Participation
Anyone can join the network. Have a 4090 sitting idle? Contribute compute and earn.
Simple Abstractions
Deploy inference workloads with one API call. We handle routing, failover, and load balancing.
Careers
Join the team.
We're a small team building the CDN of inference — distributed systems, GPU infrastructure, and developer experience. If routing around failure at global scale sounds like your kind of problem, we'd love to hear from you.
Let's build the future together
Whether you want to deploy on our network, provide compute, or join our team.