Turn idle GPUs
into private revenue.
Run a Nox node and serve encrypted inference the network can never read. Your hardware earns continuously — payouts settle automatically, every hour, in real time.
Bring the silicon. We bring the demand.
Any modern accelerator with 24GB+ of VRAM qualifies. Higher tiers route to larger models and command higher per-hour rates.
Minimum spec
Everything runs inside a hardware-isolated enclave. Meet these and you can enroll today.
- 24GB+ VRAM accelerator
- 100 Mbps symmetric uplink
- CUDA 12 / ROCm capable
- Confidential-compute support
From bare metal to earning in four steps.
No manual ops, no Kubernetes. The daemon handles isolation, attestation, scheduling, and settlement.
Install the daemon
One command pulls the signed Nox runtime and provisions the secure enclave on your GPU.
Register the node
Generate an operator keypair and bind it to your machine. Your identity, never your data.
Stake to qualify
Lock a small bond to enter the routing pool. Stake weights how much traffic you receive.
Serve & earn
Encrypted requests stream in. The enclave computes, settles, and pays out hourly — automatically.
You keep 88% of every request you serve.
Payouts are metered per inference and settled on-chain every hour — no minimum threshold, no payout schedule to wait for. Earnings scale with utilisation and uptime.
What could your hardware earn?
Pick an accelerator and how many hours a day it serves. Estimates use demand-weighted utilisation, net of the 12% network fee. Illustrative, not a guarantee.
Figures assume demand-weighted routing. Real earnings vary with model mix, region, and uptime.
One mesh. Thousands of private endpoints.
The coordinator routes each encrypted request to the lowest-latency node that meets the model's spec — your node is just one secure endpoint in a global mesh.
Plug in your GPU. Start earning by the hour.
Join 2,840 operators powering private inference. Set up in minutes, keep 88%, and never expose a single token of user data.