What is the GPU Share Network?
The GPU Share Network connects verified GPU operators with AI workloads that need compute. Operators contribute idle GPUs and earn 60% of the revenue Plugsky routes to them. Plugsky handles authentication, metering, routing, and Stripe payouts, so operators just plug in and earn.
How it works
- Check eligibility — run the local preflight:
curl -fsSL https://plugsky.com/gpu-worker/preflight.js | node -. - Register a node — add your GPU from the dashboard and run the outbound-only worker.
- Get verified — nodes pass a health and identity check before they receive traffic.
- Earn 60% — every job routed to your GPU pays out 60% of the revenue, with a minimum payout threshold.
Pricing and operator share
Plugsky GPU 27B is $0.30 per 1M input tokens and $2.40 per 1M output tokens. Operators keep 60% of routed revenue; Plugsky keeps a 40% gross platform share before payment processing and infrastructure costs. This is below the $0.50/$3.00 list rate for the full 27B model, but commodity providers can be cheaper. The tradeoff is a verified consumer-GPU pool, 32K context, and no inbound port on the operator machine.
Security and classification
Only non-sensitive workloads are routed to the shared network. Every payload is classified before routing — if a request looks sensitive, it stays on first-party infrastructure instead. This keeps the network open and the data policy honest.
GPU Share Network FAQ
What GPUs can join?
Consumer and prosumer GPUs with at least 16 GB VRAM/unified memory and 24 GB system RAM can run the current model; 24 GB VRAM is recommended for 32K context. The local preflight reports your exact eligibility before registration.
How do payouts work?
Payouts are made via Stripe when your earnings cross the minimum threshold. You can track per-node earnings and payout history in the dashboard.
Is my data shared with node operators?
No. Payloads are classified before routing, and sensitive workloads never reach the shared network.
Do I need the GPU Share Network to use Plugsky?
No. API customers are served from our first-party cloud first; the shared network adds capacity and lowers costs. Your API behavior never changes.
Get started in minutes
OpenAI-compatible API with 39 models, a 7-day money-back guarantee, and a 99.9% uptime SLA. No code changes required.
Start free → Read the docs