Earn from your idle GPUs
Serve SkyPilot Tokens on the idle GPUs on your fleet. Your own jobs always take priority.
Opt in and start earning with one click · Pause or leave any time
SkyPilot is trusted by AI teams building the frontier
“SkyPilot provides a flexible orchestration layer that adapts to Nubank's infrastructure and operational requirements, rather than constraining us to a rigid platform.”Abhishek Shivanna, Machine Learning Senior Manager, NubankRead testimonial“How Meta FAIR uses SkyPilot to manage AI workloads across their 100k+ GPU Slurm fleet.”Watch tech talk“We've been using SkyPilot since the early days of the company, for over 2 years. Great software + team, highly recommend :)”Erik Spock Gafni, Co-founder and CTO, TypeSafeRead testimonial“Abridge moved from SLURM to a scalable multi-cloud AI infra with SkyPilot, unlocking 10x faster development cycles.”Read story“We have been using SkyPilot at Unconventional AI to scale very quickly. So much pain removed, quick and easy ways to launch large jobs quickly.”Xavier Gonzalez, AI Researcher, Unconventional AIRead testimonial“SkyPilot gave us all of those features and more out of the box — gang scheduling, multi-node jobs, SSH access — in a unified interface across all our clouds.”Linden Li, Co-founder, Applied ComputeRead testimonial“Hundreds of researchers sharing thousands of Blackwell GPUs across hyperscalers and neoclouds was a full-time platform team's problem. With SkyPilot Platform, it runs itself — quotas, policies, and a dashboard instead of a queue of tickets.”Rahul A R, Principal Researcher, SarvamRead testimonial“H Company adopted SkyPilot as its standard AI infrastructure layer, scaling online RL to 2,000+ GPUs on Kubernetes — previously impossible on Slurm.”Read story“Moving from SLURM to SkyPilot was a strong win for us. A unified next-gen platform to manage all our clusters means we can scale GPUs exactly when we need them without lock-in.”Debajyoti Datta, Co-founder, Hippocratic AIRead testimonialHow it works
Token buyers save. GPU owners earn.
- Earn from the idle GPUs in your fleet
- Get paid while GPUs sit idle; they come back automatically when new workloads arrive.
- Opt in, nothing to operate
- SkyPilot Platform already schedules your fleet, so there is nothing new to run.
- Your jobs take priority
- Token endpoints are preempted within 30 seconds when your workloads need the GPUs, and come back when they go idle.
Calculate the token revenue from your idle GPUs
Kimi-K3 on 8×B300 nodes| How revenue grows | Year 1 | Year 2 | Year 3 |
|---|---|---|---|
| Token revenue | $15M | $19M | $25M |
Show the math
- Nodes512 GPUs ÷ 8 per node64
- Idle node-hours a month64 × 730 h × 30% idle14,016
- Serving node-hours14,016 × 80% serving11,213
- Revenue per node-hour32 requests at once, measured$110
- Token revenue a month11,213 × revenue per node-hour$1.2M
- Token revenue a yeartoken revenue a month × 12$15M
- Our prices
- Kimi-K3: input $1.80, cached input $0.18, output $9.00 per M tokens.
- Serving load
- 32 requests at once on one 8-GPU node: 79,778 input and 565 output tokens per second, 91% of input from cache, 42.5 ms per output token. Earns $110 per node-hour. This is the measured point with the highest throughput that keeps latency under 50 ms per output token and under 2 seconds to the first token.
- Measured by
- SemiAnalysis InferenceX's agentic-trace benchmark of Kimi-K3 on one 8×B300 node, read 2026-10-28. About the AgentX traces
- Traffic
- Real multi-turn coding-agent sessions from InferenceX's AgentX traces, about 125,400 input and 890 output tokens per request with 91% of input served from cache, so output is under 1% of tokens. Shorter contexts or longer answers earn less per node-hour at our prices, because an output token takes far longer to produce than a cached input token.
Secure and isolated on your cluster
Your compute and data stay private. Nothing on the public side gets write access to your cluster.
- Never in the write path
- Your own SkyPilot Platform deploys it. Tokens only sends signed, read-only requests.
- Your data stays untouched
- Unprivileged pods in a restricted namespace, with no access to your data or Kubernetes API.
- Anonymous to buyers
- Buyers can’t identify you. Responses are rebuilt, so no backend details reach them.
- Nothing stored, fully auditable
- No prompts on disk or in logs. SOC 2 Type II; report on our Trust Center.
FAQ
What does joining as a GPU owner involve?
You opt in through the SkyPilot Platform you already run, and you can pause or leave SkyPilot Tokens at any time. You earn on every token served from your fleet: use the calculator to estimate the token revenue.
New to SkyPilot? Sign up for SkyPilot Platform and become a verified provider.
What happens when I need my GPUs back?
Your workloads take priority and transparently preempt any Tokens jobs. They are scheduled as usual.
What actually runs on my GPUs?
A model server that is (1) deployed through the same SkyPilot Platform control plane that already manages your fleet; (2) using a limited set of hardened container images maintained by SkyPilot.
The model server runs unprivileged in a restricted namespace: no service-account tokens, egress limited to the SkyPilot Tokens proxy and the cloud storage model weights are pulled from, no access to your data volumes, and images and weights pulled only from SkyPilot-managed sources. It writes nothing to your infrastructure. Exact manifest is available for audit.
How can SkyPilot Tokens be 40% cheaper?
SkyPilot Tokens is built on a novel approach: serving tokens on SkyPilot-managed GPUs from verified customers. These customers already use SkyPilot Platform to schedule their GPU fleets, and they can opt in to allow SkyPilot Tokens serving on idle nodes.
Serving on idle GPUs (with dedicated fallback capacity) means we can offer 40% cheaper cost than most inference providers.