Compute lanes
Two lanes. One network.
Match each workload to the right hardware. Sealed compute protects sensitive prompts; open compute serves public-model jobs, embeddings, evaluations and image generation.
Sealed lane
Confidential GPU execution
COMPUTE DESIGN
Design details
TDX confidential environment
Encrypted prompts through OHTTP
Signed COSE job receipts
Eligible GPU attestation
Compatible inference API
Open lane
WORKLOAD FIT
Flexible GPU capacity
COMPUTE DESIGN
signed at the source
Design details
RTX 3090 / 4090 / 5090 class
Public-model batch inference
Embeddings, evaluations and images
Non-sensitive workloads only
Workload-aware scheduler
Shared proof
Evidence across both lanes
COMPUTE DESIGN
signed at the source
Design details
Node-key receipt signatures
Job-to-billing verification
Hourly cumulative Merkle roots
Bonded node identity
Reputation-guided routing
Evidence before accounting
One verified work ledger
Compare the design
Sealed lane
Hardware & workload
GPU class
CC
RTX
Evidence
Metering
GPU-hour
Job / tokens
Evidence
Workload-aware scheduler
Signed receipts
Receipts & execution
Signed COSE job receipts
Job evidence
Compatible inference API
Hourly cumulative Merkle roots
Root cadence
1 hour
1 hour
Evidence
Network integrity
Hardware capability checks
Non-sensitive workloads only
Bonded node identity
Receipts prove execution
One verified work ledger
Evidence before accounting
Illustrative flows
Explore the paths through the network
“Install a node. Serve suitable AI jobs. Follow the signed receipt to a payout route you choose.”
Spindle network
Illustrative flow
2
2
compute lanes in the design
Have questions? Find answers.
What does Spindle do?
Spindle is a GPU compute network design built on Robinhood Chain. Providers register GPU nodes and serve AI jobs in the open lane; sealed is in testing. Each job emits a signed receipt and verified earnings accrue in USDG. Claims pay USDG to your wallet; from there you can swap to Stock Tokens or fund a virtual card. This site presents the v1.1 specification and an illustrative dashboard; it does not display live earnings or guarantee demand.
Which GPUs and APIs are supported?
What does a receipt prove?
Which compute lane is private?
How do the payout routes work?
How does node onboarding work?