Oracle nodes
Oracle nodes are the top tier. They are designed to power the Autonomous Trading Engine (ATE). The tier is for quantitative analysts, data scientists and institutional operators with deep experience in financial markets and machine learning.
Rewards are not being paid yet, and operators join by application. See Testnet Status.
What you need to know
- Stake: 150,000+ NODR. Staked NODR has a 21-day unstaking period.
- TrustFingerprint: keep a score of at least 0.90.
- Oracle NFT: you must hold one.
- Onchain role: role admins grant the Oracle role. They are the protocol wallet and the Timelock. See Governance. There is no onchain election today.
- Uptime: 99.9%+ required, with 24/7 monitoring and alerting. Respond to critical issues in <1 hour. Apply updates and security patches regularly.
- Time: 20–30 hours per week.
Duties
- Strategy validation. Validate trading strategies with ML/AI models and assess their risk.
- Strategy analysis. Oracles will run machine learning models to find trading strategies. Today one protocol keeper submits strategies.
- Execution. Oracles will execute trades and manage capital for the vaults. No vault capital is deployed to strategies yet.
- Data. Collect and process financial market data.
- Model training. Train and update ML/AI models.
- Emergency controls. Pause and halt authority for protocol safety is planned. Today only the 2-of-3 Safe can pause.
- Governance. Oracle Chamber votes once DAO governance is active.
Hardware
The installer refuses a machine with fewer than 8 CPU cores, 16 GB of RAM or 100 GB of free disk. That is the only hardware check. The table below is the recommended specification and is not enforced.
Oracle nodes are designed around a high end NVIDIA GPU for ML/AI workloads.
| Component | Recommended |
|---|---|
| GPU | RTX 4090 (24GB VRAM), A100 (40GB or 80GB VRAM), H100 (80GB VRAM) or an equivalent professional GPU |
| GPU features | CUDA 11.8+ and Tensor Cores required |
| VRAM | 24GB or more, 40GB+ preferred |
| CPU | 32 cores, 64 threads, 3.0+ GHz base clock. Intel Xeon or AMD EPYC preferred |
| RAM | 128 GB DDR4/DDR5 ECC. ECC is strongly recommended |
| Primary storage | 4 TB NVMe SSD, PCIe 4.0, 5000+ MB/s read, 4000+ MB/s write |
| Backup storage | Extra storage for regular automated backups of ML models and node state |
| Bandwidth | 1 Gbps+ symmetric fiber. Asymmetric connections such as cable are not recommended |
| Latency | <50ms to major network hubs |
| Static IP | Dedicated static IP required |
| Redundancy | Backup internet connection with automatic failover recommended |
| Power supply | Redundant, 1200W+ each, 80+ Platinum or Titanium. GPU and CPU can draw 800–1000W under load |
| UPS | 30+ minutes runtime, with automatic shutdown scripts |
| Cooling | GPU below 80°C and CPU below 75°C under load. Good case airflow, quality fans, possibly liquid cooling |
Software
- Operating system: Ubuntu 22.04 LTS or later recommended. Debian 11+ or CentOS 8+ also work.
- ML/AI: PyTorch (primary), TensorFlow (secondary), scikit-learn, CUDA Toolkit 11.8+, Python 3.11+
- DevOps: Docker 20.10+, Kubernetes, CI/CD
- Monitoring: Prometheus, Grafana, custom alerting
- Security: firewall, VPN, SSH hardening
Costs
The NODR stake and the setup package listed on Run a node are separate.
| Item | Cost |
|---|---|
| Hardware, total | $8,000–$15,000 |
| GPU (RTX 4090 to A100/H100) | $1,500–$5,000 |
| CPU (high end Xeon/EPYC) | $1,000–$2,500 |
| RAM (128GB ECC) | $800–$1,500 |
| Storage (4TB NVMe) | $600–$1,200 |
| Motherboard, case, PSU | $1,000–$2,000 |
| Networking, UPS | $500–$1,000 |
| Monthly running costs, total | $300–$500 |
| Electricity (800-1000W sustained load) | $150–$300 |
| Bandwidth (1Gbps+ fiber) | $100–$150 |
| Cooling (HVAC) | $50–$100 |
Rewards
Oracles will share 30% of the node pool, which is 40% of protocol revenue. The target is ~20–25% APY. This is a design target, not a guarantee, and rewards are not being paid yet.
Rewards are a revenue share, not a fixed APY. They will be paid in NODR and capped by the Base-Rate Governor as a function of trailing protocol net revenue. Your share will depend on:
- Strategy performance, a share of future profits from the protocol's trading strategies. No strategy trades vault capital yet.
- Activity in the Oracle Chamber
- The accuracy and timeliness of the data you provide
- TrustFingerprint score. A higher score means a higher reward multiplier.
- Stake amount. Higher stakes earn proportionally more.
A monthly operational stipend is planned to cover costs.
Separately from node rewards, operators who stake NODR can earn a share of strategy profits through the profit share ladder in StrategyStakingV2. It is keyed to the amount staked, not to node tier: Entry 2% / Tier 1 5% / Tier 2 10% / Tier 3 15% / Max 20%.
How to join
- Prepare hardware that meets the requirements above. Applicants also need a proven track record on a lower tier and a technical interview.
- Fill in the Node Operator Application.
- Once approved and paid for the Oracle tier, the team sends you an install token.
- Run the installer with your token. See the Setup Guide.
Planned: onchain elections
Onchain Oracle elections are designed but not active. Full DAO governance is on the roadmap for Q1 2028. Once active:
- A candidate with a TrustFingerprint score of at least 0.90 will start an election.
- Current Oracle operators will vote in the Oracle Chamber.
- A candidate will need a 66% Oracle supermajority to be confirmed.