Skip to main content

​
What you charge

You advertise one rate per allowlisted token you accept. Rate is per MB, denominated in the smallest unit of the payment token. For USDC: 
advertise_rate = 1000000   # 0.000001 USDC ≈ $0.001 per MB ≈ $1/GB
Rates must fall within the governance-set [MIN_RATE, MAX_RATE] for the token.

​
What determines selection

Clients and peers select candidates by the unified score: 
score = rate_per_mb × rtt_ms × (1 / max(reputation, 0.1)²)
Lower is better. Three levers, two you control:
LeverYour controlStrategy
RateFullPrice at cost + margin; track competitive rates
RTTVia region choice and peeringLocate near your market; invest in low-latency connectivity
ReputationIndirect, via behaviorDon’t phantom-announce, don’t corrupt, sync blacklist, hit your advertised rate
A 2× rate reduction improves score linearly — halving rate halves score. A 2× reputation improvement improves score quadratically — doubling reputation cuts score 4×. Don’t cut rates if your reputation is already the bottleneck.

​
How you earn

Per-MB voucher revenue for every byte streamed:
  • Client → node — client pays you directly.
  • Node → node (cache miss) — the requesting node pays you per MB of upstream pull.
All payments flow through on-chain payment channels. Vouchers are off-chain EIP-712 signatures; settlement happens at channel close or via dispute.

​
What you pay

  • Outgoing per-MB for cache-miss pulls from other nodes (if pull_through: true).
  • Origin backend egress (if origin-backed) — your S3 / R2 / B2 bill.
  • Gas for occasional on-chain operations (channel close, stake adjustments, takedown polls).
  • Stake opportunity cost — TOKEN locked in StakingRegistry is not earning elsewhere.

​
Egress cost framing for origin-backed nodes

Origin-backed nodes set the effective price ceiling. If your S3 bucket charges $0.09/GB egress, you must charge at least that per MB pulled from you or lose money on origin pulls.
BackendEgressBreak-even rate
AWS S3 Standard$0.09/GB~90/TB≈90/TB ≈ 0.00009/MB
Cloudflare R2$0/GB (to Workers)Near zero
Backblaze B2$0.01/GB~$0.00001/MB
Self-hosted MinIOYour bandwidth billVaries
The protocol doesn’t force you to price at cost — rates are market-driven — but systematic under-pricing leads to capital loss.

​
Effective revenue model

For a given blob:
  • First serve per cache node: cache node pays origin-backed node (once). Origin-backed earns on the pull.
  • Subsequent serves per cache node: cache node serves from local cache, keeping 100% of the per-MB client revenue.
  • Amortization: if a cache node pays 0.001/MBtopulloncebutthenservesthesameblob10×at0.001/MB to pull once but then serves the same blob 10× at 0.002/MB to clients, it nets $0.019/MB amortized.
Cache nodes live or die on amortization — cache hit rate is the key operational metric.

​
Rate manipulation risk

You must not charge more per MB than you advertised. slash_sig on RateChange gossip messages creates on-chain counter-evidence. Charging 1.1× your advertised rate is a rate-manipulation offense — slashable 5% / 10% / 20% in escalating 90-day windows. If you need to change rates mid-stream, close the current session cleanly and let the client open a new session at the new rate. Never silently raise charged rate above advertised.

​
No rewards emissions

Revenue is 100% delivery-based. No ongoing emissions subsidize nodes. The filter is: operators who can’t earn on delivery economics aren’t running sustainable nodes.

​
Break-even napkin math

For a pure-cache node priced at 0.002/MBwitha900.002/MB with a 90% cache hit rate and 0.001/MB upstream pull cost:
  • Revenue per MB served: $0.002
  • Cost per MB served: 0.1 × 0.001=0.001 = 0.0001
  • Gross margin: $0.0019/MB
At 100 TB/month throughput:
  • Revenue: ~$200,000
  • Pull cost: ~$10,000
  • Less hosting + bandwidth bill (varies), stake opportunity cost, and gas.
Rough-order numbers only — actual economics depend heavily on hit rate, which depends on what content you cache. See node onboarding for the actual registration flow and reputation for score maintenance.