Skip to main content
Running a Livepeer orchestrator pays through two distinct revenue streams with different mechanics. Know both before you decide how to configure the node, grow stake, or price work.
New here? Start here for the mechanics. For what those mechanics mean in practice for your specific hardware and situation, see Feasibility and Economics. For the step-by-step operations of reward calling and commission setting, continue to Rewards and Fees.

The two revenue streams

ETH Service Fees

Gateways pay you in ETH for every transcoding or AI inference job you complete. Demand-driven: you only earn when you win jobs.

LPT Inflationary Rewards

The protocol mints new LPT every round and distributes it to active staked orchestrators. You must call Reward() each round to claim your share.
These streams behave very differently. ETH fees depend on market demand and your competitiveness. LPT rewards depend on your stake, the current inflation rate, and whether you call Reward() every round without missing one. Both matter for total income.

ETH service fees

Every time a gateway routes a job to your orchestrator and you complete it successfully, you earn ETH. The gateway sends payment via Livepeer’s probabilistic micropayment system — a series of signed tickets that represent ETH obligations. When you receive a winning ticket, you submit it on-chain to redeem the ETH. Fees exist for two workload types: AI inference commands higher fees per job because the work is more GPU-intensive. An orchestrator running popular AI models warm on a high-VRAM GPU is positioned very differently in the fee market than one doing transcoding alone.

What determines how much work you win

Gateways route jobs through a competitive filter across multiple factors at once:
Stake weight         → higher stake = larger active set position = more segments
Capability match     → do you support the requested pipeline or resolution?
Performance history  → transcoding success rate, AI inference latency, uptime
Pricing              → are your fees below the gateway's maxPricePerUnit?
Model warmth (AI)    → warm models win over cold ones for latency-sensitive jobs
None of these factors alone is decisive. A high-stake orchestrator with uncompetitive pricing wins fewer jobs than a moderate-stake node priced correctly. A well-priced node without warm models loses AI jobs to one that has them.

LPT inflationary rewards

Each round (approximately every 22 hours on Arbitrum), the Livepeer protocol mints new LPT and distributes it to active orchestrators proportional to their total stake — that is, their own self-bonded LPT plus all delegated stake from third parties. Three things must be true for you to receive that round’s rewards:
  1. You are in the active set (top orchestrators by stake weight)
  2. You successfully call Reward() on-chain that round
  3. Your node has LPT bonded (self-stake)
Inflation rate dynamics: The rate is dynamic. It rises when less than ~50% of total LPT supply is bonded, and falls when more than ~50% is bonded. The protocol targets 50% participation. As of early 2026, approximately 61% of LPT supply was bonded, so the inflation rate was gradually declining towards its floor. Treasury allocation: As of LIP-89 (March 2026), 10% of each round’s newly minted LPT goes to the Livepeer Treasury before distribution to orchestrators. The remaining 90% is distributed as before.
Missing a round’s Reward() call means that round’s inflation allocation is permanently lost. For a well-staked orchestrator, a single missed round forfeits a noticeable amount of LPT. Automate this.

Commission parameters

When you register as an orchestrator, you set two commission parameters that determine how you split earnings with delegators. Both update on-chain immediately, and frequent changes damage delegator trust.

Reward cut

The percentage of LPT inflationary rewards you keep as the operator. Delegators receive the rest, split proportionally by their stake.
Your LPT = Round issuance share × rewardCut
Delegator LPT = Round issuance share × (1 − rewardCut) × (their stake / total stake)
Example: Your orchestrator earns 100 LPT this round. Your rewardCut is 20%.
  • You keep: 20 LPT
  • Delegators split: 80 LPT (in proportion to their individual stakes)

Fee share

The percentage of ETH job fee revenue that is shared with your delegators. The remainder you keep.
Delegator ETH = Job fees × feeShare × (their stake / total stake)
Your ETH = Job fees × (1 − feeShare)
Example: You earn 0.1 ETH in fees this round. Your feeShare is 50%.
  • You keep: 0.05 ETH
  • Delegators split: 0.05 ETH
Commission rates directly affect delegator yield calculations on the Livepeer Explorer. Delegators compare orchestrators by ROI. Setting competitive rates attracts more stake, which increases your inflation share and job routing weight — a compounding dynamic. See Getting Delegates for the full strategy.

Transcoding vs AI fees in practice

The fee landscape is shifting. AI inference jobs now represent a growing share of total network fee revenue. A key distinction: for video transcoding, stake weight directly determines your position in the routing queue. For AI inference, the gateway first filters by capability and price — your stake matters less than having the right pipeline running at a competitive price. This supports two parallel strategies on a single node:
  • Transcoding income: grow stake to capture more segments
  • AI inference income: run popular models warm at competitive prices

Payment flow summary

Gateway routes job → you complete it

Gateway sends payment ticket (signed ETH obligation)

You receive winning tickets (probabilistic — not every ticket wins)

Redeemer submits winning tickets on-chain (Arbitrum)

ETH arrives in your orchestrator wallet
For the full mechanics of how probabilistic micropayments work, ticket face values, and on-chain redemption, see Payments.

Monitoring your earnings

Check your reward call history on Explorer regularly. A gap in the reward history means missed rounds and lost LPT. For monitoring setup, see Metrics and Monitoring.

Watch: Orchestrator economics overview

Continue in this section

Last modified on March 16, 2026