} ; }; Orchestrators earn from two independent sources: ETH service fees for processing jobs, and LPT inflation rewards for protocol participation. Both require Active Set membership. Earnings from fees depend on workload volume; earnings from rewards depend on total stake and the Reward Cut configuration. ## Two Revenue Streams An Orchestrator's total earnings combine two streams that are structurally independent. ```mermaid theme={"theme":{"light":"github-light","dark":"dark-plus"}} %%{init: {'theme': 'base', 'themeVariables': {'primaryColor': '#18794E', 'primaryTextColor': '#fff', 'primaryBorderColor': '#3CB540', 'lineColor': '#3CB540', 'mainBkg': '#18794E', 'nodeBorder': '#3CB540', 'clusterBkg': 'transparent', 'clusterBorder': '#3CB540', 'titleColor': '#3CB540', 'edgeLabelBackground': 'transparent', 'textColor': '#3CB540', 'nodeTextColor': '#fff'}}}%% flowchart LR subgraph FEES["Stream 1: Service Fees (ETH)"] A["Gateway sends job\n+ payment ticket"] --> B["Orchestrator\nexecutes work"] B --> C["Winning ticket redeemed\non TicketBroker"] C --> D["ETH paid to Orchestrator\n(fee cut retained;\nremainder to fee pool)"] end subgraph REWARDS["Stream 2: Inflation Rewards (LPT)"] E["Orchestrator calls\nrewards each round"] --> F["Protocol calculates\nshare based on stake"] F --> G["LPT minted and distributed\n(reward cut retained;\nremainder to delegator pool)"] end classDef default fill:#1a1a1a,color:#fff,stroke:#2d9a67,stroke-width:2px style FEES fill:#0d0d0d,stroke:#2d9a67,stroke-width:1px style REWARDS fill:#0d0d0d,stroke:#2d9a67,stroke-width:1px ``` Stream Currency Source Frequency Depends on **Service fees** ETH Gateways paying for compute (transcoding or AI inference) Per job (ticket redemption is probabilistic) Volume of work received; price configuration; uptime **Inflation rewards** LPT Protocol inflation (new LPT minted per round) Once per round (\~22 hours) via reward call Total bonded stake (self-stake + delegated); Reward Cut setting ## Service Fees (ETH) Service fees are paid by Gateways via **probabilistic micropayment (PM) tickets**. Each job segment or inference request carries a ticket with a face value and a win probability. When a ticket "wins" (determined cryptographically), the Orchestrator redeems it on the TicketBroker contract on Arbitrum for the face value in ETH. The expected value of a ticket equals its face value multiplied by the win probability. Over a large number of tickets, actual earnings converge to the expected value. ### Pricing units Fee income per job is determined by the price the Orchestrator charges and the volume it processes: Workload Pricing unit Configuration flag **Video transcoding** Wei per pixel per segment `-pricePerUnit` **Batch AI inference** Wei per pixel or per millisecond (pipeline-dependent) `-pricePerCapability` (per pipeline and model) **Real-time AI (Cascade)** Interval-based during stream duration `-livePaymentInterval` Setting prices too high means Gateways will not select the Orchestrator (their `-maxPricePerUnit` ceiling acts as a filter). Setting prices too low reduces earnings per job. The network forms a market price - use `livepeer_cli` to survey current rates before setting your own. ### Fee distribution The Orchestrator's **Fee Cut** determines what share of collected fees goes to the Orchestrator versus the shared fee pool distributed to Delegators. ``` Total fee ticket value = Orchestrator fee cut + Delegator fee pool ``` For example, at a Fee Cut of 5%: the Orchestrator keeps 5% of ticket value; 95% goes to the fee pool distributed proportionally among Delegators. ## Inflation Rewards (LPT) The Livepeer Protocol mints new LPT each round to incentivise participation. Orchestrators in the Active Set earn a share of this inflation proportional to their total bonded stake. ### The reward call To claim LPT rewards for a round, the Orchestrator (or an automated process) must call the reward function on the BondingManager contract **once per round**. Missing a reward call means forgoing that round's LPT rewards entirely - this is not recoverable. Most production Orchestrators automate reward calling. The go-livepeer node can be configured to do this automatically. See for configuration details. ### Reward distribution The Orchestrator's **Reward Cut** determines what share of the round's LPT goes to the Orchestrator versus Delegators: ``` Total round reward = Orchestrator reward cut + Delegator reward pool ``` A Reward Cut of 25% means the Orchestrator keeps 25% of the LPT earned; 75% is distributed to Delegators in proportion to their bonded stake. Reward Cut setting Orchestrator keeps Delegators receive Implication 0% None 100% Maximum incentive to attract delegated stake; operator earns nothing from rewards 25% 25% 75% Common production setting; balances operator return with Delegator appeal 100% 100% None Maximum operator return; Delegators receive no LPT rewards from this Orchestrator ### Active Set requirement Only the top 100 Orchestrators by total bonded stake (self-stake + delegated stake) are in the **Active Set** in any given round. Only Active Set members receive inflation rewards and can receive Gateway jobs. Stake below the threshold for Active Set membership means no earnings from either stream. The Active Set threshold varies with network participation. Monitor your rank on the [Livepeer Explorer](https://explorer.livepeer.org) to ensure you remain in the Active Set. ## Orchestrator Costs Earnings must be weighed against operational costs. Orchestrator costs fall into three categories: Cost category What it covers Notes **Hardware** GPU(s), server, networking Capital expenditure; amortised over Orchestrator lifetime **Infrastructure** Electricity, bandwidth, colocation or cloud fees Ongoing operating expense; electricity is the largest recurring cost for GPU operators **Staking opportunity cost** LPT locked in self-bond cannot be sold or used elsewhere Relevant when evaluating ROI against alternative LPT uses (delegating to another Orchestrator) Unlike Gateways, Orchestrators do not pay for the jobs they process - they are paid. The cost structure is infrastructure and staking, not service fees. ## Payment Flow End-to-End This is the complete flow from Gateway payment through Orchestrator receipt to Delegator distribution. ```mermaid theme={"theme":{"light":"github-light","dark":"dark-plus"}} %%{init: {'theme': 'base', 'themeVariables': {'primaryColor': '#18794E', 'primaryTextColor': '#fff', 'primaryBorderColor': '#3CB540', 'lineColor': '#3CB540', 'mainBkg': '#18794E', 'nodeBorder': '#3CB540', 'clusterBkg': 'transparent', 'clusterBorder': '#3CB540', 'titleColor': '#3CB540', 'edgeLabelBackground': 'transparent', 'textColor': '#3CB540', 'nodeTextColor': '#fff'}}}%% flowchart LR A["Gateway"] -->|"PM tickets\n(ETH on Arbitrum)"| B["Orchestrator"] B -->|"Fee cut\n(% retained)"| C["Orchestrator earnings"] B -->|"Fee pool\n(% to delegators)"| D["Delegator pool"] E["Protocol inflation"] -->|"LPT rewards\neach round"| B B -->|"Reward cut\n(% retained)"| F["Orchestrator LPT"] B -->|"Reward pool\n(% to delegators)"| G["Delegator LPT pool"] classDef default fill:#1a1a1a,color:#fff,stroke:#2d9a67,stroke-width:2px ``` ## Operator Models Orchestrators operate for different reasons. The earnings model looks different depending on context. If you have GPU hardware already running - mining, rendering, or sitting idle - adding Livepeer Orchestrator workloads converts spare capacity into ETH service fees and LPT rewards. The incremental cost is electricity and a small slice of GPU time. The marginal earnings are service fees (proportional to workload volume) plus LPT rewards (proportional to stake). This model favours operators with high-end consumer or prosumer GPUs (RTX 3090, 4090, A100, H100) and reliable connectivity. The solo-operator path is documented in the . Commercial Orchestrators serving application workloads (Daydream, Livepeer Studio, custom AI products) operate differently from solo GPU miners. Revenue is primarily from service fees rather than inflation rewards. The key metrics are throughput, latency, and reliability - not stake size. Pricing strategy matters more, and direct relationships with high-volume Gateways are common. This model requires understanding Gateway selection mechanics in depth. See for the factors Gateways weigh. Pool operators run the Orchestrator process and LPT staking infrastructure, then accept GPU workers from contributors who want to earn without managing the on-chain complexity. Revenue for pool operators typically comes from a fee charged to workers (percentage of their earnings) or from running some of the GPUs in the pool themselves. This model combines the earnings from running GPU compute with revenue from managing the protocol layer on behalf of workers. An Orchestrator that holds substantial self-stake but operates at low capacity or joins a pool earns primarily from inflation rewards rather than service fees. High Reward Cut retains more LPT per round. Low Reward Cut attracts more delegated stake, which increases the Orchestrator's share of the total round reward. This model is most relevant for large LPT holders evaluating whether to self-operate versus delegating to an existing Orchestrator. ## Configuring Prices Orchestrators set three categories of pricing parameter at startup or via `livepeer_cli`: ### Base price ```bash theme={"theme":{"light":"github-light","dark":"dark-plus"}} # Price per pixel per second for video transcoding -pricePerUnit=1000 # Enable automatic price adjustment based on network conditions -autoAdjustPrice=true ``` ### AI capability pricing ```bash theme={"theme":{"light":"github-light","dark":"dark-plus"}} # Per-pipeline, per-model AI pricing (JSON file path) -pricePerCapability=/path/to/ai-prices.json ``` Example `ai-prices.json`: ```json theme={"theme":{"light":"github-light","dark":"dark-plus"}} { "capabilities_prices": [ { "pipeline": "text-to-image", "model_id": "stabilityai/stable-diffusion-3-medium-diffusers", "price_per_unit": 4768371, "pixels_per_unit": 1 }, { "pipeline": "audio-to-text", "model_id": "openai/whisper-large-v3", "price_per_unit": 15000, "pixels_per_unit": 1 } ] } ``` ### Per-Gateway pricing ```bash theme={"theme":{"light":"github-light","dark":"dark-plus"}} # Set a different price for a specific Gateway address -pricePerGateway='{"0xGatewayAddress": 800}' ``` Per-Gateway pricing allows commercial Orchestrators to negotiate rates with high-volume Gateways independently of the base network price. See for competitive pricing guidance and current network rate benchmarks. ## Attracting Delegators Delegated stake increases an Orchestrator's share of inflation rewards and improves its Active Set ranking. Attracting and retaining Delegators is therefore an economic lever for Orchestrators. Delegators evaluate Orchestrators on: * **Reward Cut** - lower Reward Cut means more LPT to Delegators per round * **Fee Cut** - lower Fee Cut means more ETH to Delegators from service fees * **Reliability** - missed reward calls mean Delegators receive no rewards for that round * **Reputation** - track record of uptime, performance, and community engagement * **Stake already committed** - larger self-stake signals greater operator commitment The Livepeer Explorer shows all active Orchestrators and their current cut settings, stake, and recent reward-calling history. Delegators use this to compare options. See for operator strategies. ## Related Pages What Orchestrators are and how the role has evolved. Workload types and how Gateways select Orchestrators. How to set competitive prices for video and AI workloads. The buy-side view - how Gateways pay and what they maximise.