Computing Power Trading Platform

The Computing Power Trading Platform is the operational heart of the StarMiner protocol where compute supply and demand meet in real time under decentralized conditions.

It functions as a programmable exchange for GPU cycles, enabling thousands of Service Requesters to submit compute workloads, and Provider Nodes to offer GPU power all through permissionless infrastructure with market-based pricing and smart contract enforcement.

Unlike centralized clouds or auction-based DePIN protocols, StarMiner builds a continuous, bidirectional compute marketplace, where workloads are not just matched, but dynamically prioritized, priced, and verified.


Platform Components

1. Supply Registry

  • A live, decentralized index of all available Provider Nodes.

  • Includes hardware profiles (e.g., A100, H100, RTX 4090), region metadata, real-time availability, energy profile, and reputation score.

  • Continuously updated via on-chain and off-chain telemetry.

2. Demand Queues

  • As Service Requesters submit tasks, jobs are streamed into categorized queues:

    • AI Training

    • Inference & Prediction

    • 3D Rendering

    • Simulation & Research

  • Each task includes metadata: AGPU budget, latency preference, urgency, and task type.

3. Matching Engine

  • An algorithmic router that pairs tasks to nodes based on:

    • Budget-to-capacity fit

    • Reputation & past performance of providers

    • Proximity (region, latency)

    • TEE/Privacy requirement compliance

  • Routing occurs in milliseconds, enabling high-throughput task settlement without bidding delays.

4. Execution Contracts

  • Each matched task generates a Compute Service Contract (CSC), deployed on-chain.

  • Defines job ID, cost, expected runtime, completion proof hash, and payment conditions.

  • Escrows AGPU tokens from requester and releases them upon verified task completion.


Workflow Overview

  1. Node Registration: Providers register with benchmarking info and AGPU staking (optional).

  2. Job Submission: Users submit compute tasks via dashboard, API, or SDK.

  3. Job Routing: StarMiner matches jobs to optimal nodes based on technical and market conditions.

  4. Smart Contract Deployment: A CSC is created for each job.

  5. Execution: The Provider Node processes the task and submits result + proof.

  6. Settlement: AGPU is released from escrow and paid to the node; job result is returned to the user.


Trading Model

  • StarMiner uses a streaming trade model rather than a batch-based auction.

  • Tasks are not bid for by nodes — instead, the protocol dynamically prices compute via the Multi-Tier Pricing (MTP) engine, and assigns jobs based on score-fit rather than raw price.

  • This eliminates gas wars, front-running, or sniping behavior, and allows lower-friction matching.

Optional Features:

  • Compute Vaults: Requesters can pre-fund vaults to receive automated compute on defined triggers (e.g., weekly inference jobs).

  • Reserved Liquidity: Nodes can stake AGPU into resource pools that receive preferred routing during demand surges.


Economic Efficiency

The trading platform ensures:

  • High utilization of idle GPUs

  • Price discovery that adapts to actual usage patterns

  • Low latency for urgent or enterprise-grade compute needs

  • Transparent cost structure without platform overhead or hidden markup


Compliance and Auditability

  • All trades and task histories are recorded on-chain (task hash, metadata, timestamps, payment).

  • For sensitive jobs, off-chain compute proofs can be recorded with optional ZK validation or TEE attestations.

  • DAOs and enterprise clients can audit compute cost, performance, and provider reliability over time.

Last updated