kWh rate×GPU watts=price No markup
GPUShare
Built a self-hosted platform for trusted users to share local AI inference and Blender rendering, with optional cloud routing and estimated usage-cost allocation.
Result The application implements auth, local and optional cloud inference, MCP tool routing, queued rendering, and append-only usage accounting. Host availability and a stale installer migration command remain operational limits; no adoption claim is made.