Sunk Cost is a tool designed to help you calculate how long it takes for a local AI machine to pay for itself compared to using commercial APIs. It evaluates which open models fit your hardware, how fast they run, and calculates the exact token volume required to recover your hardware investment.

The platform addresses the financial uncertainty of purchasing dedicated hardware for local artificial intelligence by factoring in electricity costs, hardware pricing, and API cost depreciation. It provides honest, data-driven estimates that can even conclude if a machine will never break even.

It is built for developers, hobbyists, and professionals considering a hardware investment to run local large language models instead of relying entirely on cloud-based APIs.