Sunk Cost Calculator Tracks Local LLM Break-Even — SkimNews
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Sunk Cost is a Show HN project that calculates the break-even point for running a local LLM rig versus paying API costs
- The tool is organized into three sections: the break-even calculation, how the user would put the rig to work, and assumptions the user can change
- Local inference speed is estimated as memory bandwidth divided by bytes read per token, with these estimates explicitly labeled where no direct measurement exists
- API speed only enters the time comparison between local and cloud approaches, not the cost calculation itself
Why it matters: For developers weighing self-hosted LLMs against API subscriptions, this tool replaces back-of-envelope math with a structured model, but the memory-bandwidth shortcut means users must validate the underlying tokens-per-second estimate against their own hardware before trusting any break-even figure.
Ask SkimNews
