# Bring-your-own-key pricing must prove value beyond inference cost ## Claim When customers supply the model key, an AI product has to price the value of its workflow rather than mark up inference it no longer buys. <!-- atomicity: 1 idea ✓ --> ## Evidence > [!quote] > "You can't like charge usage on top of something that is not running your cost center." The discussion contrasts a large annual surcharge for bring-your-own-key access with a pricing story based on the value of resolved work rather than reimbursement for model usage. ## Scope & conditions The product may still charge for orchestration, support, storage, evaluation, and business outcomes. The constraint is explanatory: once inference moves onto the customer's bill, usage volume can't justify a vendor markup unless the vendor can name a separate service that scales with that usage. ## Links - [[3. Resources/Permanent Notes/Outcome-Based Pricing Transfers AI Cost Risk From Buyer to Vendor]] — BYOK moves inference risk back to the buyer, making outcome value the clearer basis for the vendor's remaining price - [[3. Resources/Permanent Notes/Subscription Inference Is Sold Below Cost Creating a Tokenmaxxing Arbitrage]] — removing bundled inference also removes the subsidy that rewards customers for maximizing token burn - [[3. Resources/Permanent Notes/A Mature Plugin License Can Cost Less Than Rebuilding Its Integration Network]] — both force software pricing to name the durable value around execution cost rather than the cheap code or tokens themselves - [[3. Resources/MOCs/AI Cost & Tokenomics MOC]] — home for the shift from metered inference to value-based AI pricing ## Source - [[3. Resources/Readwise/Articles/141 The Era of Sol]] — argues that BYOK prevents a vendor from justifying price as a markup on model usage it doesn't fund