Pricing AI: A Unit-Based Problem?

Pricing AI products is challenging due to the complex and costly nature of running AI models, which involve significant compute and GPU power. The process is broken down into four key considerations: units, tiers, terms, and implementation. Units can range from requests, tokens, successes, to physical pricing, each offering different ways to measure AI usage. Tiers include model and subscription tiering, with variations depending on the complexity and cost of AI models, while terms involve managing billing cycles and mitigating risks related to customer payments. Implementation focuses on accurately tracking AI usage through methods like snapshot and event recording, ensuring that billing aligns with operational costs. The overall difficulty in pricing AI products stems from balancing the need to cover high operational expenses while offering clear and fair pricing models to customers.