AI coding tools are grappling with pricing models that often frustrate users while threatening company profitability. Companies like Cursor have faced backlash after poorly communicated pricing changes, highlighting a broader issue: as AI usage becomes more expensive, companies struggle to balance customer expectations for unlimited access with the reality of unsustainable costs. The landscape has shifted; instead of prices dropping as anticipated, expenses have risen, especially for advanced models that require significant token consumption. Pricing strategies in this space are complex. Companies must maintain margins, ease the buying process for customers, and remain competitive. The article notes that many AI coding startups undercharge, failing to capture the value they provide. For example, Cursor's previous pricing model was unsustainable because it didn't account for the variability in token consumption across different tasks. Experts suggest that pricing should align with user behavior, making it easy for individuals to make purchasing decisions without needing finance approvals for unpredictable costs. Common pricing strategies include pay-per-use and credit-based models. Pay-per-use, as seen with Claude API, offers transparency and direct correlation between usage and costs but complicates revenue forecasting and can be unstable. On the other hand, credit-based pricing, exemplified by Replit, allows for predictable costs but can lead to user dissatisfaction if they feel they are paying for unused credits. Each model has its advantages and drawbacks, and as user demands evolve, companies must continuously adapt their pricing strategies to stay viable in a competitive market.