Gemini's Pricing Design: Google's Strategic Contraction in AI-as-a-Service, Not Openness

The afternoon before yesterday, our platform team received an urgent help request: a business line was using Gemini Pro for real-time conversation summaries, and three days after launch, the API suddenly started throwing 429 errors. Engineers spent ages investigating only to discover that the free tier's requests-per-minute limit had dropped from 60 to 10, while the business side completely missed Google's update email. This isn't an isolated case. When teams plug in AI capabilities like utilities (water/electricity), hardly anyone bothers to read those dense rate-limiting docs.