Google has quietly swapped the old prompt-based usage quotas on its Gemini Apps for a new system tied to compute resources. The change, which applies to both free and paid tiers, is already squeezing power users who rely on the AI for heavy workloads.
How the new quotas work
Instead of counting the number of prompts a user sends, the new limits track the computational cost of each request. Complex tasks — like analyzing long documents, generating code, or processing images — consume more compute units than simple queries. That means a single advanced request can eat up the same allocation as dozens of basic ones.
Google hasn't published a detailed breakdown of how many compute units each action costs, leaving users to guess what triggers a cap. The company says the shift is designed to ensure fair access and prevent abuse, but the lack of transparency has frustrated many.
Impact on power users
Developers, researchers, and content creators who push Gemini to its limits are feeling the pinch. Some report hitting daily limits after just a few hours of intensive use. A user on a paid subscription told a tech forum that their quota ran out before lunch, forcing them to wait until the next day or pay for extra compute credits.
The change is especially hard on those who use Gemini for batch processing or iterative tasks. A single debugging session that involves multiple code rewrites can burn through a week's worth of compute in an afternoon. Google has not announced any adjustments for high-volume users, and the company's support pages offer only generic advice about managing usage.
Competitive landscape reshuffles
The move is reshaping the AI services market. Rivals like OpenAI and Anthropic still use prompt-based or token-based pricing, which some users find more predictable. Startups that built workflows around Gemini's old quotas are now scrambling to adapt or migrate to other platforms.
Smaller AI providers see an opening. Several have started marketing their services as “no-surprise compute” alternatives, emphasizing fixed pricing per task. The shift could accelerate a broader trend toward usage models that favor casual users over heavy adopters.
Google's decision also puts pressure on enterprise customers who were evaluating Gemini for internal tools. Without clear compute costs, budgeting for AI usage becomes a guessing game. The company has not said whether it plans to offer enterprise-specific pricing or volume discounts.
For now, power users are left to monitor their compute meters and adjust their habits. Whether Google will refine the system or competitors will capitalize on the frustration remains an open question.
