BigQuery Token Quotas: Daily AI Spend Caps by Project and User

Google has made daily token quotas generally available for the generative AI functions in BigQuery, giving enterprises a way to cap AI consumption by project and by individual user. InfoWorld reported the change, which is aimed at making spending on AI functions more predictable.

What happened

The quotas cover both input and output tokens and can be set at the project level or the individual-user level, with limits applied per day. Administrators manage them through the IAM and Admin tab in the Google Cloud Console. The controls apply to seven functions: AI.GENERATE, AI.GENERATE_TABLE, AI.GENERATE_TEXT, AI.CLASSIFY, AI.IF, AI.AGG and AI.SCORE.

The move builds on AI-based SQL functions that Google added to its managed data warehouse and analytics platform last November. Those functions use large language models to structure unstructured data, such as text, so it can be analysed with SQL. Previously, SQL could not interpret that kind of content, and preparing it was slow. As usage of these functions grows, the new quotas give organisations a direct mechanism for controlling how much is consumed.

Why it matters

For organisations running analytics at scale, AI functions inside the data warehouse change how data teams work. A query that once required a separate pipeline to prepare unstructured content can now call a model directly from SQL. That makes AI a routine part of analytics rather than a specialist project, and it shifts the question from whether AI can be used to how much of it should be used, by whom and for what.

Usage-based pricing is hard to forecast when consumption is spread across teams and individuals. Daily quotas at both project and user level give finance and platform leaders a way to set guardrails, and they give data practitioners a clearer picture of the limits they work within. Taken together, this points to cost governance becoming a design question for AI-enabled data platforms, not just a billing concern.

By the numbers

• Seven BigQuery generative AI functions are covered by the daily token quotas, from AI.GENERATE through to AI.SCORE.
• Two levels of control are available: the project and the individual user.
• Daily limits apply to both input and output tokens, managed from the IAM and Admin tab in the Google Cloud Console.

The Renascence take

Most coverage will treat this as a cost-control feature. The more important point is what it reveals about how AI consumption behaves inside an organisation: it is invisible at the point of use, and each query feels trivial to the person running it.

Budget caps are a blunt tool. The real risk is that AI calls become a habit that no one designed, so the people closest to the data have no feedback on what each action costs. Teams should decide which tasks justify a model call and which can be handled with simpler SQL, then set quotas to match those decisions. Treat quotas as the boundary of a service design, not just a spending ceiling.