Google BigQuery Adds Daily Token Quotas for Generative AI Functions

Google has made daily token quotas generally available for BigQuery’s generative AI functions. This feature allows enterprises to limit artificial intelligence consumption by project and individual user, providing a mechanism to manage costs and resource usage within the managed data warehouse platform.

The quotas cover both input and output tokens at the project and individual-user levels per day. They apply to several AI-based SQL functions, including AI.GENERATE, AI.GENERATE_TABLE, AI.GENERATE_TEXT, AI.CLASSIFY, AI.IF, AI.AGG, and AI.SCORE. According to InfoWorld, these limits help organizations control the volume of tokens processed by large language models integrated into their analytics workflows.

Managing Quotas and Defaults

Administrators can manage these limits via the IAM and Admin tab in the Google Cloud Console. By default, input tokens are capped at 200 billion per project and 150 billion per user per day. Output tokens, which include reasoning tokens used by the large language model, are capped at 20 billion per project and 15 billion per user per day. Enterprises can request increases to these default quotas if their workload requires higher throughput.

Google also supports programmatic management of these quotas through the Cloud Quotas API, the gcloud command-line interface, and Terraform. This allows infrastructure teams to automate quota enforcement and align AI resource allocation with existing governance frameworks. The company noted in a blog post that these tools provide granular control over how different teams consume AI capabilities within the BigQuery environment.

Impact on Queries and Embedding Functions

Embedding functions such as AI.GENERATE_EMBEDDING, ML.GENERATE_EMBEDDING, and AI.EMBED are not subject to these new quotas. Google stated that these functions run on different, substantially cheaper models, so they are managed separately from the primary generative AI functions. This distinction ensures that cost-effective embedding operations do not interfere with the capacity limits applied to more resource-intensive text generation and classification tasks.

When a query exceeds the established token quota, BigQuery rejects it with a “QuotaExceeded” error. If the quota is exhausted while a query is already running, the platform stops sending new requests to the large language model. For AI.IF queries, BigQuery may return partial results if the number of unprocessed rows remains within a specified error limit. If that limit is exceeded, the query fails entirely.

For AI.GENERATE queries, execution may complete even if the quota is hit during processing. However, rows that could not be processed will have no generated output and will carry an error message in their status field. This behavior allows administrators to identify which specific data points failed due to quota constraints, facilitating troubleshooting and capacity planning.

What should you do?

Review your current BigQuery usage to identify which AI functions are consuming the most tokens. Check whether your default quotas align with your expected workload volumes. Monitor query logs for “QuotaExceeded” errors or partial results, which indicate that your current limits are too restrictive for your operational needs.

Background: Generative AI

Generative AI examples include text completion, image synthesis and code generation. These systems predict the next likely token based on vast training datasets. Understanding the underlying mechanics helps you identify risks like data leakage and hallucination in your infrastructure.

Read the full guide: Generative AI Examples: Mechanisms, Risks and Control

What to do and how to stay safe: Google

• Audit your BigQuery projects to identify which users or workloads are approaching default token limits.
• Configure alerts for quota utilization to receive notifications before limits are reached.
• Request quota increases through the Google Cloud Console if your analytics pipelines require higher throughput.
• Review query error logs to determine if partial results are affecting data integrity or downstream processes.

Step-by-step guide: Prompt Injection Warning Signs: How to Spot AI Manipulation

General security guidance from the Threat Times newsroom. It is not confirmed advice from the organisations named in this story.

Frequently asked questions

Which BigQuery AI functions are subject to daily token quotas?

Quotas apply to AI.GENERATE, AI.GENERATE_TABLE, AI.GENERATE_TEXT, AI.CLASSIFY, AI.IF, AI.AGG, and AI.SCORE functions.

Are embedding functions included in these token limits?

No, embedding functions like AI.GENERATE_EMBEDDING are excluded because they use cheaper models.

What happens if a query exceeds its token quota during execution?

BigQuery stops sending new requests to the LLM, potentially resulting in partial results or query failure.