Getting Started with Snowflake Per-User Quotas for User-Level Cost Management of Warehouses and AI Features

Snowflake's Per-User Quota, now in Public Preview, offers a way to manage user-level consumption of warehouse and AI resources. This feature allows administrators to set credit limits for individual users, addressing issues such as a few users consuming disproportionate resources, delayed cost overrun awareness, the need for manual intervention when limits are breached, and difficulty tracking resource consumption.

Per-User Quota enables setting credit limits on a defined cycle, with warehouses supporting only monthly limits and AI-related limits set monthly, daily, or both. Custom cycles are not currently supported. Users can be targeted all at once, excluding specific tagged users, or reassigned to different quotas by changing their tags. When a limit is reached, custom actions can be invoked for warehouses and a built-in block mechanism is available for AI-related requests.

The resources that can be targeted include warehouses, AI Functions, Snowflake Intelligence, Cortex Agents, CoCo, and Snowflake CoCo. These are referred to as "domains" when configuring a Per-User Quota. You can monitor either the Warehouse domain or AI domains within a single quota, but not both, due to the difference in measurement units.

Compared to CoCo's built-in limits, Per-User Quota offers a broader scope, supporting multiple AI domains, and integrates with Snowsight's budget management capabilities. While CoCo is simpler for controlling CoCo costs alone or applying a uniform account-wide cap, Per-User Quota is ideal for cross-cutting control across warehouses and AI features. The feature also provides flexibility in targeting via tags and supports Snowsight's budget management with notifications and limit-triggered actions.