AI cost management and governance¶

Snowflake gives you a consistent way to understand, monitor, and manage AI usage alongside the rest of your platform activity. Across AI features, pricing is primarily based on consumption, including token-based usage where applicable, so teams can align spend to actual usage instead of fixed capacity. To support cost transparency, Cortex AI provides usage views that help you analyze activity over time, break down consumption, and connect usage to billing workflows already used across your organization. These views can be used for reporting, governance, showback, and internal monitoring. For detailed pricing by feature, model, and unit of consumption, refer to the consumption table, which provides the current pricing structure across Snowflake AI capabilities.

Usage views¶

Snowflake provides usage views to help you track AI consumption using the same core approach used across the platform. These views support analysis of usage over time and can help teams understand how AI activity maps to overall spend, whether they are monitoring adoption, reviewing trends, or supporting internal reporting. This allows finance, platform, and engineering teams to work from a common system of record when evaluating usage. Pricing details remain available in the consumption table, which outlines how individual AI features are billed. Together, usage views and pricing documentation provide a foundation for understanding and managing AI costs across your Snowflake environment.

Usage views for total cost¶

These views should be used when calculating AI usage and AI-related spend. Together, they provide the standard foundation for cost reporting across AI features.

Coming Soon: Cortex AI Guardrails Account Usage View

_{[1] UTC Converted to local means if your account is altered to local time it will display in local time. The underlying data is still in UTC.}

_{[2] CORTEX_REST_API_USAGE_HISTORY is billed in dollars and is not currently shown in account level DAILY_METERING_HISTORY.}

_{[3] CORTEX_SEARCH_DAILY_USAGE_HISTORY includes embeddings which need to be excluded from combined calculations as they are also shown in CORTEX_AI_FUNCTIONS_USAGE_HISTORY.}

Usage views for additional analysis¶

Use these views when you need more granular or feature-specific insight. They complement the primary views, but are not intended to serve as the standard source for AI cost totals.

Total Cost of Operations¶

Tokens, Messages, and others are not the only ways in which you are billed for Cortex AI, you also are billed for the query, warehouse time, and any other associated Snowflake charges. Through query_id, warehouse_id, user_id you should be able to calculate your total cost of operation. For more details please see the associated usage view or contact support.

Budget features¶

Snowflake budgets help organizations monitor credit usage and respond when spending approaches or exceeds configured thresholds. These features can support internal planning, alerting, and broader governance processes for AI usage as part of an overall cost management strategy. A budget defines a monthly spending limit for an account or for a custom group of Snowflake objects. Budgets can send notifications when spend is projected to exceed the configured limit, and Snowflake also supports custom actions for budgets based on either projected or actual consumption. This allows teams to pair spend monitoring with operational responses, using the same core budgeting model across Snowflake cost management workflows.

Resource budgets for AI features¶

Resource budgets let administrators define a monthly credit limit for a tagged Cortex Agent object and evaluate spend against that budget on a periodic basis. Because they use Snowflake’s tag-based cost attribution model, they fit into broader governance and budget management patterns already used across the platform. Snowflake also announced resource budgets for Snowflake Intelligence on the same date, extending this model across additional AI experiences.

Shared resource budgets for AI Features¶

A shared resource budget lets you track and control credit consumption for AI features – such as AI Functions, Cortex Agents, Cortex Code, and Snowflake Intelligence – broken down by the team or cost center consuming them. Instead of budgeting a resource that belongs to a single owner or with a single budget, this budget tracks AI features that are used by specific users. Those users are identified with tags, so you can group them into logical units like a cost center or team. For example, if both an engineering team and a finance team call the same AI function, you can set up separate budgets that each track only the credits consumed by their respective tagged users, even though both teams are using the same underlying AI feature.

Budget capability by feature¶

Not supported nor planned: Cortex Analyst, Cortex Fine-tuning.

Budget timing, enforcement, and automated actions¶

For Resource Budgets and Shared Resource Budgets you can attach stored procedures that are executed when spending reaches specific thresholds, which are expressed as a percentage of the spending limit and apply to the monthly budget period. Budget evaluation and enforcement are calculated periodically rather than instantaneously. After a budget threshold is exceeded, actions can take up to eight hours to take effect under normal operation, or up to two hours when using the latency-optimized option. Budgets are useful for ongoing spend management and policy enforcement, while still being part of a broader cost governance strategy that may also include usage monitoring and internal operational review.