ADA-C01 Question 425
Single answerMonitor AI and ML function costsA Snowflake administrator needs to identify which workloads are driving spend from AI/ML-related functions so the finance team can charge costs back to business units. Several teams run SQL that uses Cortex AI functions and some teams also invoke ML-related processing from SQL. The administrator wants a reliable way to monitor these costs over time at the query level without manually reviewing every worksheet. Which approach should the administrator use?
- A
Query ACCOUNT_USAGE.QUERY_HISTORY and use the CREDITS_USED_CLOUD_SERVICES column to measure AI/ML function costs for each query.
- B
Query ACCOUNT_USAGE.METERING_DAILY_HISTORY and attribute all daily cloud services credits to AI/ML workloads based on warehouse usage patterns.
- C
Query ACCOUNT_USAGE.CORTEX_FUNCTIONS_USAGE_HISTORY and join it with query history or tags to attribute AI function consumption to teams and workloads.
- D
Use WAREHOUSE_METERING_HISTORY because AI/ML function costs are billed as warehouse credits whenever SQL calls an AI function.
Show answer and explanation
Correct answer: C
Explanation
To monitor AI and ML-related function costs in Snowflake, administrators should use the dedicated Account Usage telemetry designed for those services rather than trying to infer costs from warehouse or generic cloud services metering. For Snowflake Cortex AI functions, ACCOUNT_USAGE.CORTEX_FUNCTIONS_USAGE_HISTORY provides the relevant consumption history. For operational chargeback, best practice is to correlate this usage with QUERY_HISTORY, query tags, users, roles, and business metadata so costs can be assigned to the correct team or application. By contrast, METERING_DAILY_HISTORY is useful for high-level account cost trends, and WAREHOUSE_METERING_HISTORY is appropriate for warehouse compute analysis, but neither is sufficient for accurately attributing AI function spend. This aligns with Snowflake guidance to use Account Usage views specific to the billed service when monitoring and allocating specialized feature costs.
- A. Incorrect.
Incorrect. QUERY_HISTORY is useful for identifying which SQL statements invoked particular functions and for associating activity with users, roles, warehouses, and query tags. However, CREDITS_USED_CLOUD_SERVICES is not a reliable measure of AI/ML function consumption. AI-related function usage is tracked through dedicated usage views rather than inferred from generic cloud services credits.
- B. Incorrect.
Incorrect. METERING_DAILY_HISTORY provides account-level daily credit consumption by service type, which is too coarse for attributing AI/ML spend to specific workloads or teams. It may help with total cost trending, but it does not provide the query-level or function-level visibility needed for accurate chargeback.
- C. Correct.
Correct. CORTEX_FUNCTIONS_USAGE_HISTORY is the appropriate Account Usage view for monitoring Snowflake Cortex AI function consumption. In practice, administrators can combine this usage data with QUERY_HISTORY, users, roles, databases, or query tags to identify which teams and workloads are responsible for the spend and to support internal chargeback or showback.
- D. Incorrect.
Incorrect. Warehouse metering tracks compute used by virtual warehouses. AI function charges such as Cortex function usage are not simply represented as warehouse credits, so WAREHOUSE_METERING_HISTORY alone will miss or misattribute those costs. This is a common mistake when administrators assume all SQL-related cost is warehouse-based.