ADA-C01 Question 429
Single answerMonitor AI usage and costsA Snowflake administrator needs to help the FinOps team understand which users and workloads are driving spend for Snowflake Cortex AI features. The team wants a daily report that can be joined to other account usage data and filtered by time range, user, and object. Which approach should the administrator use?
- A
Query the ACCOUNT_USAGE.CORTEX_FUNCTIONS_USAGE_HISTORY view to analyze AI function usage and credits over time, then join it to other ACCOUNT_USAGE views as needed.
- B
Use WAREHOUSE_METERING_HISTORY because all Cortex AI costs are billed through the virtual warehouse that executed the SQL statement.
- C
Query QUERY_HISTORY and sum compilation_time and execution_time for statements that call AI functions, because Snowflake does not expose AI usage in dedicated account usage views.
- D
Use DATABASE_STORAGE_USAGE_HISTORY because Cortex AI charges are recorded as database storage growth caused by generated embeddings and completions.
Show answer and explanation
Correct answer: A
Explanation
For monitoring Snowflake AI usage and costs, administrators should use the dedicated ACCOUNT_USAGE telemetry Snowflake provides for Cortex usage rather than trying to infer cost from warehouse, query duration, or storage views. In this scenario, the best choice is ACCOUNT_USAGE.CORTEX_FUNCTIONS_USAGE_HISTORY because it is designed for reporting on AI function consumption over time and can be correlated with other ACCOUNT_USAGE views for broader governance and FinOps reporting. This matches Snowflake best practices: use service-specific usage views for cost attribution, and use QUERY_HISTORY only as a supplementary source to understand the SQL statements involved. Relevant Snowflake documentation includes the ACCOUNT_USAGE views for Cortex/Snowflake AI observability and the general guidance on using ORGANIZATION_USAGE or ACCOUNT_USAGE for cost monitoring and chargeback reporting.
- A. Correct.
Correct. Snowflake exposes dedicated ACCOUNT_USAGE telemetry for Cortex AI usage, including CORTEX_FUNCTIONS_USAGE_HISTORY, which is intended for monitoring usage and cost trends for supported AI functions. This approach supports time-based reporting and can be joined with other ACCOUNT_USAGE views for broader analysis by user, workload, or related objects. It aligns with the requirement to build a daily report that integrates with existing usage reporting.
- B. Incorrect.
Incorrect. WAREHOUSE_METERING_HISTORY tracks compute credits consumed by virtual warehouses. Cortex AI service charges are not simply inferred from the warehouse that ran the SQL. Relying only on warehouse metering would miss the dedicated AI service usage details and would not accurately attribute AI-specific consumption.
- C. Incorrect.
Incorrect. QUERY_HISTORY is useful for identifying which statements invoked AI functions, but it is not the authoritative source for AI cost reporting. Summing query timing fields does not translate to billable AI usage or credits. This is a common misconception because query history helps with operational troubleshooting, but Snowflake provides dedicated usage views for service billing analysis.
- D. Incorrect.
Incorrect. DATABASE_STORAGE_USAGE_HISTORY reports storage consumption, not AI service usage. Although some AI-related workflows may create persisted artifacts, Cortex function charges are not represented as storage growth in this view. Using storage history would not meet the requirement to track AI usage and cost by user and time range.