SnowPro Specialty: Gen AI Question 255
Single answerSnowflake Service Consumption TableA Snowflake administrator needs to show monthly chargeback for a GenAI proof of concept that uses Snowflake Cortex features across several teams. Finance wants a report that identifies which AI service generated consumption and lets them aggregate usage by day and by service type from Snowflake-provided telemetry. Which approach should the administrator use?
- A
Query the Snowflake Service Consumption Table and aggregate rows by usage date and AI service dimensions to attribute Cortex-related consumption.
- B
Query WAREHOUSE_METERING_HISTORY because all Cortex AI service usage is billed only as warehouse credits, and no service-level telemetry is available.
- C
Use QUERY_HISTORY alone and sum query execution time, because AI service charges are derived directly from SQL runtime rather than service consumption records.
- D
Query the ACCOUNT_USAGE.METERING_DAILY_HISTORY view, because it exposes detailed per-feature AI service identifiers needed for chargeback reporting.
Show answer and explanation
Correct answer: A
Explanation
The best answer is to use the Snowflake Service Consumption Table for service-level chargeback analysis. In GenAI scenarios, Snowflake-managed capabilities such as Cortex may produce consumption that is distinct from standard virtual warehouse usage. For finance and operational reporting, administrators should use Snowflake-provided service consumption telemetry to group usage by date and service dimensions rather than trying to infer costs from warehouse metering or query runtime. This aligns with Snowflake best practices of using ACCOUNT_USAGE and organization/account telemetry views or tables for the specific billing domain being analyzed: warehouse views for warehouse spend, and service consumption telemetry for managed-service usage. Candidates should recognize that service-level attribution is the key requirement in the scenario.
- A. Correct.
Correct. The Snowflake Service Consumption Table is intended to expose service-level consumption telemetry, including Snowflake-provided services such as AI/ML-related services, so administrators can analyze usage and cost attribution beyond warehouse metering alone. For a Cortex chargeback scenario, aggregating service consumption by date and service dimensions is the appropriate method when finance needs reporting based on Snowflake-managed service usage.
- B. Incorrect.
Incorrect. This reflects a common misconception that all GenAI-related charges appear only as warehouse consumption. Snowflake Cortex and other managed services can generate service consumption that is tracked separately from warehouse metering. WAREHOUSE_METERING_HISTORY is useful for warehouse credit analysis, but it does not provide the needed service-level AI consumption detail for this scenario.
- C. Incorrect.
Incorrect. QUERY_HISTORY can help identify who ran queries and when, but it is not the authoritative source for service consumption billing attribution for Snowflake-managed AI services. Execution time does not reliably map to service charges, especially for managed service consumption such as Cortex features.
- D. Incorrect.
Incorrect. ACCOUNT_USAGE.METERING_DAILY_HISTORY provides account-level daily metering summaries, but it is not the best source when the requirement is to identify which AI service generated the usage at a more specific service-consumption level. The scenario explicitly asks for Snowflake-provided telemetry that distinguishes service type, which is what the Service Consumption Table is intended to support.