Snowflake Cortex is a suite of fully managed AI and ML services within the Snowflake Data Cloud, accessible via SQL and Python. It includes seven key capabilities: Cortex Agents (orchestrating structured and unstructured data queries), Cortex AI Functions (serverless SQL/Python AI tasks like summarization and classification), Cortex Analyst (natural language to SQL), Cortex Fine-Tuning (PEFT-based model customization), Cortex Search (hybrid vector and keyword search for RAG), Snowflake Copilot (in-UI SQL assistant), and Snowflake Intelligence (natural language data exploration with visualizations). Pricing for Cortex Search involves virtual warehouse compute, embedding token compute, serving compute (billed per GB/month even when idle), storage, and cloud services compute, with worked examples provided. Key limitations include unpredictable token-based costs, lack of native resource monitors, idle service charges, limited observability, developer experience gaps, notebook collaboration restrictions, and Git integration friction. The post concludes with a promotion of Finout's FinOps platform for granular Cortex cost attribution and anomaly detection.
