Semantic Aggregation Over Millions of Rows in BigQuery with AI.AGG
Google, Thursday, November 5th, 2026
BigQuery's AI.AGG, in preview, runs natural-language semantic aggregation over millions of rows in a single SQL query.
AI.AGG is a preview function in BigQuery GoogleSQL that performs multi-level semantic aggregation over large volumes of text and multimodal data.
It addresses LLM context window limits that previously forced teams to build external map-reduce pipelines for summarizing millions of rows.
BigQuery validates row sizes, batches rows to fit within the chosen Gemini endpoint's token limits, applies the prompt to batches in parallel via Vertex AI, and recursively combines intermediate summaries.
This lets data teams run natural-language instructions over data such as logs, transcripts and reports in one query.