Neo4j Virtual Graphs (released June 2026 as a preview) let you query BigQuery data using Cypher and visualize it as a graph without any ETL. The post walks through setting up the IMDB dataset (~100M rows) in BigQuery, creating a graph model in Neo4j Aura, and running both a native Neo4j instance and a Virtual Graph instance side by side. It demonstrates finding the Bacon number using Cypher's SHORTEST path on native Neo4j and Dijkstra's algorithm via GDS Sessions on the Virtual Graph. Performance comparisons show both approaches complete fixed-length traversal queries in seconds for this dataset. The post also covers inspecting the generated SQL via EXPLAIN, optimizing the graph model with dedicated labels and relationship types, and using BigQuery's native GQL graph support as an alternative.
