Bringing the borderless Lakehouse into the agentic era with innovations for Gemini

Google Cloud introduced a native integration between its borderless Lakehouse and Gemini, allowing AI agents to directly access and reason over enterprise data stored across multiple clouds, SaaS platforms, and on-premises systems without moving it. The integration, currently in preview, supports data sources like Databricks Unity Catalog and Snowflake Horizon through Lakehouse catalog federation, enabling natural language queries and agent-driven insights across cloud boundaries. Business users can now query records spanning clouds while AI agents ground their reasoning and generate code from the same data, accelerating AI adoption and simplifying access to critical information.

The borderless Lakehouse now supports Apache Iceberg V3 specifications in BigQuery, including binary deletion vectors and sort orders, improving efficiency for row-level updates and deletes by eliminating the need to rewrite entire data files. BigQuery’s execution engine has been upgraded with a FlatBuffers-based Parquet reader and parallelized I/O to reduce CPU overhead and serialization bottlenecks, narrowing the performance gap between open formats and native storage. Cross-Cloud Interconnect provides dedicated, private connections between Google Cloud and AWS or Azure at 10 Gbps or 100 Gbps per connection, ensuring low-latency, predictable performance for agent workloads.

New features include graph support for Iceberg in BigQuery for analyzing relationship networks like supply chains, and AlloyDB lakehouse federation for querying live database records alongside historical warehouse data without copying information. The Lakehouse runtime catalog now offers a standard Hive Metastore-compatible endpoint, enabling engines like Apache Spark and Trino to discover metadata without architectural changes. Knowledge Catalog integrates with the Lakehouse to index distributed data assets, providing precise business metadata and semantic definitions for agents to reason over unstructured data in Cloud Storage.

Customer use of Apache Iceberg has tripled since early 2025, with enterprises reporting significant impact: Etsy unified three petabytes of data, accelerating pipeline queries by 60%; Commerzbank transformed distributed data into governed, agent-ready access; and Roundel rebuilt its Target Product Ads stack using the borderless Lakehouse. Organizations can migrate from legacy Dataproc Metastore to the serverless Lakehouse runtime catalog via an automated, zero-downtime path, preserving existing configurations.