Google Cloud announced a new Gemini integration with the borderless Lakehouse, currently in preview. Through Lakehouse catalog federation, Gemini can directly access data sources such as Databricks Unity Catalog and Snowflake Horizon, allowing business users to query records across clouds with natural language and allowing AI agents to ground their reasoning on that data without moving it. The announcement is aimed at enterprises whose AI usage is shifting from experimental chatbots to autonomous agents that need to reason over data spread across clouds, SaaS platforms, and on-premises systems.

Alongside the preview, several borderless Lakehouse features introduced earlier this year are now generally available: Cross-Cloud Interconnect and cross-cloud caching for AWS and Azure; BigQuery support for Apache Iceberg V3; read-write interoperability on Iceberg tables for BigQuery and Managed Service for Apache Spark; AlloyDB lakehouse federation; bi-directional caching with AWS Glue, Databricks Unity Catalog, and Snowflake Horizon Catalog via Lakehouse runtime federation; integration with Knowledge Catalog; zero-copy data integration with Salesforce, SAP, Workday, and Confluent Tableflow; and synchronization of Databricks Delta Lake assets through the Lakehouse console using Dataflow.

The post argues that agent context windows typically ingest hundreds of kilobytes at a time, while enterprise workflows may join petabytes of data across clouds. Using MCP for such joins is limited to megabyte scale because of context windows and payload size. The borderless Lakehouse avoids that limit by integrating at each layer of the data stack: at the networking layer, Cross-Cloud Interconnect provides a dedicated private connection between Google Cloud and AWS or Azure at 10 Gbps or 100 Gbps per connection, and cross-cloud caching serves repeat reads from inside Google Cloud so agents can query data in other clouds without migrating it.

Storage and query-engine work focuses on open formats. BigQuery now supports Apache Iceberg V3, including binary deletion vectors and sort orders, which handle row-level updates and deletes without rewriting entire data files. A new FlatBuffers-based Parquet reader and parallelized I/O reduce CPU overhead and serialization bottlenecks, narrowing the performance gap between open formats and native storage. Read-write interoperability on Apache Iceberg tables for BigQuery and Managed Service for Apache Spark lets data teams work off the same tables in Google Cloud Storage. BigQuery graph support for Iceberg enables relationship analysis, such as supply chains, directly on open table datasets. For operational data, AlloyDB lakehouse federation lets agents query live database records and historical warehouse data together from a single connection, without copying information.

At the metadata layer, the Lakehouse runtime catalog, based on the Apache Iceberg REST Catalog, provides bi-directional federation with AWS Glue, Databricks Unity Catalog, and Snowflake Horizon Catalog. A new Hive Metastore-compatible endpoint allows engines like Apache Spark, Trino, and Presto to discover metadata without architectural changes. Knowledge Catalog integration indexes distributed data assets, including unstructured data in Cloud Storage, to provide business metadata and semantic definitions for agents while controlling token costs. Zero-copy SaaS integrations include Salesforce, SAP, Workday, and the newly GA Confluent Tableflow, which syncs Iceberg tables into the Lakehouse runtime catalog so they appear directly in BigQuery without copying or ETL. Databricks-centric repositories can import, transform, and synchronize Delta Lake assets into the Lakehouse through the console using Dataflow.

Customer examples cited in the post: Etsy unified three petabytes with borderless Lakehouse, joining data in place to accelerate pipeline queries by 60%; Commerzbank transforms raw distributed data into governed, agent-ready access; Hightouch integrates with borderless Lakehouse for marketing activation; and Roundel rebuilt its Target Product Ads technology stack on it. For existing Dataproc Metastore users, Google Cloud offers an automated, zero-downtime migration path to the serverless Lakehouse runtime catalog, preserving schemas and lineage.