Google is expanding its Agentic Data Cloud, a platform it describes as a System of Action for running production agents against enterprise data. The new releases focus on activating data across clouds and SaaS systems, grounding agents in governed business context, and reworking practitioner tooling. The company also notes that Apache Iceberg data under management has more than tripled over the past year.

The borderless Lakehouse, built on the open Apache Iceberg format, now lets Gemini operate directly on data across clouds. Business users can query records in natural language while agents use the Lakehouse for grounding. Cross-cloud interconnect and caching for AWS and Azure are generally available, offering private, secure connections with zero variable egress fees; remote data blocks are cached locally in Google Cloud. Customers can also use open Iceberg catalog federation with Databricks Unity, Snowflake Horizon, and AWS Glue, plus zero-copy integrations with SAP, Workday, and Salesforce. Knowledge Catalog integration indexes distributed and unstructured data in cloud storage to give agents context while controlling token costs. A quoted Delivery Hero talabat engineer says implementing the borderless Lakehouse to federate single-copy Apache Iceberg data from Amazon S3 tables to BigQuery reduced data-movement costs by about 40% and made streaming transactional events queryable across both clouds within minutes.

New AlloyDB integrations link transactional and analytical data. Direct Iceberg access in AlloyDB lets you query Iceberg tables in cloud storage without ETL or copies and join live operational data with historical lakehouse data in a single PostgreSQL query. Tables registered in the Lakehouse runtime catalog appear in Knowledge Catalog, so agents get the same governed context. Google says AlloyDB direct Iceberg access is up to 7x faster on aggregation queries than Amazon’s comparable PostgreSQL-compatible service. Spanner Omni is now generally available for hybrid agentic applications across on-premises, edge, and multi-cloud environments.

For grounding, Knowledge Catalog is Google Cloud’s universal context engine. New self-learning enrichment, in preview, mines unstructured sources like Google Workspace, GitHub, Jira, Confluence, and SharePoint to close context gaps, and packages domain knowledge into reusable Open Knowledge Format (OKF) context sets. A new conversational studio experience lets data stewards explore context, visualize the data estate, and review and approve enrichment updates with versioning and rollback. Knowledge Catalog is now on by default in Gemini, in preview, grounding data questions in governed business context under each user’s access permissions. Google’s internal benchmarks claim Knowledge Catalog makes AI agents nearly 4x more accurate, with up to 33% lower latency and up to 75% fewer tokens. Semantic models from BigQuery, Looker (LookML), Databricks, Snowflake, dbt, and SAP are now generally available without moving data, so agents apply the same governed formulas as business teams. Metadata discovery and federation extends to Oracle AI Database@Google Cloud, Snowflake, Databricks Unity Catalog, and third-party catalogs like Atlan and Collibra.

Practitioner experiences are being repositioned around agents. In Gemini, data and ML engineers can describe outcomes in plain language, and Gemini generates PySpark code, provides notebooks, trains models, and troubleshoots pipelines. Business users can generate operational reports from natural language using BigQuery and Knowledge Catalog, with saved reports running without token costs. Data Agent Kit is generally available, integrating Google Data Cloud tools into IDEs and CLIs such as VS Code, Claude Code, and Antigravity, using native MCP tooling to let agents generate, tune, and test pipelines.

Infrastructure for agent scale includes AlloyDB Express, a preview edition with sub-2-second provisioning, scale-to-zero, private IP, observability, and a small footprint. PostgreSQL for agents in AlloyDB, in preview, supports thousands of collaborating agents with workload isolation, sub-millisecond I/O, terabit-per-second aggregated scan throughput, and over 3 million queries per second, with compute scaling based on active reasoning sessions. Managed Service for Apache Spark extends price-performance advantage by over 3x through Lightning Engine, with Spark 4.0 improvements. BigQuery queries are over 40% faster with over 50% lower processing cost, BigQuery Graph is generally available for petabyte-scale graph analytics on relational tables, and TimesFM 3.0 and TabFM are preview models running natively in BigQuery. The post notes these numbers come from Google’s product claims, several features are in preview, and the overall pitch is that enterprises like PayPal, Siemens, Telus, Vodafone, and Walmart are already using the platform.