Google and Oracle connect Gemini to business databases, with key tools still in preview

Google Cloud and Oracle are promising AI agents that can work directly with business data, but the pieces are arriving at different stages. At Oracle AI World in Las Vegas on October 27, the companies announced deeper database and Gemini integrations, mixing available tools with previews and a future government-certification effort.

The new Gemini connector for Oracle AI Database@Google Cloud is in preview. It lets business users and developers ask natural-language questions about their Oracle databases from inside the Gemini app. Google says responses are grounded in the Oracle data each user is authorized to access.

Behind that interface is Knowledge Catalog, whose Oracle integration is also in preview. It catalogs database structures and business metadata so systems can discover information, track where it came from and apply access controls. The connector uses that catalog to find relevant data and ground its answers within the user's permissions, Google says.

Another foundation is generally available: the Remote Model Context Protocol server and ExecuteSQL Data API. They provide a standard interface through which agents, Google Cloud services and applications can run SQL, the language used to query databases. Google says credentials are resolved through Secret Manager, while its identity and access-management policies govern access.

The catalog and database-access capabilities were first announced at Google Cloud Next '26. They are supporting infrastructure for this partnership update, not entirely new inventions at Oracle AI World.

Oracle in Gemini Enterprise includes conversational analytics, cross-cloud database connectors and an Oracle AI Database Agent. Google describes more than 135 prebuilt operational and developer skills, intended to help coordinate agents across JD Edwards, E-Business Suite, NetSuite and Oracle Fusion without custom SQL or middleware.

The connection also runs in the other direction. Google says Gemini models are embedded in Oracle Fusion AI Agent Studio, Fusion Claw, Oracle Fusion Applications and NetSuite to support automated work across transactional systems. Partners including IBM, Infosys and TCS are contributing operational skills, connectors and industry deployment blueprints.

The data-sharing changes offer different ways to connect Oracle records with BigQuery and AI workloads. A new bidirectional lakehouse preview lets Oracle Autonomous AI Database and Google's borderless Lakehouse both query and write to shared Apache Iceberg tables in Google Cloud Storage. Google says this avoids copying the data or maintaining a separate transfer pipeline.

For customers that want Oracle changes reflected inside BigQuery, the companies describe streaming through OCI GoldenGate. It reads changes from Oracle's transaction logs and sends them over Google Cloud private networks into BigQuery tables. Google claims sub-second freshness without adding overhead to transactional workloads; that is a company performance claim.

A third option, BigQuery query federation with Oracle, is in preview. It lets BigQuery run queries against live Oracle databases without first moving their records. Unlike shared analytical storage, this approach can affect the Oracle transactional system depending on the query, according to Google's architectural comparison.

Where access is widening

• Regional availability: Oracle AI Database@Google Cloud is now live in Turin, Italy, and Querétaro, Mexico, adding locations for customers with local data-residency needs.
• Government workloads: Oracle plans to pursue FedRAMP High certification for the service. The proposed setup uses dedicated infrastructure within Google Cloud Assured Workloads. This is a certification plan, not an announcement that the service has obtained certification.