Oracle Corporation announced on April 24, 2026, a significant expansion of its multi-cloud partnership with Google Cloud, introducing the Oracle AI Database Agent for Gemini Enterprise. Simultaneously, industry reports confirmed that Oracle has canceled a substantial order for Nvidia GB300 NVL72 racks previously placed with Super Micro Computer, Inc. These developments highlight a dual strategy of deepening software-level AI integrations while re-evaluating hardware supply chain partnerships.
The new Oracle AI Database Agent for Gemini Enterprise is designed to allow corporate customers to connect their private data stored in Oracle Databases directly to Google’s Gemini large language models. According to Oracle Executive Vice President of Database Autonomous Services, Juan Loaiza, the integration utilizes retrieval-augmented generation (RAG) to provide enterprise-grade accuracy and security. This move expands on the Oracle Database@Google Cloud initiative launched in 2024, which allows customers to run Oracle database services within Google Cloud data centers. The service is now available across 15 global regions, including North America, Europe, and Asia-Pacific.
The AI Database Agent automates the process of converting natural language queries into SQL, enabling non-technical users to extract insights from complex datasets. Oracle stated that the tool includes built-in security protocols to ensure that proprietary data used to ground AI responses remains within the customer’s controlled environment and is not used to train Google’s foundation models. This deployment follows Oracle’s broader strategy to offer its database technology across all major hyperscale cloud providers, including Microsoft Azure and Amazon Web Services.
In a separate development, internal sources and supply chain analysts reported that Oracle has terminated a multi-billion dollar procurement agreement with Super Micro Computer for Nvidia-based hardware. The canceled order specifically involved the GB300 NVL72 liquid-cooled racks, which are built on Nvidia’s Blackwell architecture. This order was originally intended to bolster Oracle’s Cloud Infrastructure (OCI) capacity for generative AI training and inference. While Oracle has not issued an official statement regarding the cancellation, the move coincides with reports of shifting procurement toward other original design manufacturers and internal engineering adjustments.
The shift in hardware procurement comes as Oracle continues to invest heavily in its own data center infrastructure. In its most recent fiscal report, Oracle noted capital expenditures of 6.9 billion dollars for the quarter, largely driven by AI demand. The company currently operates 85 customer-facing cloud regions, with another 32 under construction. Oracle CEO Safra Catz previously stated that the company aims to reach 65 billion dollars in annual revenue by fiscal year 2026, a goal supported by the rapid scaling of its AI-integrated cloud services.