Larry Ellison, Oracle's CEO and co-founder, has said the owners of high-level private corporate and government data will be the landlords of the AI landscape.

Oracle’s recent moves indicate a significant pivot toward what Ellison calls the 'second phase' of the AI revolution: moving beyond public internet data to Private Enterprise Data (PED).

The Oracle Federal Forum in Washington, D.C., on March 31, 2026, officially launched the Oracle AI Data Platform for Federal Government. 

This platform is specifically designed to bridge the gap between high-level generative AI and the highly sensitive, esoteric data held by government agencies.  

Oracle is betting on the concept of Isolated Cloud Regions which are air-gapped environments physically disconnected from the public internet.  

These regions are designed to host Secret and Top Secret classified workloads.  

By keeping the infrastructure entirely local and disconnected, Oracle addresses the data sovereignty concerns often preventing federal agencies from adopting cloud-based AI.  

These can be deployed in dedicated government data centers or portable units (like the Roving Edge Device) for tactical use in the field. 

According to Ellis, Oracle's platform enables agentic applications. Unlike standard chatbots that simply summarize text, this architecture is built for Multi-Step Reasoning.  

The platform pulls from government hives—disparate databases like OCI, Oracle Autonomous AI Database, and OCI Enterprise AI.  

In a defense context, the AI is designed to analyze real-time mission data to suggest tactical maneuvers or identify perimeter breaches.  

For agencies like the Treasury or HHS, it’s being used to automate complex financial audits and reasoning through massive datasets to identify anomalies that a human might miss.

Larry Ellison’s stance is that foundational models trained only on public data (like GPT or Gemini) are becoming commoditized. He argues that the true "AI race" will be won by whoever can securely unlock PED.  

In January of 2026, Ellison stated training AI models on public data is the fastest growing business in history but AI models drawing from private data will be even more lucrative.

Because a vast majority of the world's high-value enterprise and government data already sit in Oracle databases, the company is positioning its AI Data Platform as the essential security layer allowing agencies to query their own private data without ever exposing it to the public web.  

This shift suggests a move away from "all-knowing" general AI toward highly specialized, hyper-secure "Sovereign AI" tailored for the specific mission requirements of the federal government.