Oracle Puts Its Entire Application Stack Inside ChatGPT

The latest:

Oracle has built a single entry point to all of its applications inside ChatGPT, letting users pose a business question and receive answers drawn from connected Oracle tools and data. Karan Batta, senior vice president of Oracle Cloud Infrastructure, described the product, Oracle AI Workspace, to The Wall Street Journal Leadership Institute ahead of its unveiling at Oracle AI World 2026 in Las Vegas.

Details:

• How it works: Oracle said users can describe a business goal in plain language, and plugins inside ChatGPT assemble the relevant tools and information to carry it out. The company said the move makes the full breadth of its application capabilities agent-ready, rather than requiring users to enter each product separately.
• The enterprise angle: Batta said the partnership is OpenAI’s coming out party to be enterprise-ready, and that the announcement is what allows the model maker to demonstrate it. He tied the timing to competitive pressure from Anthropic, which recently filed an S-1.
• Altman onstage: OpenAI chief executive Sam Altman, appearing with Oracle chief executive Clay Magouyrk at Tuesday’s conference, said the company’s push into the enterprise has produced a number of surprises. He did not detail what those surprises were.
• The security product: Oracle and OpenAI are co-engineering a cybersecurity offering called Oracle AI Security Mesh, powered by OpenAI’s Daybreak models, Batta said. He described it as a managed service built to help enterprises find and fix vulnerabilities across their software.
• Patching problem: Batta said some companies patch software frequently and build harnesses for AI models while others do not. Altman said it may be too hard to patch every bug across the supply chain, and added that he is shocked by how many vulnerabilities AI systems are uncovering.
• Chips, part one: Oracle said Google’s Tensor Processing Units will be available on its cloud infrastructure, which the company said marks the first time Google TPUs are offered on a major cloud platform outside Google’s own.
• Chips, part two: Cerebras Systems inference hardware will become available on Oracle’s cloud platform starting in 2027, the company said. Oracle is also deploying OpenAI’s Jalapeño AI chip, without a disclosed timeline for that rollout.
• The supply logic: Batta said supply is the decisive factor as the chip supply chain tightens, and that giving customers choice is comparatively easy. Oracle framed the wider lineup as a way to offer flexibility while AI compute remains constrained.
• The adoption gap: Enterprises have struggled to link their corporate data and applications together and then expose them to AI systems. Connecting Oracle applications and data through ChatGPT could be pivotal in helping businesses actually use AI’s functionality.
• Also: Oracle will also offer Google TPUs outside Google’s own cloud for the first time.

Between the lines:

Both halves of the announcement point the same way: Oracle is positioning itself as the layer enterprises buy instead of building their own AI plumbing. Hosting a rival’s TPUs and a startup’s inference chips alongside OpenAI silicon signals that Oracle values being the place compute is bought over loyalty to any one supplier, at a moment Batta says supply decides outcomes.

What’s next

Cerebras hardware is due on Oracle’s cloud in 2027. Watch Anthropic’s S-1 process for the enterprise comparison Batta invoked, and Microsoft and Alphabet earnings, reported after the bell, for cloud demand signals.