Oracle Expands OpenAI Partnership and AI Chip Lineup

Oracle said on the 27th local time at Oracle AI World 2026 in Las Vegas that it would expand its work with OpenAI across enterprise AI products and add more types of AI chips to Oracle Cloud Infrastructure (OCI). The plan combines OpenAI models and development tools with Oracle applications and data, while working with several companies to meet growing demand for AI computing.

At the keynote, Oracle co-CEO Clay Magouyrk announced plans for collaboration with OpenAI and other major technology companies. Oracle and OpenAI have also worked together on AI infrastructure, including the Stargate data center in Abilene, Texas. The companies introduced Oracle AI Security Mesh, which uses OpenAI models; Oracle AI Workspace, which connects ChatGPT to Oracle business systems; and Oracle App Composer, a tool for developing AI applications.

Magouyrk said AI now writes more than 70% of the software Oracle produces, and that the pace of change is accelerating. He said Oracle built App Composer to address the need to deploy applications safely and under control as more developers build apps. OpenAI CEO Sam Altman said AI had advanced in a short time from chatbots to coding agents and then to systems that create innovations themselves. He said people would be able to use better software and higher-quality services at lower cost.

Oracle plans to make Snowflake’s AI Data Cloud a standard OCI service, with the goal of reducing data transfers between clouds and the costs they bring.

Oracle also plans to broaden the AI accelerators available on OCI. Magouyrk said Oracle had begun delivering its first Rubin systems to customers and described Vera Rubin as the most powerful system the company had deployed to date. Oracle plans to add Google’s TPUs, specialized chips for AI computing, and other accelerators to OCI. It will be the first cloud provider outside Google to offer TPUs, with deliveries set to begin in 2027 and expand in stages. Oracle also plans to offer Cerebras’ high-speed inference service on OCI that year.

Oracle also plans to bring Halapeno, an inference chip jointly developed by OpenAI and Broadcom, to OCI. Altman said AI is good at chip design and that AI assistance enabled chip development at an extraordinarily fast pace.

Mahesh Thiagarajan, who leads OCI, said accelerators can be bottlenecked by network, storage and data-delivery performance. Oracle aims to address the issue by integrating the entire cloud and enabling customers to get more work done for every dollar they spend.

Oracle said on the 27th local time at Oracle AI World 2026 in Las Vegas that it would expand its work with OpenAI across enterprise AI products and add more types of AI chips to Oracle Cloud Infrastructure (OCI). The plan combines OpenAI models and development tools with Oracle applications and data, while working with several companies to meet growing demand for AI computing.

At the keynote, Oracle co-CEO Clay Magouyrk announced plans for collaboration with OpenAI and other major technology companies. Oracle and OpenAI have also worked together on AI infrastructure, including the Stargate data center in Abilene, Texas. The companies introduced Oracle AI Security Mesh, which uses OpenAI models; Oracle AI Workspace, which connects ChatGPT to Oracle business systems; and Oracle App Composer, a tool for developing AI applications.

Magouyrk said AI now writes more than 70% of the software Oracle produces, and that the pace of change is accelerating. He said Oracle built App Composer to address the need to deploy applications safely and under control as more developers build apps. OpenAI CEO Sam Altman said AI had advanced in a short time from chatbots to coding agents and then to systems that create innovations themselves. He said people would be able to use better software and higher-quality services at lower cost.

Oracle plans to make Snowflake’s AI Data Cloud a standard OCI service, with the goal of reducing data transfers between clouds and the costs they bring.

Oracle also plans to broaden the AI accelerators available on OCI. Magouyrk said Oracle had begun delivering its first Rubin systems to customers and described Vera Rubin as the most powerful system the company had deployed to date. Oracle plans to add Google’s TPUs, specialized chips for AI computing, and other accelerators to OCI. It will be the first cloud provider outside Google to offer TPUs, with deliveries set to begin in 2027 and expand in stages. Oracle also plans to offer Cerebras’ high-speed inference service on OCI that year.

Oracle also plans to bring Halapeno, an inference chip jointly developed by OpenAI and Broadcom, to OCI. Altman said AI is good at chip design and that AI assistance enabled chip development at an extraordinarily fast pace.

Mahesh Thiagarajan, who leads OCI, said accelerators can be bottlenecked by network, storage and data-delivery performance. Oracle aims to address the issue by integrating the entire cloud and enabling customers to get more work done for every dollar they spend.