Oracle (opens in new tab) on Tuesday announced an expanded collaboration with OpenAI that brings ChatGPT and Codex into Oracle (opens in new tab)’s business applications, CNBC (opens in new tab) reported. At Oracle (opens in new tab)’s AI World conference in Las Vegas, Mahesh Thiagarajan, president of Oracle (opens in new tab) Cloud Infrastructure, told CNBC (opens in new tab) that using ChatGPT “allows anybody to build an application, deploy it, manage it.”

Oracle (opens in new tab)’s release lists four offerings. Oracle (opens in new tab) AI Workspace is designed to let employees use ChatGPT to reach Oracle (opens in new tab) applications, data and approved workflows. Oracle (opens in new tab) AI Security Mesh, a managed service, uses OpenAI cyber models to help customers find and fix vulnerabilities. Oracle (opens in new tab) App Composer lets employees build applications with Codex, with Oracle (opens in new tab) managing deployment. The Oracle (opens in new tab) AI SMB Bundle is free to start and provides $12,000 in Oracle (opens in new tab) cloud credits a year for small businesses.

“OpenAI is transforming how enterprises solve business problems, from chat, to coding, to analytics, to autonomous agents,” Oracle (opens in new tab)’s Clay Magouyrk said in the release. OpenAI’s Greg Brockman, its president and co-founder, said: “More businesses are choosing OpenAI as we make it easier to put AI to work through the tools and systems they already use.”

Oracle (opens in new tab) also announced changes to the chips behind its cloud. Reuters reported that Oracle (opens in new tab) said it will begin offering Google’s tensor processing units, or TPUs, and CNBC (opens in new tab) reported it is the first time Google’s custom chips will run on another hyperscaler’s cloud. CNBC (opens in new tab) also reported that OpenAI’s Jalapeño chip and hardware from Cerebras Systems will run on Oracle (opens in new tab) Cloud Infrastructure. Reuters reported that Oracle (opens in new tab) plans to offer chips from SambaNova Systems, which is backed by Intel, as well as Cerebras. Both companies make chips for inference, the stage at which an AI system responds to user requests.

Reuters reported that Oracle (opens in new tab) and Nvidia said Oracle (opens in new tab) will expand to “multi-gigawatt scale” with Nvidia’s Vera Rubin servers. The companies did not disclose the deal’s value, though Nvidia CEO Jensen Huang has previously said Vera Rubin systems can generate up to $40 billion in revenue per gigawatt for Nvidia. Reuters also said Oracle (opens in new tab)’s Nvidia-based systems will skip some Nvidia networking chips in favor of Oracle (opens in new tab)’s own “Acceleron” chips.

Asked by CNBC (opens in new tab) whether the additions mark a move away from Nvidia’s GPUs, Thiagarajan said no, calling Nvidia’s chips “the best, most fungible.” A Cerebras spokesperson told CNBC (opens in new tab) the company would not disclose the scale of its deal because it is in a quiet period. Cerebras CEO Andrew Feldman told Reuters: “That’s a massive expansion of our customer base of tier-one customers.”

Oracle (opens in new tab) did not disclose how it plans to pay for the chip purchases, Reuters reported. During an earnings call last month, Oracle (opens in new tab) said customers are increasingly paying for chips and kept its capital spending target of about $90 billion to $95 billion for fiscal 2027. Oracle (opens in new tab) plans to hold an investor day this week.