Oracle integrates OpenAI and Google TPU at the core of its cloud

Oracle unveiled in Las Vegas a suite of tools combining OpenAI and its own applications, while adding Google’s TPUs to OCI. The goal: attract companies seeking more AI, more hardware choices, and greater control.

Oracle unveiled at Oracle AI World in Las Vegas a series of announcements that bring OpenAI, Google, and its own cloud closer together. On one hand, the company is adding ChatGPT and Codex to its application stack for enterprise customers; on the other, OCI will host Google's TPUs, as well as other AI chips like those from Cerebras and OpenAI's Jalapeño line [to be verified]. The message is clear: Oracle wants to become a hub for AI workloads, from software development to production inference. Sam Altman is also set to take the stage with Clay Magouyrk during the event.

OpenAI s’invite dans les applications Oracle, pas seulement dans les démonstrations

The new Oracle x OpenAI offering is not limited to a logo on a slide. Oracle detailed four services: Oracle AI Workspace, Oracle AI Security Mesh, Oracle App Composer, and Oracle AI SMB Bundle. Together, they cover office use, cybersecurity, application creation, and support for small businesses. The idea, according to Oracle, is to have OpenAI models work directly with existing applications, data, and enterprise controls.

Clay Magouyrk, CEO of Oracle, sums up the logic: “OpenAI is transforming how enterprises solve business problems, from chat, to coding, to analytics, to autonomous agents,” he said. Greg Brockman, president and co-founder of OpenAI, states that more companies are choosing OpenAI because it is becoming easier to connect AI to the tools they already use. The message is classic, but the product is more concrete than a trade-show partnership: Workspace aims to unify models, Oracle applications, and business data in a single interface, while App Composer wants to enable application creation with Codex without coding expertise.

Google TPU chez Oracle : le cloud se vend aussi au rayon silicium

Oracle and Google also announced that Google's Tensor Processing Units will be available via Oracle Cloud Infrastructure in 2027. According to their statement, OCI will be the first hyperscale cloud platform outside Google to offer these TPUs. Oracle provides the physical infrastructure — power, cooling, datacenter networking — to deploy these chips at scale. On paper, this allows customers to choose between multiple AI architectures within the same cloud, depending on their cost, performance, or latency needs.

Mahesh Thiagarajan, president of Oracle Cloud Infrastructure, summarized the idea in one sentence: “Different AI workloads have different compute needs,” he said. Amin Vahdat, from Google, added that the next generation of AI will require a broader diversity of computing. In this context, the arrival of TPUs in OCI is not a technical detail. It shows that public clouds no longer compete only on software or storage, but also on the chips themselves. The market loves to complicate things whenever it can.

Ce que changent les nouveaux services Oracle

Oracle AI Workspace is intended to serve as a unified working layer between ChatGPT, Oracle Fusion applications, NetSuite, Oracle AI Database, and OCI. In practice, an employee can start a task in ChatGPT, access approved data, then trigger actions in company systems with the appropriate permissions. Oracle cites cases such as customer order tracking, financial variance analysis, or employee onboarding across HR, IT, and finance.

Oracle AI Security Mesh is designed for controlled offensive security. The service uses OpenAI cyber models to test vulnerabilities in applications and infrastructures, including on other clouds and on-premises, but only within a perimeter authorized by the client. Oracle describes a fairly precise chain: system mapping, validated test plan, specialized agents, cross-verification, correction recommendations, then final control after modification. The goal is to prevent a security test from turning into a production incident — which would be unfortunate for a security offering.

App Composer takes a different direction: it targets employees capable of imagining a business tool without knowing how to code. The created application can then be deployed and operated by Oracle on OCI, with identity, permission, network, audit, and lifecycle management. Oracle also says it supports APEXlang, an in-house technology meant to offer the speed of generative development without producing opaque, hard-to-maintain code.

Une stratégie de plateforme, pas une addition de logos

Taken together, these launches tell a fairly clear strategy. Oracle wants to sell a stack where generative AI, business applications, security, and infrastructure are intertwined rather than connected by hastily cobbled APIs. Customers get more hardware choice with Google's TPUs, more automation with OpenAI, and a promise of centralized governance that remains very important for large enterprises. It remains to be seen, in the long run, if this integration will hold up when use cases move from demonstration to daily operation.

Key points

  • Oracle AI World served as the stage for announcements on October 27, 2026.
  • Google TPUs will arrive on OCI in 2027, according to Oracle and Google.
  • Oracle launches four offerings around OpenAI and its applications.
  • App Composer is already in limited availability.
  • OCI becomes an access point to multiple AI architectures.

By the numbers

  • 4 offers — Oracle AI Workspace, Security Mesh, App Composer, SMB Bundle.
  • 12,000 dollars — annual cloud credits included in the AI SMB Bundle.
  • 2027 — announced availability year for Google's TPUs on OCI.
  • 250 dollars — maximum credit per workspace for the NetSuite QuickStart extension.

To read