Oracle Brings Google TPUs, OpenAI's Jalapeño and Cerebras to OCI, and ChatGPT to Its Apps

Oracle and Google announced on October 27, 2026 that Google's Tensor Processing Units (TPUs) will be available through Oracle Cloud Infrastructure (OCI) in calendar year 2027. It is the first time Google's TPUs will be offered on a hyperscale cloud platform outside Google, the companies said in Oracle's release.

TPUs are Google's custom chips for training and running AI models, one of the few large alternatives to Nvidia GPUs among AI chips. At its Oracle AI World conference in Las Vegas, Oracle also said OpenAI's Jalapeño chip and Cerebras hardware are coming to OCI and unveiled a product suite built on ChatGPT and Codex, CNBC reports. The TPU deal lands one week after OpenAI's models arrived on Google Cloud in a private preview on October 21.

What did Oracle and Google announce?

Oracle and Google said on October 27, 2026 that Google TPUs will be available on OCI in calendar year 2027, ramping over a multi-year period to meet customer demand. Oracle will provide the power, cooling and data center networking to deploy the chips at scale inside its own infrastructure, according to the release.

The release pitches TPUs for large model training and for production inference in reasoning, agentic and long-context applications. It describes Google's eighth generation, with a TPU 8t chip for training and a TPU 8i chip for inference, but it does not say which TPU generation OCI will offer or at what price.

Bringing Google TPUs to OCI gives customers access to multiple leading AI compute architectures in one cloud.

Google has also been making TPUs easier to adopt: on October 20, 2026 it released TorchTPU, which runs PyTorch code natively on the chips.

Which other chips are coming to Oracle's cloud?

Oracle also said on October 27, 2026 that OpenAI's Jalapeño chip and Cerebras hardware will run on OCI, according to CNBC. A Cerebras spokesperson told CNBC the deal is material but that the company will not disclose its scale because it is in a quiet period.

OCI plans to offer the Cerebras service in 2027, the companies said, and Oracle did not say how much Cerebras hardware it will use, according to The Next Web. The move comes seven months after Amazon Web Services announced plans to bring Cerebras hardware into its data centers, CNBC notes.

Jalapeño is OpenAI's first custom accelerator, built with Broadcom for running large language models. When the companies unveiled it on June 24, 2026, Broadcom CEO Hock Tan said it would enable gigawatt scale data centers "with Microsoft and other partners beginning in 2026," naming no other partner. Oracle is now publicly tied to the chip as well, per CNBC. Neither Oracle's OpenAI release nor OpenAI's post on the partnership mentions the chip, and CNBC gives no timing for Jalapeño on OCI.

Nvidia stays at the center of OCI. Nvidia and Oracle said the same day that Vera Rubin NVL72 rack systems and Nvidia's Vera CPUs are now generally available on OCI, and that Oracle plans to scale its Vera Rubin clusters to multiple gigawatts, according to Nvidia's release. Thiagarajan told CNBC the new chips are not a pivot away from Nvidia, whose chips he called "the best, most fungible."

Hardware	Maker	Status on OCI	Timing	Source
Tensor Processing Units	Google	Announced by Oracle and Google	Calendar year 2027, multi-year ramp	Oracle
Jalapeño accelerator	OpenAI with Broadcom	Oracle said it is coming, per CNBC	Not disclosed	CNBC
Cerebras systems	Cerebras	Announced, scale not disclosed	2027	CNBC, The Next Web
Vera Rubin NVL72 and Vera CPUs	Nvidia	Generally available	Now, multi-gigawatt expansion planned	Nvidia
HGX Rubin NVL8 servers	Nvidia	Planned	Not disclosed	Nvidia

Why is Oracle adding so many chip suppliers?

Oracle frames the new chips as customer choice at a time when AI chips are still scarce. Thiagarajan told CNBC that chip scarcity remains an ongoing issue and that customers such as OpenAI benefit from having different accelerators to choose from.

OpenAI shows the scale involved: it trained and served GPT-6 Astra on an OCI cluster in Abilene, Texas with more than 100,000 Nvidia Blackwell GPUs, according to Nvidia.

The spending behind it is large. Oracle plans to spend $70 billion on capital projects in the fiscal year to May 2027, after $55.7 billion the year before (The Next Web). Oracle shares rose nearly 5% on October 27 and Cerebras stock climbed about 8% after the news, though Oracle shares are down more than 30% this year, partly on concerns about how it will finance its data center buildout, according to CNBC.

What changes for Oracle's application customers?

Oracle said on October 27, 2026 that it is building OpenAI models, ChatGPT and Codex into its applications through four offerings. Oracle co-CEO Clay Magouyrk shared the keynote stage with OpenAI CEO Sam Altman that day (CNBC).

According to Oracle's release, the suite includes:

• Oracle AI Workspace, which uses plugins in ChatGPT to let employees look up information and complete approved workflows across Oracle Fusion Applications, NetSuite, Oracle AI Database and OCI.
• Oracle AI Security Mesh, a managed service that uses OpenAI's cyber models to find and fix vulnerabilities across OCI, other clouds and on-premises systems, within a scope the customer authorizes.
• Oracle App Composer, which lets employees build applications with Codex while Oracle handles deployment on OCI, identity and access controls.
• Oracle AI SMB Bundle, a free-to-start package for small businesses with $12,000 in Oracle cloud credits each year, ChatGPT Business, Codex and NetSuite QuickStart.

Oracle is also a large OpenAI user itself: it reported 130,000 active ChatGPT users and more than 95,000 active Codex users, according to OpenAI.

What we don't know yet

• When Jalapeño will run on OCI, at what scale, and whether outside customers or only OpenAI will be able to use it.
• Which TPU generation OCI will offer in 2027, at what price, and in which regions.
• How much Cerebras hardware Oracle is buying, which the company has not disclosed.
• Release dates and pricing for Oracle AI Workspace, AI Security Mesh and App Composer.