Oracle Will Rent Out Google’s TPUs on Its Cloud From 2027

Google’s Tensor Processing Units (TPUs) will be available through Oracle Cloud Infrastructure (OCI), Oracle and Google said in a joint release dated 27 October 2026, at Oracle AI World in Las Vegas. The release says this marks “the first time Google TPUs will be offered on a hyperscale cloud platform outside Google.”

The release says Google TPUs on OCI “will be available in calendar year 2027, ramping over a multi-year period to meet customer needs.” It gives no capacity, price, region or customer names.

Short explainer · Google TPUs on OCI

A vertical news Short narrated from the published text of this piece and related FourWeekMBA pieces.

Business Pill · Distribution advantage

An evergreen Business Engineer explainer on distribution advantage. It explains the general idea and does not discuss Oracle or Google.

The key insight: As we read it, the joint release makes OCI the first hyperscale cloud outside Google to offer Google TPUs, with Oracle supplying the power, cooling and networking.

Update, 28 Oct: OpenAI products can now be bought through Oracle. In a 27 October post, OpenAI said “Oracle customers in the U.S. can purchase access to the OpenAI API, Codex, and ChatGPT Work through Oracle Marketplace, using existing procurement processes and eligible Oracle Universal Credits.”

The post says Oracle “reported 130,000 active ChatGPT users and more than 95,000 active Codex users.”

It also lists Oracle offerings built with OpenAI models and tools: Oracle AI Security Mesh, which uses OpenAI cyber models to find and fix vulnerabilities; the Oracle AI SMB Bundle, which “combines ChatGPT Business and Codex with Oracle applications and recurring Oracle Cloud Infrastructure credits”; and Oracle App Composer, for building apps with Codex connected to approved Oracle data.

What Oracle and Google Announced

Under the announcement, OCI “will provide the power, cooling, and datacenter networking to deploy Google TPUs at scale within its infrastructure,” according to the release. Customers will be able to choose Google TPUs within OCI for training, post-training and production inference, the companies said.

“Bringing Google TPUs to OCI gives customers access to multiple leading AI compute architectures in one cloud,” said Mahesh Thiagarajan, president of Oracle Cloud Infrastructure, in the release.

“The next generation of AI will require a diverse range of compute,” said Amin Vahdat, senior vice president and chief technologist, AI & Infrastructure, at Google, in the same release.

Compute performance per pod of Google's TPU 8t training chip against the previous TPU generation, as Google st
Compute performance per pod of Google’s TPU 8t training chip against the previous TPU generation, as Google states it in the Oracle and Google joint release of 27 October 2026. Google’s own claim, not an independent benchmark.
Google TPUs on Oracle's cloud: the first hyperscale cloud outside Google to offer them; TPU 8t at nearly 3x th
Google TPUs on Oracle’s cloud: the first hyperscale cloud outside Google to offer them; TPU 8t at nearly 3x the previous generation’s compute per pod; Google’s eighth TPU generation. Source: Oracle and Google joint release, 27 Oct 2026.

What Was Posted on X

In a clip posted on X by @firesidealpha, Sam Altman says “AI is really good at chip design” and that OpenAI made a chip called Jalapeño “that we’re deploying in OCI”, which Altman calls “the best AI chip in the world for inference”, built “faster than would have ever been imaginable without AI”. The comparison is Altman’s own claim, not an independent benchmark.

CNBC posted a report that Google is altering its cloud playbook, enlisting partners to sell more of its AI chips.

TestingCatalog posted that OpenAI extends its partnership with Oracle to bring Oracle apps into ChatGPT, that Oracle users in the US can buy API, Codex and ChatGPT Work via Oracle Marketplace, and that alongside Oracle AI Workspace, Oracle is announcing Oracle AI Security Mesh, Oracle AI SMB Bundle and Oracle App Composer.

Google Cloud’s own account reposted the partnership at 17:00 UTC on 8 November: Google TPUs come to Oracle Cloud Infrastructure in 2027, and the post says OCI will be the first hyperscaler outside Google to offer TPUs. It restates the plan this piece reported, with no new figures.

Which TPUs

TPU 8t compute per pod against the previous TPU generation, as Google states it; TPUs on OCI from 2027. Source
TPU 8t compute per pod against the previous TPU generation, as Google states it; TPUs on OCI from 2027. Source: Oracle and Google joint release, 27 Oct 2026.

The release describes Google’s eighth-generation TPU as two chips: TPU 8t for training and TPU 8i for inference. It says TPU 8t delivers “nearly 3x the compute performance per pod over the previous generation,” while TPU 8i “is designed to handle the iterative work of many specialized agents.” These are Google’s own figures, not an independent benchmark.

The release does not say which TPU generation OCI will offer first, or whether Oracle will buy the chips or Google will supply them as a service inside Oracle’s data centres.

Update, 27 Oct: more chips on OCI. Google’s TPUs were not the only accelerator Oracle announced for OCI on 27 October. Cerebras Systems and Oracle announced plans to make Cerebras’ fast inference available on OCI, and Cerebras said Cerebras Inference on OCI “is planned to be available in calendar year 2027.”

CNBC reported that OpenAI’s Jalapeño chip will also run on OCI, and that a Cerebras spokesperson said the Cerebras deal is material but that the company won’t disclose its scale because it is in a quiet period. Asked by CNBC whether the additions mark a pivot away from Nvidia GPUs, Oracle’s Mahesh Thiagarajan said no, calling Nvidia chips “the best, most fungible.”

Reuters’ headline on the day reads “Oracle expands chip supply deal with Nvidia, adds Google and Cerebras offerings”; we could not read the article’s text, so it is cited for its headline only.

Update, 27 Oct: share prices. By 12:57 p.m. New York time, Oracle shares were at $131.25, up about 4.1% from Monday’s $126.06 close, and Cerebras shares were at $118.29, up about 9.1% from $108.37, according to intraday exchange data on Yahoo Finance (percentages are our arithmetic).

TradingKey headlined the Cerebras move as a surge of over 9% tied to Oracle deploying its systems in the first half of 2027, and Benzinga headlined that Oracle stock jumped. Intraday prices move through the session; these are the readings at that time.

Update, 27 Oct: the Nvidia side. Nvidia published its own release the same day. Nvidia and Oracle announced plans to expand OCI Superclusters with Nvidia Vera Rubin NVL72 systems “to multi-gigawatt scale,” and said Vera Rubin NVL72 and Nvidia’s Vera CPUs are now generally available on OCI.

According to the release, OCI also plans to introduce HGX Rubin NVL8: eight Rubin GPUs in an x86 server with two 64-core Intel Xeon 6 CPUs, 4TB of DDR5 memory and 6.4 Tbps of RDMA networking. The release says OpenAI trained and served GPT-6 Astra in Abilene on an OCI Supercluster with more than 100,000 Nvidia Blackwell GPUs, and quotes OpenAI’s Uday Ruddarraju saying OpenAI is “bringing Vera into production on OCI” for its agentic sandbox workloads.

The release does not give a gigawatt total, a dollar value or a timeline for the Vera Rubin expansion.

Oracle and OpenAI, Same Day

In a separate release the same day, Oracle announced an expanded collaboration with OpenAI. It includes Oracle AI Workspace, which brings Oracle applications into ChatGPT through plugins, and Oracle AI Security Mesh, a managed service that the release says “uses OpenAI cyber models” to help customers find and fix vulnerabilities.

Oracle also announced an Oracle AI SMB Bundle that, according to the release, provides “$12,000 in Oracle cloud credits each year” and bundles ChatGPT Business, Codex and OCI.

Connecting the Coverage

Our TPU sales piece reported on 27 October that a former MediaTek employee, interviewed by an expert network, estimates Google’s external TPU hardware revenue at about $51 billion in 2027; that estimate is the former employee’s, not Alphabet’s. Alphabet reports its third quarter on 28 October.

Our Broadcom piece reported on 26 October that a $63 billion special purpose vehicle will buy Broadcom chips and lease them to Anthropic. The Oracle release is a second route by which Google’s TPU architecture reaches buyers outside Google’s own cloud.

The Structural Read

As we read it, the release splits the work: Google supplies the TPU architecture, and OCI says it will provide the power, cooling and datacenter networking to deploy it at scale.

As we read it, the timing matters as much as the headline: the release says availability starts in calendar year 2027 and ramps over a multi-year period, and it names no customers or capacity.

As we read it, Oracle widened two relationships on one day: TPUs from Google for OCI, and OpenAI models, ChatGPT and Codex inside Oracle’s applications, according to its two releases.

Mahesh Thiagarajan, president, Oracle Cloud Infrastructure, joint release, 27 October 2026

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

Three Implications

FOR GOOGLE The joint release says this is the first time Google TPUs will be offered on a hyperscale cloud platform outside Google.

FOR ORACLE Oracle says OCI will provide the power, cooling and datacenter networking for the TPUs, alongside the other AI compute it offers.

FOR BUYERS The release says customers will be able to choose Google TPUs within OCI for training, post-training and inference, from calendar 2027.

The Business Engineer Lens

This story maps onto the Business Engineer framework The State of The Inference Economy.

The framework puts it this way: “Inference now accounts for approximately two-thirds of all AI compute in 2026, up from one-third in 2023 and half in 2025 (Deloitte).”

As we read it, the joint release’s split between a training chip and an inference chip fits that shift. The framework does not discuss Oracle or Google TPUs on OCI.

What Is Not Established

The joint release gives no dollar value, chip count, capacity in megawatts, regions or named customers. It does not say how revenue is shared between Oracle and Google. Oracle’s OpenAI release gives no pricing beyond the SMB credits.

This Week In The AI Supercycle

Execution Fault Lines, Concentrated Demand, and Geopolitical Firewalls

This week’s issue: the AI build is not short of demand for its parts. It is short of cheap money, and its end revenue rests on a narrow base of large, often venture-funded buyers.

Read the full issue →

Business Engineer Framework

The State of The Inference Economy

A Business Engineer map of where AI compute is spent as inference grows.

Read the Map of AI →

The Bottom Line

Oracle and Google say OCI will offer Google TPUs from calendar 2027, the first hyperscale cloud outside Google to do so. On the same day, Oracle widened its OpenAI collaboration across ChatGPT, Codex and security.

95,000+ executives read Business Engineer for the AI strategy frameworks cited by ChatGPT, Claude, and Perplexity.

A note on sourcing. This piece is based on the Oracle and Google joint release and Oracle’s OpenAI release, both dated 27 October 2026 and read in full on 27 October 2026, and on our own earlier reporting. We did not contact Oracle, Google or OpenAI.

Sources: Oracle and Google joint release, 27 Oct 2026 · Oracle release on OpenAI, 27 Oct 2026

Scroll to Top
FourWeekMBA