TL;DR: OpenAI plans to buy Google’s AI chips, known as Tensor Processing Units (TPUs), according to an October 27 report from The Information that Reuters relayed. No price, volume or delivery date has been disclosed, and neither company has confirmed a deal: OpenAI declined to comment and Google did not immediately respond to Reuters. The report landed the same day Oracle and Google announced that Google TPUs will be offered on Oracle Cloud Infrastructure starting in 2027, the first time Google’s chips will be sold through another hyperscale cloud.
The news that OpenAI plans to buy Google TPUs is notable because OpenAI is one of the biggest customers of Nvidia’s graphics processors, and because Google is one of its biggest rivals in chatbots and search. If the plan goes ahead, it would be another sign that the companies building frontier AI models no longer want to rely on a single chip supplier. Here is what has actually been reported, what is still unknown, and why it matters.
What the report says
The Information reported on Tuesday, October 27, citing a person familiar with the matter, that OpenAI, “a longtime Nvidia customer,” plans to buy Google’s AI chips. Reuters picked up the story the same evening and added several details:
- The report did not disclose financial terms or other details of the purchase.
- OpenAI declined to comment, and Google did not immediately respond outside business hours.
- Reuters reported in June 2025 that OpenAI had already begun renting Google’s chips through Google Cloud.
- The week before this report, OpenAI said it would make its frontier AI models available through Google Cloud.
- Google began selling TPUs directly to some customers this year, and deals with customers including Anthropic have driven demand.
In other words, this is a reported plan, not an announced contract. Until OpenAI or Google confirms it, the size and timing are unknown.
Renting vs buying: what changes
The difference between renting and buying matters. When OpenAI rents TPUs, it pays Google Cloud for time on chips that sit in Google’s data centers. Buying chips outright, or committing to a dedicated supply, usually means a longer and larger commitment. Back in June 2025, an OpenAI spokesperson told Reuters that the company was in early testing with some TPUs but had “no plans to deploy them at scale right now.” A purchase plan would suggest that position has changed.
| June 2025 | October 2026 (reported) | |
|---|---|---|
| Arrangement | Renting Google TPUs via Google Cloud | Plans to buy Google TPUs |
| OpenAI’s public stance | Early testing, no plans to deploy at scale | Declined to comment |
| Disclosed size or terms | None | None |
| Main chip supplier | Nvidia | Still Nvidia, plus its own Broadcom-designed chip |
Why OpenAI wants more than Nvidia
Every response from ChatGPT is a round of AI inference, and serving hundreds of millions of users takes enormous amounts of compute. Reuters notes that Amazon, Microsoft and Alphabet plan to spend as much as $600 billion combined this year on AI infrastructure, while investors keep asking when that spending will pay off. Having more than one chip supplier can help a company like OpenAI on several fronts:
- Supply. Nvidia’s top chips are in heavy demand. A second source of accelerators reduces the risk of waiting in line.
- Price. Reuters says TPUs have gained traction “as companies seek cheaper options.” Competition between suppliers can improve negotiating power.
- Workload fit. Some chips are tuned for training and others for serving. Google now makes separate TPU designs for each.
OpenAI is also building its own hardware. Reuters says OpenAI unveiled its first custom AI chip, designed with Broadcom, in June and planned to deploy it by the end of this year as part of a multi-generation chip program. So a TPU deal would sit alongside Nvidia GPUs and OpenAI’s in-house chip, not replace them.
Oracle becomes the first outside cloud to offer TPUs
The report coincided with a related announcement. At its AI World event in Las Vegas on October 27, Oracle and Google announced that Google TPUs will be available through Oracle Cloud Infrastructure (OCI). Oracle says it is “the first time Google TPUs will be offered on a hyperscale cloud platform outside Google.”
Key details from the companies’ announcement:
- Timing: TPUs on OCI will be available in calendar year 2027, “ramping over a multi-year period.”
- Division of labor: OCI will provide the power, cooling and data center networking; the TPU systems are Google’s.
- Hardware: Google’s eighth generation comes in two versions, TPU 8t for training and TPU 8i for inference. Oracle’s release says TPU 8t delivers “nearly 3x the compute performance per pod over the previous generation,” a claim made by the companies.
- Not disclosed: capacity, pricing, regions and named customers, as TSN and others pointed out.
“Different AI workloads have different compute needs,” Mahesh Thiagarajan, president of Oracle Cloud Infrastructure, said in the release. “Bringing Google TPUs to OCI gives customers access to multiple leading AI compute architectures in one cloud.”
The Information also ran a briefing headlined “Google to Sell AI Chips to OpenAI, Oracle,” tying the two developments together. Oracle is already a major compute partner for OpenAI, so OCI could become one place where OpenAI runs TPUs, though no source has confirmed how or where a purchase would be deployed.
What it means for Google and Nvidia
For Google, selling TPUs to outside buyers turns an internal technology into a product. Reuters reports that Alphabet CFO Anat Ashkenazi has said the company expects to begin recognizing a small percentage of revenue from TPU sales agreements by the end of this year, with the vast majority expected in 2027. Alphabet reports quarterly results later on Wednesday, October 28, so analysts may press for details. Google has also been widening the TPU software story, including open-sourcing TorchTPU for PyTorch developers.
For Nvidia, the report does not mean OpenAI is leaving. Reuters describes OpenAI as “among the largest buyers” of Nvidia GPUs, which it uses for both training and inference. The bigger trend is that frontier labs now spread their bets across Nvidia, Google, AMD and custom chips, as seen in deals such as Nvidia and AMD’s investment in Meta’s stock offering and the financing lined up for Anthropic’s Broadcom chips.
What we still don’t know
- How many TPUs OpenAI plans to buy, and at what price.
- Which TPU generation is involved and when deliveries would start.
- Whether the chips would run in Google’s data centers, Oracle’s or OpenAI’s own sites.
- Which workloads, training or serving, OpenAI would put on them.
Until OpenAI or Google confirms the arrangement, treat the purchase as a credible report rather than a done deal.
The bottom line
OpenAI reportedly moving from renting Google’s TPUs to buying them would mark a real shift for a company that was publicly cautious about the chips in 2025. Combined with Oracle’s plan to offer TPUs from 2027, it shows Google’s chips becoming a genuine alternative in a market Nvidia still leads. For readers new to the topic, our explainers on large language models and AI inference cover why this compute matters so much.
The Teck Hustlers Editorial Team covers technology, business, finance and gaming news for readers who want the story behind the headline. Every article is researched, written and edited by our in-house team, checked against company announcements, filings and reputable reporting, and updated when new facts emerge. We focus on what each development means for founders, professionals, investors and everyday users. Have a tip or a correction? Reach us through our contact page.
