OpenAI’s push toward purchasing Google’s Tensor Processing Units marks a strategic shift from renting cloud capacity to owning chips and expanding alternatives to Nvidia processors.
OpenAI’s plans to purchase Tensor Processing Units and transition from renting to owning
Hiring engineers to optimize models and develop the Halapeno chip with Broadcom
Google’s expansion in making its chips available via Oracle Cloud and the Anthropic deal
Implications of the deal for Alphabet shares and cloud computing economics
Frequently asked questions
OpenAI’s plans to purchase Tensor Processing Units and transition from renting to owning
OpenAI plans to buy Tensor Processing Units from Google, according to a report by The Information on Tuesday, October 27, 2026, citing a person familiar with the matter. The report did not disclose financial terms or other details, according to Reuters.
In a separate report, Bloomberg stated that OpenAI intends to buy artificial intelligence computing chips from Alphabet’s Google, citing its own source who clarified that the final terms of any deal are not yet known, while both OpenAI and Google declined to comment.
Direct purchasing will transition OpenAI from renting Google chips to owning them, and indicates a diversification of its sources away from Nvidia, whose graphical processors it relies on heavily. Reuters reported in June 2025 that the maker of ChatGPT had begun renting artificial intelligence chips from Google.
Hiring engineers to optimize models and develop the Halapeno chip with Broadcom
OpenAI also posted a job opening for an engineer specialized in optimizing its flagship models to work on Tensor Processing Units, noting that experience in training or inference is welcome according to Yonhap News Agency, which The Information interpreted as a sign that optimization efforts are still in an early stage.
When OpenAI began renting those units last year, it stated at the time that it “has no specific plans to use Tensor Processing Units at the present time,” according to the Seoul Economic Daily.
According to Bloomberg, this plan is part of a broader push by OpenAI to quickly add computing processing capability, which includes its in-house chips called Halapeno developed in collaboration with Broadcom, which it plans to start using later this year. Richard Ho, who leads the chip team at OpenAI, previously worked on Tensor Processing Units at Google.
Google’s expansion in making its chips available via Oracle Cloud and the Anthropic deal
This report came on the same day Oracle announced it will offer Tensor Processing Units through its cloud infrastructure, marking the first time a major cloud provider other than Google provides these chips. The service is scheduled to launch by 2027, according to The Information. Google executive Amin Vahdat said in a statement announcing the Oracle deal: “The next generation of artificial intelligence will require a diverse set of computing capabilities.”
Anthropic also disclosed in its initial public offering filing this week a five-year, $161 billion agreement with Broadcom to rent equipment powered by Google’s Tensor Processing Units, according to The Herald Business, while OpenAI announced last week the availability of its flagship models via Google Cloud.
Implications of the deal for Alphabet shares and cloud computing economics
Alphabet shares rose in after-hours trading following the publication of the report ahead of its quarterly earnings announcement, according to Yahoo Finance, as investors await signs that its heavy artificial intelligence spending continues to drive growth in its cloud division, according to Reuters.
In a comprehensive analytical reading of these fast-paced developments, it becomes clear that transforming the Tensor Processing Units developed by Google from an exclusive internal tool into processors sold and rented to major competitors like OpenAI and Anthropic, and made available via Oracle Cloud, redraws the competitive landscape in the semiconductor market and grants major artificial intelligence companies negotiating chips and alternative technical backups to Nvidia processors.
Furthermore, the timing of reports from The Information, Bloomberg, Reuters, and Yonhap regarding OpenAI’s plans to own Google’s Tensor Processing Units, alongside Oracle’s announcement to provide those chips via the cloud by 2027 and Anthropic’s disclosure of the $161 billion deal with Broadcom, confirms that the artificial intelligence infrastructure market is witnessing a profound structural shift toward diversifying computational accelerators and combining Nvidia processors with custom chips like Halapeno and Google units.
As part of the comprehensive editorial review for the Friday, October 30, 2026 newsletter regarding the “Google chip purchase” file, journalistic fact-checking based on primary sources and documented news reports emphasizes the methodological necessity of distinguishing between officially announced financial and technical figures and future expectations or preliminary tests under trial, ensuring the Arab reader receives an integrated and accurate professional coverage spanning all aspects of the event and its direct and indirect repercussions on the global artificial intelligence sector and future competition among major tech companies.
Frequently asked questions
Question: What did reports from The Information and Bloomberg reveal regarding OpenAI and Google chips?
Answer: They revealed that OpenAI plans to purchase Google’s Tensor Processing Units and own them directly rather than just renting them, without final financial terms being announced yet.
Question: How is OpenAI developing its own chips alongside purchasing from outside sources?
Answer: The company is developing internal chips named Halapeno in cooperation with Broadcom and plans to start using them later this year under the supervision of chip team head Richard Ho.
Question: What other agreements did Oracle and Anthropic make regarding Google chips?
Answer: Oracle announced the availability of Tensor Processing Units through its cloud by 2027, while Anthropic disclosed a 5-year, $161 billion agreement with Broadcom to rent equipment based on those chips.
Question: Why is OpenAI seeking to buy Tensor Processing Units from Google?
Answer: To accelerate the addition of computing capabilities necessary to train and run its models and to diversify supply chains to reduce exclusive reliance on Nvidia processors.
