The Pulse
OpenAI Plans to Buy Google’s AI Chips, The Information Reports
OpenAI plans to buy Google’s Tensor Processing Units, according to an October 27 report by The Information, but the report gives no price or purchase details. The move would extend OpenAI’s use of Google chips as the company hires engineers
AI.info Team · Written with AI · fact-checked ·
“The next generation of AI will require a diverse range of compute,”
Amin Vahdat, senior vice president and chief technologist for AI and Infrastructure at Google, in an October 27 announcement with Oracle.
OpenAI plans to buy Google’s artificial-intelligence chips, The Information reported on October 27, citing a person familiar with the matter. The report did not identify the chips’ quantity, price or delivery schedule, and neither company’s announcement confirms a purchase. If completed, the deal would add a direct procurement relationship to OpenAI’s existing use of Google’s Tensor Processing Units, or TPUs.
From renting TPUs to a reported purchase
The Information previously reported that OpenAI had begun using Google TPUs, reducing some reliance on Nvidia hardware. In its latest article, the publication says OpenAI also rents Nvidia chips through Google Cloud. A direct chip purchase would be a different arrangement from renting computing capacity through a cloud provider, though the report does not explain where OpenAI would install the chips or how it would operate them.
The earlier TPU use did not amount to a public commitment to expand it. After The Information reported on the initial arrangement, an OpenAI spokesperson told the outlet the company did not have “current plans to use TPUs specifically.” The new report indicates a change in those plans, but its sourcing remains a person familiar with the matter rather than a public confirmation from OpenAI or Google.
OpenAI is hiring for TPU optimization
OpenAI’s current job listing for a software engineer on its TPU compilers team adds evidence of technical work on the hardware. The listing says the hire would “lead the optimization of OpenAI’s state-of-the-art frontier models on TPUs” and describes responsibilities including profiling workloads, improving latency and throughput, and optimizing kernels. It welcomes experience in either model training or inference, but does not say the role is tied to a particular purchase or deployment.
The job posting points to a practical requirement: software designed for one accelerator does not automatically run efficiently on another. OpenAI says the engineer would work across model code, kernels and hardware execution. That makes TPU-specific engineering consistent with the reported buying plan, without independently confirming the plan or its scale.
Google is widening TPU access beyond its cloud
On October 27, Google and Oracle announced that Google TPUs will become available through Oracle Cloud Infrastructure, the first time Google says the chips will be offered on a hyperscale cloud platform outside Google. Oracle says the service is scheduled to begin in calendar year 2027 and ramp over multiple years. The announcement does not identify OpenAI as a customer, and it does not establish whether OpenAI’s reported purchase would use Oracle’s infrastructure.
The separate announcements show Google expanding how customers can obtain TPU capacity while OpenAI builds expertise in running its models on the chips. Google Cloud also announced on October 21 that OpenAI models would be offered to select enterprise customers through its Model Garden, creating a commercial relationship between the two companies alongside their competition in AI products. None of those public announcements provides terms for a chip sale to OpenAI.
The deal’s size and structure remain unknown
The report’s central claim is specific—OpenAI plans to buy Google AI chips—but leaves the transaction’s mechanics unresolved. It does not say whether the deal covers training, inference or both, how many TPUs OpenAI might acquire, or when they could be installed. Those details matter because buying chips requires more than securing silicon: a buyer also needs compatible systems, power, cooling and software tuned to the hardware.
For now, the public evidence is an anonymously sourced report, an OpenAI job listing devoted to TPU optimization and Google’s expanding routes to make TPUs available. The listing confirms that OpenAI is seeking TPU expertise; it does not confirm a purchase. The Information’s report supplies the purchase claim, while leaving its financial terms and deployment plan undisclosed.