OpenAI is reportedly preparing to buy artificial-intelligence chips from Google, a move that could give the ChatGPT maker another source of computing capacity beyond Nvidia. The Information reported the potential purchase, citing a person familiar with the matter, while neither OpenAI nor Google has confirmed the plan.
The price, number of chips and delivery schedule have not been disclosed. OpenAI is already one of Nvidia’s largest customers, and the reported shift comes as companies developing advanced AI systems seek access to increasingly large amounts of specialized computing hardware.
What OpenAI is reportedly planning
The hardware at the center of the report is Google’s tensor processing unit, or TPU. Google designed the processor specifically for the mathematical workloads used by AI models and first disclosed it in 2016. Unlike a general-purpose processor, an accelerator such as a TPU is built to handle particular calculations involved in training models and generating responses.
OpenAI reportedly began renting Google TPUs in June 2025. At that time, the arrangement was described as early testing, without a plan to deploy the chips at scale. Buying the systems would represent a different type of relationship: OpenAI would own the hardware rather than access it through an hourly rental arrangement.
A recent OpenAI job listing adds support to the possibility that the company is preparing for broader TPU use. The San Francisco-based role calls for an engineer to lead the optimization of OpenAI’s frontier models on Google’s chips. The work includes compilers, the software that converts an AI model into instructions a particular processor can execute.
That requirement highlights why moving between chip platforms is not a simple hardware substitution. Models and the software supporting them often need to be adapted to the architecture of the processor on which they run.
Richard Ho, who leads OpenAI’s hardware program, previously spent nearly nine years at Google. He was among the leaders of Google’s Cloud TPU project before joining OpenAI.
A broader push to secure computing capacity
OpenAI’s reported interest in Google’s hardware follows several major arrangements involving AI-chip suppliers. In September 2025, Nvidia signed a letter of intent to invest as much as $100 billion in OpenAI as the AI company deployed at least 10 gigawatts of Nvidia systems. A gigawatt equals one billion watts and is commonly used to describe the scale of data-center projects.
Nvidia ultimately invested $30 billion in OpenAI’s funding round, below the amount stated in the earlier letter of intent. Nvidia Chief Executive Jensen Huang said in March that the investment could be the company’s last such commitment before OpenAI goes public.
OpenAI also reached an arrangement with AMD in October 2025. Under that agreement, OpenAI received the right to buy as many as 160 million AMD shares, representing roughly 10% of the company, with the arrangement tied to the deployment of 6 gigawatts of AMD chips.
The company is developing hardware of its own as well. In June, OpenAI introduced Jalapeño, its first custom chip, in partnership with Broadcom. The processor is intended for inference—the stage at which a trained AI model produces an answer—and its initial deployment is planned for the end of 2026.
Google and the competitive TPU market
Google unveiled two eighth-generation TPUs in April: the 8t, designed for training AI models, and the 8i, designed for inference. Google’s published performance comparisons measure those chips against its previous-generation Ironwood processor, rather than against Nvidia hardware.
Google is also expanding the way it supplies TPU systems. Alphabet’s chief financial officer said the company began delivering TPU systems to customers’ own data centers in the second quarter. The CFO has also said most revenue from TPU system sales is expected to come in 2027. Alphabet was scheduled to report third-quarter earnings after the U.S. market closed on Wednesday.
OpenAI is not the only major AI company linked to Google’s accelerators. Anthropic, a principal OpenAI rival, signed an agreement with Google and Broadcom in April covering multiple gigawatts of next-generation TPU capacity, which it expects to begin coming online in 2027.
Google Cloud also began offering OpenAI’s GPT-6 models to selected customers in private preview the previous week, adding another commercial connection between the two companies even as OpenAI reportedly considers purchasing Google-designed hardware.
Moneyiar market picture
Moneyiar market data at the stated times show mixed performance among the listed digital assets:
- Ethena (ENA) was priced at $0.2092782867 as of October 25, 2026, at 22:39 +03:30. It was up 8.19% over 24 hours and 12.35% over seven days.
- Compound (COMP) stood at $21.1287447338 at the same time, down 10.42% over 24 hours and up 0.30% over seven days.
- USD.AI (CHIP) was $0.0443846724, down 0.29% over 24 hours and up 9.36% over seven days.
- OpenLedger (OPEN) was $0.0976845525 as of October 24, 2026, at 15:23 +03:30. It gained 2.83% over 24 hours and 3.00% over seven days.
