On October 28, it was reported that AI leader OpenAI plans to purchase Google’s proprietary AI chips to accelerate the diversification of its computing power supply chain and reduce reliance on traditional chip giants and cloud service providers.

Author and source: AIBase

OpenAI is preparing to buy Google’s AI chips.

On October 27, the U.S. tech media The Information, citing sources familiar with the matter, reported that OpenAI plans to purchase Google’s custom AI chip, the TPU (Tensor Processing Unit). The financial terms, scale of procurement, and other details of this potential deal have not been disclosed, and neither OpenAI nor Google has publicly commented.

OpenAI’s collaboration with Google on computing power did not begin today. In June 2025, OpenAI started leasing TPUs via Google Cloud to provide computing support for products like ChatGPT—marking OpenAI’s first substantial use of chips other than NVIDIA. At the time, OpenAI’s primary goal was to reduce inference costs, and Google did not lease its most powerful TPUs to OpenAI. This time, OpenAI has moved from leasing to purchasing, signaling a significant deepening of their partnership. A notable indicator is that OpenAI is currently hiring software engineers in San Francisco specializing in TPU compilers to lead optimizations for running OpenAI’s cutting-edge models more efficiently on TPUs. This demonstrates that OpenAI’s use of Google’s chips has progressed from initial test-based leasing to foundational tuning tailored for its own models. Just last week, OpenAI announced it would make its frontier models available through Google Cloud, further intensifying the commercial ties between the two companies.

OpenAI doesn't want to bet solely on NVIDIA

As a leading player in the large model field, OpenAI has long been one of NVIDIA’s most important customers, heavily relying on NVIDIA GPUs for both training and inference, and carrying substantial commitments for computing power procurement. However, with NVIDIA’s chips being expensive and in short supply, OpenAI has been seeking additional alternatives.

In addition to adopting Google TPUs, OpenAI is also advancing its own chip development. OpenAI is collaborating with Broadcom to design custom chips and has previously stated its intention to begin using these in-house chips to support model operations later this year. Richard Ho, who leads OpenAI’s chip team, previously worked on the development of Google’s TPUs. By procuring TPUs, developing custom chips, and continuing to purchase NVIDIA GPUs, OpenAI is pursuing multiple parallel paths to expand its computing power and reduce dependence on any single supplier.

Google TPU acceleration is expanding into external markets

For Google, acquiring OpenAI would be equally significant. Google began developing TPUs over a decade ago, initially to enhance the speed and efficiency of its search engine, and these chips have since become the core hardware underpinning AI computing. This year, Google began directly selling TPUs to select customers, with Anthropic, a competitor of OpenAI, being a major client. On October 27, Google also announced that Oracle would become the first major cloud provider outside of Google to offer TPUs to customers; Oracle will purchase TPUs directly and deploy them on its own cloud platform. Alphabet’s Chief Financial Officer, Anat Ashkenazi, previously stated that revenue from certain TPU sales contracts will begin to be recognized by the end of this year, with the majority of revenue expected to materialize next year.

OpenAI and Google are direct competitors in large models, AI assistants, and even search services, with ChatGPT and Gemini head-to-head. Yet, faced with demands for computing power, both have chosen to collaborate. OpenAI’s plan to purchase Google’s chips reflects not only its accelerating efforts to diversify its computing supply chain but also the emerging landscape in which tech giants simultaneously compete and cooperate in AI infrastructure. As Google’s TPU customer base continues to grow, NVIDIA’s dominant position in the AI chip market is facing increasing challenges.