OpenAI’s chip hedge gets juicier as Google TPUs challenge Nvidia’s dominance

OpenAI plans to purchase AI chips from Google, according to a person familiar with the matter who spoke with The Information. Neither company has confirmed the deal, and the price, number of chips and delivery dates have not been disclosed. The reported acquisition would reduce OpenAI’s reliance on Nvidia, which supplies most of the hardware behind the AI boom and counts OpenAI as one of its biggest customers.

Google’s tensor processing units, or TPUs, are designed specifically for the mathematical operations used to train and run AI models. Google revealed the chip in 2016, saying it had already deployed TPUs in its data centers for more than a year. Reports in June 2025 said OpenAI had begun renting Google’s TPUs, describing the work as early testing with no plans to deploy them at scale. Buying the hardware would give OpenAI ownership rather than hourly access, although other details remain private.

An OpenAI job listing in San Francisco seeks an engineer to lead optimization of the company’s frontier models on TPUs. The role involves compilers, software that translates AI models into instructions a specific chip can execute. OpenAI hardware lead Richard Ho spent nearly nine years at Google and helped lead its Cloud TPU project before joining OpenAI.

OpenAI has expanded its hardware partnerships while preparing for large-scale deployment. In September 2025, Nvidia signed a letter of intent to invest up to $100 billion in OpenAI as it deployed at least 10 gigawatts of Nvidia systems, with one gigawatt representing a billion watts. Two weeks later, AMD gave OpenAI the right to buy up to 160 million AMD shares, roughly 10% of the company, tied to OpenAI deploying 6 gigawatts of AMD chips. Nvidia ultimately invested $30 billion in OpenAI’s funding round, below the $100 billion letter of intent. CEO Jensen Huang said in March that it might be the last such investment before OpenAI goes public. In June, OpenAI unveiled Jalapeño, its first custom chip, built with Broadcom for inference, with initial deployment planned for the end of 2026.

Google unveiled two eighth-generation TPUs in April: the 8t for training AI models and the 8i for inference. At its Cloud Next 2026 conference in Las Vegas on Wednesday, Google introduced two AI processors as part of its eighth generation of custom silicon targeting Nvidia’s AI-chip dominance. The training-focused TPU 8t delivers nearly 3x the compute performance per pod over its predecessor, with a single superpod scaling to 9,600 chips and providing 121 ExaFlops of compute capacity. Google says the architecture offers 2.8x better price-to-performance, with performance measured against the previous-generation Ironwood chip rather than Nvidia hardware.