OpenAI plans to buy AI chips from Google, according to a person familiar with the matter who spoke with The Information. The move would reduce its reliance on Nvidia, which supplies most hardware powering the AI industry.
Neither company has confirmed the plan. The price, quantity, and delivery dates remain undisclosed.
The chips in question are Google's tensor processing units, or TPUs. Google unveiled these processors in 2016, designed specifically for the mathematical operations AI models require. The company had already deployed TPUs in its data centers for over a year before announcing them.
Reports from June 2025 said OpenAI began renting Google's TPUs for early testing with no plans for large-scale deployment. A purchase would mean owning hardware instead of paying Google hourly rates.
A job listing from OpenAI in San Francisco seeks an engineer to lead optimization of its frontier models on TPUs. The role involves compilers, software that translates AI models into instructions specific chips can execute. Switching chips presents technical complexity beyond simple hardware swaps.
Richard Ho leads OpenAI's hardware program. He spent nearly nine years at Google and led its Cloud TPU project before joining OpenAI.
In September 2025, Nvidia signed a letter of intent to invest up to $100 billion in OpenAI tied to deployment of at least 10 gigawatts of Nvidia systems. Nvidia ultimately invested $30 billion in OpenAI's funding round, below the letter of intent amount. CEO Jensen Huang said in March this might be the last such investment before OpenAI goes public.
Two weeks after Nvidia's announcement, 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.
In June, OpenAI unveiled Jalapeño, its first custom chip built with Broadcom. The chip targets inference workloads, with initial deployment planned for end of 2026.
Google released two eighth-generation TPUs in April. The 8t handles model training and the 8i handles inference. Performance figures measure against Google's previous-generation Ironwood chip, not Nvidia hardware.
Anthropic signed an agreement with Google and Broadcom in April for multiple gigawatts of next-generation TPU capacity expected online starting in 2027.
Alphabet began delivering TPU systems to customers' own data centers in the second quarter. Google Cloud started offering OpenAI's GPT-6 models to select customers in private preview last week.