OpenAI seeks to buy Google’s AI chips to reduce reliance on Nvidia amid competitive hardware landscape
The sought-after chips are Google’s tensor processing units, designed specifically for AI computation. Google first introduced TPUs in 2016 after successfully running them in its data centres for over a year.
Reports from June 2025 indicated that OpenAI had started testing Google’s TPUs but did not plan to use them extensively. A successful acquisition would shift OpenAI from renting to owning the necessary hardware.
OpenAI purchase
OpenAI’s job listing for an engineer to optimise models on TPUs adds credibility to the purchasing rumours.
The role focuses on compilers that facilitate the operation of models on designated chips, indicating the complexities involved in changing hardware.
Richard Ho, leading OpenAI’s hardware efforts, previously spent nearly a decade at Google, where he played a key role in the Cloud TPU project.
In September 2025, Nvidia expressed intent to invest up to $100 billion in OpenAI, aiming to supply at least 10 gigawatts of systems.
Two weeks later, AMD granted OpenAI the option to acquire up to 160 million shares, contingent on deploying 6 gigawatts of AMD chips.
Ultimately, Nvidia contributed $30 billion to OpenAI’s funding, falling short of its initial investment plans.
In June, OpenAI announced Jalapeño, its first custom chip developed with Broadcom for AI inference, with deployment slated for the end of 2026.
Google also introduced two new TPUs in April, targeting AI training and inference without directly comparing their performance to Nvidia’s hardware.
