The Information: OpenAI plans to buy Google TPU chips to reduce dependence on Nvidia
According to The Information, people familiar with the matter said OpenAI plans to purchase Google's AI chips. If OpenAI decides to buy Google's tensor processing units (TPUs), it would be a major development for Google's chip business. It is not yet clear what the financial terms of the purchase and other details are.
OpenAI is hiring an engineer to lead efforts to optimize the company's most advanced frontier models so that they can run on TPUs. The job description shows that the role will be responsible for full forward-pass performance optimization, from model code to kernels to hardware execution.
The Information reported last year that OpenAI had begun using Google TPUs, thereby reducing its dependence on Nvidia chips. After that report, an OpenAI spokesperson said the company currently had no plans to use TPUs specifically. This job posting suggests that model optimization work for TPUs is still at an early stage. OpenAI is currently one of Nvidia's major buyers of GPUs for model training and inference.
Original sources:
Title: OpenAI plans to buy Google's AI chips, The Information reports Publisher: Reuters Publication date: 2026-10-28 URL: https://www.reuters.com/business/openai-plans-buy-googles-ai-chips-information-reports-2026-10-28/ Source type: media report (citing The Information) Verification status: full report read in this session, verification date 2026-10-28 Supporting details: confirms that The Information reported OpenAI plans to buy Google AI chips, no financial terms disclosed, OpenAI declined to comment
Title: Software Engineer, TPU Compilers Publisher: OpenAI careers page (republished across platforms) Publication date: from 2026-10-20 URL: https://www.foundrole.com/jobs/software-engineer-tpu-compilers-at-openai-01a15ae5-eb41-7d9f-a23e-65e1012f8b2d Source type: company hiring information Verification status: job description read in this session, verification date 2026-10-28 Supporting details: confirms hiring of an engineer to lead optimization of frontier models to run on TPUs, with responsibilities including full forward-pass performance optimization
ABAB AI Insight
This report continues the path from mid-2025, when OpenAI began renting TPUs through Google Cloud for inference. At that time, an OpenAI spokesperson explicitly said there were no plans for large-scale deployment, while the current hiring suggests the company is making technical preparations for a possible purchase, enabling its models to run efficiently on TPU hardware.
Google's TPUs have long been used mainly by Google itself and some external customers; selling them directly to OpenAI would expand its external market. For Nvidia, OpenAI is an important GPU customer, and any substantial shift in purchasing could affect its share in inference. The key indicators to watch are whether the hiring turns into actual purchase orders, and the scale and use case of the purchases (inference or training).
OpenAI is also advancing its own chips (in collaboration with Broadcom). If TPU purchases materialize, they would form part of its diversification of compute sources rather than a single replacement.
ABAB News · Cognition Law
The hiring of TPU optimization engineers before any formal purchase disclosure shows that technical adaptation work comes before contract completion.
The follow-up report after the spokesperson's denial reflects the gap between the company's staged statements and its actual preparations on its path to diversifying compute.