OpenAI has announced that it has no plans to adopt Google AI chips yet, and early testing is underway



At the end of June 2025, it was reported that OpenAI had begun working to adopt Google's AI chip 'TPU' for AI inference, but OpenAI has now stated that 'at this time, we have no plans to deploy TPUs on a large scale.'

OpenAI says it has no plan to use Google's in-house chip | Reuters
https://www.reuters.com/business/openai-says-it-has-no-plan-use-googles-in-house-chip-2025-06-30/

Google began developing processors specialized for machine learning in the 2010s, and in 2016 announced the existence of a machine learning specialized processor 'Tensor Processing Unit (TPU)'. TPUs began to be made available to general users via Google's cloud services in 2017, and generation updates are announced regularly. The 7th generation TPU 'Ironwood', which appeared in April 2025, is equipped with 192GB of memory per chip and has a processing speed of 4614 TFLOPS per chip. This makes it possible to run various inference models at high speed.

Google announces 7th generation TPU 'Ironwood' optimized for processing inference models, with 192GB of memory per chip and performance per pod more than 24 times that of the strongest supercomputer 'El Capitan' - GIGAZINE



OpenAI is known as a major customer of NVIDIA, and has built a large-scale AI infrastructure by purchasing large quantities of NVIDIA AI chips such as H100 and H200. On June 28, 2025, Reuters reported that OpenAI has begun renting Google's TPUs to process its own AI. At that time, Reuters pointed out that 'This agreement is part of Google's move to expand external use of TPUs, which had been limited to internal use, and will be the first time that OpenAI will meaningfully use chips other than NVIDIA. OpenAI has shown its intention to move away from its reliance on Microsoft's data centers.'

OpenAI reportedly begins using Google's TPU for inference processing such as ChatGPT, the first time it has used an AI processing chip other than NVIDIA - GIGAZINE



On June 30, 2025, an OpenAI spokesperson told Reuters, 'OpenAI's lab is conducting early testing with Google's TPUs. However, we currently have no plans to deploy TPUs on a large scale.' This confirmed that OpenAI was testing TPUs, but also made it clear that the company was not yet at the stage where TPUs would be used in actual products. However, Google declined to comment.

In response to the Reuters report, NVIDIA commented, 'We look forward to continuing our collaboration with OpenAI to shape the future of AI.'




Related Posts:

in Hardware, Posted by log1o_hf