OpenAI moves to buy Google's AI chips, reducing its reliance on Nvidia
- Input
- 2026-10-28 14:42:07
- Updated
- 2026-10-28 14:42:07

On the 27th (local time), the U.S. IT news outlet The Information reported that “OpenAI plans to buy Google’s AI chip, the Tensor Processing Unit (TPU).” Specifics such as the size of the purchase, contract value and delivery schedule have not been disclosed. OpenAI and Google also did not provide specific comment on the reports.
Until now, OpenAI had been using Nvidia graphics processing units (GPUs) on a large scale to handle the computing needed to train AI models and run ChatGPT. But as use of AI services surged, securing chips and covering operating costs emerged as major challenges.
Against this backdrop, OpenAI has been testing the option of renting TPUs through Google Cloud since last year. The UK media explained that “if this purchase plan comes to fruition, the partnership would expand beyond renting Google’s AI chips through the cloud to securing them directly.”
In this regard, Google’s in-house-developed TPUs are chips specialized for AI computing and are currently attracting attention as an alternative to Nvidia GPUs. They have been seen as potentially improving cost efficiency, particularly in inference tasks, in which already-trained AI models answer users’ questions.

The report came a day before Google parent Alphabet’s quarterly earnings announcement. Alphabet is scheduled to report its third-quarter results after the New York stock market closes on the 28th. Investors are watching to see whether Google’s massive investments in AI infrastructure are translating into higher revenue and improved profitability for its cloud business.
[email protected] Hong Chae-wan Reporter