OpenAI Explores Google’s TPUs to Reduce AI Inference Costs
Table of Contents
By Amelia Hernandez | SAN FRANCISCO – 2025/06/28 11:54:39
In a move that could diversify its computing resources, OpenAI is reportedly testing Google’s Tensor Processing Units (TPUs) for artificial intelligence tasks, possibly reducing its reliance on Nvidia and Microsoft.
OpenAI, the company behind ChatGPT, is exploring alternative chip solutions for its AI operations. According to a source familiar with the matter,OpenAI has recently rented Google’s TPUs (tensor Processing Units) for AI processing. This marks the first time OpenAI has utilized chips from a manufacturer other than Nvidia, signaling a potential shift away from its dependence on Microsoft’s data centers.
The insider, who spoke to the Reuters news agency, suggests that TPUs could become a more cost-effective alternative to nvidia’s GPUs. OpenAI aims to leverage Google Cloud to address its increasing demand for computing power. By renting TPUs through Google Cloud,OpenAI hopes to lower the expenses associated with AI inference. While Google declined to comment, OpenAI has not yet responded to requests for comment from Reuters.
Potential Impact on the AI Chip Market
This advancement could have significant implications for the AI chip market, potentially driving competition and innovation among chip manufacturers. The move suggests that OpenAI is seeking to optimize its infrastructure costs and diversify its supply chain.
“Openai hopes that the TPUs, which rents it via Google Cloud, will reduce the costs of inferences.”
Understanding TPUs and the AI Chip Landscape
