OpenAI Reportedly Plans to Buy Google’s AI Chips

OpenAI is reportedly getting ready to buy Google’s AI chips, giving Nvidia another reason to watch its rivals closely. The ChatGPT maker needs enormous computing power to train and run its AI models. Turning to Google’s processors could give OpenAI another option as demand for AI infrastructure continues to climb.

OpenAI reportedly wants Google’s Tensor Processing Units

According to The Information, OpenAI plans to purchase Google’s Tensor Processing Units, commonly known as TPUs. The report cites a person familiar with the matter. The publication didn’t disclose the potential deal’s financial terms or other details. Neither company has publicly confirmed the reported purchase.

Google developed TPUs specifically for AI workloads, making them an alternative to Nvidia’s widely used graphics processing units. These specialized processors can help run AI models, including the computing-intensive process known as inference.

OpenAI already relies heavily on Nvidia hardware. However, adding Google’s chips could diversify its computing resources and potentially give the company more flexibility as its infrastructure needs grow.

Google could gain ground against Nvidia

The reported purchase would also represent a significant opportunity for Google’s chip business. The company has increasingly made its TPUs available beyond its own cloud infrastructure, attracting customers looking for alternatives to Nvidia.

OpenAI has reportedly rented Google’s AI chips through Google Cloud since 2025. The latest report suggests the company could take another step toward expanding its use of the hardware.

That said, the move would not mean OpenAI is abandoning Nvidia. The company has substantial computing commitments involving Nvidia hardware and continues developing its own custom processors with Broadcom. Moreover, OpenAI is also hiring engineers to optimize its advanced models for TPUs, according to The Information. This suggests the company is exploring how effectively its models can run on Google’s architecture.