OpenAI has announced that it has no active plans to use Google’s in-house AI chips, known as Tensor Processing Units (TPUs), to power its products at scale. This statement comes just two days after reports surfaced suggesting that the AI lab was turning to its competitor’s chips to meet growing demand.
A spokesperson from OpenAI clarified on Sunday that while the company is experimenting with Google’s TPUs, there are no plans to deploy them widely at this time. “It’s common for AI labs to test different hardware solutions, but scaling up new technology requires significant adjustments in architecture and software support,” the spokesperson said.
Currently, OpenAI relies heavily on Nvidia’s Graphics Processing Units (GPUs) and AMD’s AI chips to handle its increasing computational needs. Additionally, the company is developing its own AI chip, aiming to reach the “tape-out” milestone this year, which marks the finalization of the chip’s design before manufacturing begins.
In a surprising move, OpenAI has also signed up for Google Cloud services to support its growing requirement for computing capacity, despite being major competitors in the AI sector. Most of OpenAI’s computing power continues to come from GPU servers operated by CoreWeave, a company specializing in scalable cloud solutions.
Google, on the other hand, has been expanding external access to its TPUs, which were previously reserved for internal projects. This strategy has attracted notable clients like Apple and startups such as Anthropic and Safe Superintelligence, both of which are competitors to OpenAI founded by its former leaders.
The AI industry is witnessing rapid growth and collaboration, even among competitors, as companies seek the best hardware solutions to advance their technologies. For now, OpenAI remains committed to its current hardware providers while exploring all options to meet future demands.
