OpenAI Opts Out of Google TPU for Now
OpenAI has clarified that it has no current plans to adopt Google’s in-house chips, despite recent media reports suggesting otherwise, according to Reuters.
On Saturday, several media outlets reported that OpenAI was testing Google’s Tensor Processing Units (TPUs) and might use them to meet its growing computational demands. However, in a statement issued on Sunday, the company said, “While we’ve conducted some early testing, we don’t currently have plans to use Google’s TPUs more broadly.”
At present, OpenAI relies on Nvidia’s Graphics Processing Units (GPUs) and AMD chips for its operations. The company is also in the process of developing its own custom chip, which is expected to reach the tape-out phase — the final step in chip design — later this year.
While OpenAI recently began using Google Cloud, the bulk of its computing power still comes from GPU servers provided by CoreWeave.
Meanwhile, Google continues to expand access to its TPU chips to external users, including major companies like Apple.
