<p>OpenAI has clarified that it currently has no plans to deploy Google’s in-house artificial intelligence chips at scale, despite reports suggesting otherwise. The statement comes just days afterReuters and several other outlets claimed the AI research lab was turning to Google’s Tensor Processing Units (TPUs) to support its expanding compute needs.
A spokesperson for OpenAI, which is behind ChatGPT, said on Sunday that while the company is “conducting early testing” with Google’s TPUs, there is no active intention to scale their use. “We have no plans to adopt TPUs broadly at this stage,” the spokesperson confirmed.
Google, when approached for comment, declined to respond.
Notably, it is not unusual for AI companies to experiment with different hardware configurations, but rolling out new chip infrastructure across production systems would require significant architectural changes and software adaptation, something that typically takes time and considerable resources.
For now, OpenAI continues to rely primarily on Nvidia’s graphics processing units (GPUs), which are considered the industry standard for AI workloads. The company is also utilising advanced chips from AMD to meet rising computational demand. Simultaneously, OpenAI is pushing ahead with developing its proprietary AI chip, which is expected to hit the “tape-out” phase later this year, a critical step where the chip design is finalised for fabrication.
Earlier in June,Reuters reported that OpenAI had signed up for Google Cloud services, signalling an unexpected partnership between the two tech rivals. However, sources indicate that the majority of OpenAI’s cloud-based computing still runs on servers operated by CoreWeave, a fast-growing infrastructure company offering GPU-powered solutions.
The American tech giant has recently begun offering its custom-designed TPUs to external customers, expanding beyond their previous internal use. The move has attracted major clients, including Apple and AI startups like Anthropic and Safe Superintelligence, both founded by former OpenAI executives.
(With inputs from Reuters)
