OpenAI has clarified that it at present has no plans to deploy Google’s in-house synthetic intelligence chips at scale, regardless of studies suggesting in any other case. The assertion comes simply days afterReuters and a number of other different retailers claimed the AI analysis lab was turning to Google’s Tensor Processing Models (TPUs) to help its increasing compute wants.
A spokesperson for OpenAI, which is behind ChatGPT, mentioned on Sunday that whereas the corporate is “conducting early testing” with Google’s TPUs, there isn’t a lively intention to scale their use. “We’ve got no plans to undertake TPUs broadly at this stage,” the spokesperson confirmed.
Google, when approached for remark, declined to reply.
Notably, it’s not uncommon for AI companies to experiment with completely different {hardware} configurations, however rolling out new chip infrastructure throughout manufacturing methods would require vital architectural adjustments and software program adaptation, one thing that usually takes time and appreciable sources.
For now, OpenAI continues to rely totally on Nvidia’s graphics processing items (GPUs), that are thought-about the business normal for AI workloads. The corporate can also be utilising superior chips from AMD to satisfy rising computational demand. Concurrently, OpenAI is pushing forward with growing its proprietary AI chip, which is predicted to hit the “tape-out” part later this 12 months, a crucial step the place the chip design is finalised for fabrication.
Earlier in June,Reuters reported that OpenAI had signed up for Google Cloud providers, signalling an sudden partnership between the 2 tech rivals. Nonetheless, sources point out that almost all of OpenAI’s cloud-based computing nonetheless runs on servers operated by CoreWeave, a fast-growing infrastructure firm providing GPU-powered options.
The American tech big has lately begun providing its custom-designed TPUs to exterior clients, increasing past their earlier inside use. The transfer has attracted main purchasers, together with Apple and AI startups like Anthropic and Secure Superintelligence, each based by former OpenAI executives.
(With inputs from Reuters)
