OpenAI sticks with Nvidia, holds off on Google's AI chips

OpenAI has officially stated that it has no immediate intentions to incorporate Google's specialized artificial intelligence chips, known as Tensor Processing Units (TPUs), into its offerings. This clarification comes in response to recent claims that the company was considering Google's hardware to address the surging demand for AI computing. A spokesperson for OpenAI communicated to Reuters on Sunday that while the company is in the preliminary stages of testing some of Google's TPUs, there are currently "no plans to deploy them at scale." It is standard for AI firms to experiment with various chip technologies, but implementing new hardware on a large scale usually necessitates considerable adjustments to both system architecture and software support. OpenAI remains predominantly dependent on Nvidia's graphics processing units (GPUs) while also utilizing AMD chips to accommodate its increasing computational needs. Earlier reports indicated a partnership between OpenAI and Google Cloud, a notable collaboration given their status as competitors in the AI landscape. Despite these partnerships, the majority of OpenAI's computational resources are projected to come from GPU servers supplied by CoreWeave, a specialized cloud service provider. As the competition in AI intensifies, OpenAI is actively seeking to minimize its reliance on third-party chips. The company is in the process of developing its own custom AI processor, aiming to achieve a significant “tape-out” milestone later this year. This milestone marks the completion of a chip's design, ready for manufacturing. Reports from Taiwanese media, including Commercial Times, suggest that OpenAI's in-house chip, which is being developed with support from Broadcom and Taiwan Semiconductor Manufacturing Company (TSMC), could potentially be released by the end of 2025. In contrast, Google has broadened access to its previously internal TPU chips for external clients, attracting significant players such as Apple and rival AI firms like Anthropic and Safe Superintelligence, founded by ex-OpenAI executives. The escalating demand for robust AI chips, particularly those produced by Nvidia, has dramatically transformed the technology sector, propelling Nvidia to the status of the world's most valuable publicly traded company earlier this year. As costs rise and competition becomes fiercer, OpenAI and its peers are seeking innovative strategies to diversify their hardware resources.