OpenAI Develops Custom Chip with Broadcom and TSMC Amid Chip Shortages

OpenAI is collaborating with Broadcom and TSMC to develop its inaugural in-house chip tailored for its artificial intelligence systems. This move, alongside integrating AMD chips with Nvidia’s, is in response to the company’s expanding infrastructure requirements, as reported by sources to Reuters.

The rapidly growing ChatGPT developer has weighed various options to diversify its chip supply while managing costs. Initially, OpenAI considered building its network of chip manufacturing foundries, but has opted against this costly and time-consuming path for now. Instead, the focus will shift towards in-house chip design.

This strategy leverages industry partnerships and incorporates both internal and external approaches, resembling tactics employed by major tech players like Amazon, Meta, Google, and Microsoft. OpenAI’s strategy to source chips from diverse manufacturers while developing its custom chip could shape broader trends in the tech sector.

Following the news, Broadcom’s stock surged over 4.5%, with AMD also seeing a rise. OpenAI, a key consumer of chips, largely relies on Nvidia GPUs for training and inference tasks. However, there is a current emphasis on inference chips due to increasing AI applications.

OpenAI’s collaboration with Broadcom also ensures manufacturing capacity via TSMC for its chip, anticipated to launch by 2026, although timelines may shift. Additionally, with Nvidia having a significant market share, OpenAI’s exploration of alternatives reflects an industry trend amid rising costs and shortages.